Technological Platforms and Diversification

Author(s): Dong-Jae Kim and Bruce Kogut

Source: Organization Science , May - Jun., 1996, Vol. 7, No. 3, Special Issue Part 1 of 2:
Hypercompetition (May - Jun., 1996), pp. 283-301
Published by: INFORMS

Stable URL: https://www.jstor.org/stable/2635091

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Organization
Science

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
 Technological Platforms
and Diversification

Dong-Jae Kim * Bruce Kogut

University of Illinois at Urbana-Champaign, 350 Commerce West, 1206 South Sixth Street,
Champaign, Illinois 61820
The Wharton School, University of Pennsylvania, 2000 SH-DH, Philadelphia, Pennsylvania 19104

Abstract specified to estimate the effect of technological histories on

As the invention of fundamental new sciences spawns subse- subsequent diversification. The results confirm the relation-

quent research, discovery, and commercialization, core tech- ship between relatedness and directionality of technologies

nologies branch into new applications and markets. Some of and the industrial path of diversification. The finding that

them evolve over time into many derived technologies, diversification depends on technological experience and mar-

whereas others are essentially "dead ends." The pattern of ket opportunity has important implications for firms' entry

evolution and branching is called a "technological trajectory." decisions. The authors discuss those implications by describ-

An intriguing question is whether some firms can ride the ing experience as generating options on future opportunities

trajectory by developing proprietary experience in a "plat- and distinguishing between the historical path by which the

form technology." Because the knowledge is proprietary, firms stock of knowledge is accumulated and the path by which

that originate in industrial fields based on a platform tech- new knowledge is generated and commercialized.

nology acquire the technological skills to diversify and to (Platform Technology; Diversification; Technological
mimic the branching of the underlying technological trajec- Trajectory; Knowledge Accumulation; Option)
tory.
The ability to compete in hypercompetitive markets de-
pends on the acquisition of know-how that is applicable to a
wide set of market opportunities. Such capabilities serve as
platforms into quickly evolving markets. To respond rapidly Competing in rapidly evolving industries poses the
to market changes, a firm must have already acquired funda- complex problem of choosing what capabilities should
mental competitive knowledge. In a high-technology industry,
be developed for highly uncertain and volatile markets.
such knowledge invariably is derived from experience with
Freeman (1987) has observed that forecasting the class
the underlying science and related technological fields.
of future technologies has proven to be easier than
The authors examine capabilities as platforms by analyzing
identifying future markets and products. The implica-
the temporal sequence of diversification as contingent on
market opportunities and previous experience. The pattern tion of that seemingly innocuous observation is rather
of diversification of firms reflects the evolutionary branching radical. Developing competence in new but broad-
of underlying technologies. In that sense, the aggregate deci- based technological skills is an investment in a plat-
sions of firms are driven by the technological trajectories form to participate, by a process of expansion and
common across an industrial sector. Certain technologies diversification, in the evolution of future opportunities.
have wider technological and market opportunities, and con- In contrast, forecasting demand for specific products
sequently experience in those technologies serves as a plat- may lead to the development of capabilities poorly
form for expansion. The authors propose that a firm's experi-
suited for the markets that eventually prove to be
ence in platform technologies increases the likelihood of
economically interesting.
diversification when environmental opportunities are favor-
Industries in which competitive advantages of inno-
able.
vations quickly erode are commonly called "Schumpe-
The proposition is tested with the sample of 176 semicon-
ductor startup companies founded between 1977 and 1989. terian" or "hypercompetitive." In such an environment,
Evidence from multidimensional scaling of expert opinion the capability to upgrade products and diversify into
and from an analysis of patent records was gathered to related segments is often built on the accumulation of
identity relatedness among subfields and the evolutionary experiential know-how that allows for expansion during
direction of the technologies. A discrete hazard model is windows of opportunity.1 In that sense, the initial

1047-7039/96/0703/0283/$01.25
Copyright X 1996. Institute for Operations Research
and the Management Sciences ORGANIZATION SCIENCE/VO1. 7, No. 3, May-June 1996 283

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

experiences of firms provide platforms, or options for differences in the frequency, speed, and subfield diver-
diversifying into new markets when the timing is right. sity of patent activity according to the technological
We explain the pattern of diversification of startup origins of the startup firm.
firms in rapidly growing industries as linked to the Longitudinal regression tests were applied to a sam-
evolutionary branching of technologies and market op- ple consisting of the diversification and survival histo-
portunities. A central idea in research on the evolution ries of 176 startup semiconductor firms founded be-
of firm capabilities is that the local search process tween 1977 and 1989. We find strong support for the
builds on previous knowledge that can be transferred idea that particular subfields, as suggested by the tech-
to new fields. Our conceptual contribution is to link nical literature and by the MDS and patent results,
the path of a firm's diversification to the development serve as platforms into other markets. Our discussion
of the underlying technological trajectory. The notion broadens the investigation by considering whether firms
of a trajectory implies that technologies differ in their make entry decisions myopically or with foresight.
potential to generate new paths of development and to
serve as platforms into new markets. Some technolo-
Overview
gies provide a starting point or platform for the explo-
The argument we present consists of three parts:2
ration and development of new resources and have a
1. Technologies follow an observable trajectory.
high potential to spawn new markets. Other technolo-
Some technologies are more likely to serve as plat-
gies are specific to certain applications and represent
forms into new markets than others. The trajectory and
essentially dead ends for further major development.
the platform can be identified empirically.
We predict that firms with experience in platform
2. Organizations differ in their points of origin and
technologies have a significantly greater opportunity
in their accumulation of know-how with different tech-
and tendency to diversify into related subfields than
nologies. Those that gain experience in platform tech-
firms with products narrowly dedicated to a market.
nologies will be more likely to diversify into new
Experience with a platform technology generates the
markets than those with experience in nonplatform
organizational capability to compete in related fields,
technologies.
but it does not guarantee that a firm will, in fact,
3. Organizations tend to diversify into new markets
diversify or succeed. That process is not uniquely driven
when the markets are growing.
by technology, but by firms' efforts to grow in the
presence of competition and the expansion of new
A common approach in studies of diversification is
market opportunities. In that sense, the accumulation
to relate a firm's portfolio of businesses to a measure
of experience provides platforms to enter future but
of relatedness, or to relate a firm's diversification deci-
uncertain markets whereas the timing of diversification
sion to such a measure. We have two wider objectives.
should depend on the growth of the related subfields.
The first is to explain dynamically the path of a firm's
We develop and test the proposition that the pattern
diversification history in terms of its acquisition of
of diversification reflects the branching of the underly-
related knowledge in a particular industry or subfield.
ing technologies contingent on market opportunity. An
The second is to show that the pattern of diversifica-
overview of the literature on technological evolution
tion at the firm level corresponds to a broader techno-
establishes two points: that a trajectory is the cumula-
logical trajectory in which particular technologies act as
tive knowledge resulting from local search, and that
platforms into more specialized applications. A key
the knowledge may be specific and proprietary to firms
element in that explanation is the relationship between
because of experimental learning or system effects. It is
firm-level diversification into distinct markets and
only when the knowledge from the search is propri-
macro technological trajectories. To make that connec-
etary and firm-specific that technological branching
tion, we develop three points: that trajectories are
and diversification should be related. After a brief
defined by both technological and market opportuni-
review of the history of technological branching in the
ties, that knowledge accumulates in response to a firm's
semiconductor industry, we analyze the subfields to
experience in solving particular problems, and that a
determine their "relatedness" and "directionality." An
firm's diversification history is linked to the trajectory
application of multidimensional scaling to question-
because knowledge is proprietary.
naire data shows that experts differentiate subfields of
semiconductors into distinguishable clusters. We then Technological Trajectories
analyze the subfield distribution of patent histories of Potential technological developments can be seen as a
53 startup firms. The analysis shows clear directional transition matrix in which current technologies evolve

284 ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platfonms and Diversification

into new ones. The matrix can be considered the Local Adaptation and Diversifilcation
potential technological space. If there were no logic to A trajectory is tied to the local search efforts of firms.
technological development, the transition would be There is an interesting analogy between the trajectory
essentially random. The notion of a trajectory suggests in technologies and the observed pattern in diversifica-
the contrary: that technological evolution is influenced tion. Diversification is the outcome of temporally or-
by the past. The term "technological opportunity" cap- dered decisions to adapt or acquire knowledge for
tures the idea that not all transitions are equally likely exploitation in new markets. The fact that diversifica-
and that some are favored by the inherent richness, or tion from a given industry is not randomly distributed
evolvability, of a technology. over all other industries suggests that industry experi-
Nelson and Winter (1982: 255-6) make a similar ence guides the evolution of technological and product
observation: market choices of individual firms (Teece et al. 1994).
The dependence of the capability to expand by di-
In many technological histories the new is not just better than
versification on the history of industry experience of
the old; in some sense the new evolves out of the old. One
firms supports a view of evolution by adaptive search.
explanation for this is that the output of today's searches is
Firms are not strictly inert, but diversify and adapt by
not merely a new technology, but also enhances knowledge
learning within the developmental constraints of their
and forms the basis of new building blocks to be used tomor-
row.
current capabilities. An organization is developmen-
tally constrained in its acquisition of both new informa-
The choice of a technology is determined not only by tion and new ways of doing things. Consider, for exam-
current capabilities and technological bottlenecks to be ple, the research on the acquisition of knowledge by
solved. Technologies must also be accepted for use. technology transfer and innovation. A common obser-
The selection process may be determined abstractly by vation in studies on technology transfer has been the
a market, but market forces are outcomes of the insti- importance of prior experience in determining the
tutional environment, including government policies adopter's capabilities to absorb the technology (Teece
(which were influential in the early history of semicon- 1976; Contractor 1981; Pavitt 1987). Similar conclu-
ductors). The historical record supports the importance sions have been reached about the process by which
of the coincidental effects of current technical capabili- knowledge is accumulated and learned (Cohen and
ties and socioeconomic conditions on opportunities for Levinthal 1990; Henderson and Clark 1990; Helfat,
future expansion (Mowery and Rosenberg 1981, Bijker 1994).
and Pinch 1987). History in a particular industry appears to govern
The union of organizational factors that create a the opportunities for diversification for individual firms.
"technological push" and environmental conditions that In industries where technological capability is an im-
create a "market pull" underlies the notion of what portant determinant of competitive advantage and
Nelson and Winter (1977, 1982) and Dosi (1982) have market success, the diversification pattern of a firm is
called "technological trajectories." Such trajectories coupled to its ability to tap into the technological
represent the expansion of a core set of solutions (e.g., trajectory of its industry. In that case, the analogy
mechanization of production) that are favored by the between diversification and technological evolution is
institutional forces that condition their selection and not metaphorical, but reflects a link between the diver-
survival. The revealed trajectory is the conjunction of sification path of the firm and a technology's trajectory.
both technological and market factors, and is deter- Because search is local, it is not surprising that
mined by neither individually. innovations are often recombinations of existing ideas
A trajectory implies that early investments in a tech- and practices (Kogut and Zander 1992). Schumpeter
nology reap increasing returns. One expression of that (1968: 65-66) argued that, in general, innovations are
dynamic is the tendency toward increasing specializa- new combinations of existing "materials and forces."
tion, or divergence. Occasionally, specialized technolo- Usher (1971: 50) called the process of invention a
gies converge in new combinations, which then follow a "cumulative synthesis of many items which were origi-
pattern of specialization. The repeated application of a nally independent."3 The explanation of why innova-
particular set of technologies or organizing principles tions tend to be recombinations lies partly in the do-
eventually exhausts the set of potential combinations main of what individuals can know and partly in the
and market opportunities. Because opportunities are fact that what organizations know and learn is delim-
limited, the expansion of a trajectory is both con- ited by their rigidity in changing their information and
strained and finite. structuring of roles.

ORGANIZATION SCIENCE/VO1. 7, No. 3, May-June 1996 285

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGU'.' Technological Platforms and Diversification

Platform Technologies and Diversification new markets and applications. A platform technology
A trajectory suggests direction in technological devel- represents the development of a capability that maps
opment. It is a simple step to proceed from the notion onto a wide variety of market opportunities, a capabil-
of a trajectory branching into new applications and ity that is consequently characterized by a high degree
markets to an explanation for the diversification path of intertemporal relatedness to a wide expanse of new
of firms. However, one point in the logic must be markets. We call such dynamic relatedness the "direc-
clarified: Why and when should an individual firm tionality" of a technology.
replicate the branching of a trajectory through its own What makes a technology a platform is its formative
diversification pattern? influence on a newly evolving trajectory that is charac-
The notion of an adaptive search is part of the basis terized by increasing returns to investment in further
for understanding the bridge between trajectories and exploration. The increasing returns are achieved either
diversification paths of firms. In general, the diversifi- through network externalities of components in a wider
cation literature, following Penrose (1959), has stressed technological system (e.g., the rotary converter in elec-
the importance of related diversification to perfor- trical grids) or through the accumulation of learning by
mance. In the tradition of the seminal study by Rumelt doing that is useful for the generation of new products.
(1974), that literature has focused primarily on the Those two dynamic factors of network evolution and
cross-sectional relationship between relatedness and learning, separately and jointly, lead to self-organizing
performance. The evidence tends to show that firms and irreversible historical patterns (Arthur 1985, David
enter industries related to their current lines of busi- 1975, Silverberg et al. 1988). When they are propri-
ness (Montgomery and Hariharan 1991). etary, they tie the expanding trajectory to the diversifi-
The findings on relatedness are consistent with a cation paths of firms.
process by which a firm's diversification path is coupled Proprietary learning in design and production is the
to a technological trajectory.4 However, there is a basis for viewing some semiconductors as "technology
difference. A trajectory implies that diversification drivers." As Howell and his coauthors (1988: p. 28)
should follow a broad direction. Relatedness at a point explain: 5
of time is only a cross-section of the dynamic related-
ness across time. The points of divergence and conver-
The significance of learning economies has imparted a
gence in a trajectory represent opportunities for estab-
tremendous strategic importance to certain types of semicon-
lished firms to diversify into new subfields. The
ductors for which significant unit demand exists-and which
creation of those critical junctures often depends on can thus be produced in high volumes-but which are also
the efforts of new firms, but the subsequent develop- highly complex. The mass production of such extremely com-
ment of the resultant subfields tends to show a clear plex devices generates "learning" about complex production
mixture of entry by both entrepreneurial and estab- processes which can be applied to a much wider range of
lished firms (Tushman and Anderson 1986). Related- other device types which are produced by a company in lower
ness does not establish that a given firm needs to span volume. Such high complexity, high volume "technology
different markets. drivers" are widely viewed as a prerequisite to remaining
competitive in semiconductors over the long run. Conse-
However, both relatedness and direction are neces-
quently, these product areas have been the principal areas of
sary to isolate a trajectory. Any temporal series of
international competition in microelectronics since the early
changes in technology or diversification will suggest a
1970s.
path. For a sufficient number of observed changes, a
conservative approach is to test whether the series
follows a random path. Such tests often plagued by low Udayagiri (1993) has shown that learning economies in
power in a statistical sense. An alternative method is to one generation of semiconductors lowers the costs of
identify relatedness of a technology at time 1 to that at subsequent generations. When those learning
time 2 by looking at measures of their relatedness. economies are also useful for application to other
(That approach is equivalent to saying there is system- markets, we should expect the branching of a platform
atic serial correlation in diversification over time.) It is technology into new and related subfields.
the combination of direction with a determination of System dependence and proprietary learning, inde-
relatedness that defines a trajectory empirically. pendently or jointly, tie the opportunity of a firm to the
One way to think about those issues is to distinguish overall success of its trajectory. Only if the platform is
between technologies that are specialized to particular proprietary will technological relatedness be reflected
subfields and those that can serve as platforms into systematically in the diversification pattern of the firm.

286 ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

If the benefit of the externality or learning is entirely startup companies in the semiconductor industry to
public, trajectories will still occur, but there would be record their initial subfield entry and their subsequent
no advantage for firms experienced in the technology timing of diversification. The event of diversification is
to be the engine of growth through diversification. For the dependent variable to be related to a firm's techno-
that reason, proprietary learning in the absence of logical experience.
network externalities commonly serves as a platform
into new markets. In the absence of a learning advan-
tage, network externalities can serve as an entry into Sample and Data
other markets for a firm in an environment where The main source of our data is Dataquest, a market
patent or regulatory protection is provided. The micro- research firm based in San Jose, California. Primary
processor used in personal computers is a classic exam- data and technical details about semiconductors were
ple of a network externality that expanded over time collected through field research. One author visited
with the proliferation of computers and is proprietary three semiconductor plants; the other conducted 50
because of patent protection and the installed base. hours of telephone interviews and served as teaching
We can summarize our discussion in two proposi- assistant for a course on circuitry taught in the depart-
tions. Technologies develop on the basis of current ment of electrical engineering. Presentations of the
capabilities and the influence of market opportunities. research were made to a meeting of industry analysts
Those two forces-the push of current capabilities and and to the planning staff of a large electronics firm.
the pull of environmental signals-shape the direction The initial list of 180 startup firms founded between
of search. By the argument of firm adaptability, we 1977 and 1989 was acquired from Dataquest publica-
would expect the variances in problem-solving experi- tions (1988, 1990). Because startup firms enter with a
ences to lead to differential diversification capabilities restricted product range, the research design has the
conditional on environmental opportunities. We there- obvious advantage of tracking the effect of a firm's
fore suggest the following propositions. initial experience on its subsequent path of diversifica-
tion. The design also has the advantage of concentrat-
P1. Technological Platform: A firm with experience ining on a sample of firms where technology is likely to
a platform technology is more likely to diversify into the be one of the most important considerations; in a
exploration and generation of new markets than a firm wider sample of incumbents, other capabilities such as
that has developed narrowly based skills. marketing skills would certainly require further analy-
sis. Using the Dataquest list, we contacted the startup
P2. Market Timing: The pattem and timing of entry firms and collected complete histories of entry at the
into a subfield are related positively to the growth of the subfield level (i.e., product diversification records). We
market. could not collect complete data for four of the firms,
which were dropped from the sample. Therefore, 176
startups firms form the sample for our study.
The propositions correspond closely to notions of
Table 1 provides a list and description of the primary
hyper- and Schumpeterian competition. In rapidly
subfields of semiconductor devices. The shipment data
evolving industries, competitive capabilities are transi-
for the subfields were retrieved from unpublished or
tory and opportunities are quickly closed by competi-
proprietary records. Digital signal processors (DSP) did
tors. An implication of a trajectory is that the ability to
not have shipment data. Gallium arsenide devices
compete on timing is contingent on the acquisition of
(GaAs) had shipment data only for 1984 through 1988.
the relevant know-how. In environments where future
Both of those subfields were dropped in the final
opportunities are uncertain, there is value in owning
regression analysis.
proprietary know-how that acts as a platform for rapid
In Table 2, the raw ratios of initial and subsequent
expansion.
entry to total entries for the sample firms are given.
Substantial variation is evident, with memory and gal-
lium arsenide devices showing the highest proportion
Data and Methods of initial entrants. Because subsequent entries occur in
To test the claims, we first establish the trajectory of significant proportions only for application-specific in-
technology in a given industry, semiconductors, and tegrated circuits (ASICs), microcomponents, optoelec-
identity the technologies that can be characterized as tronics, and telecommunication, the regression analysis
platforms. We then construct event history records of applies to those subfields.

ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996 287

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Table 1 Technological Subfields-Primary Product Descriptive Overview

Segments The notion of a trajectory can be illustrated by analyz-
ing the recent history of technological development in
Subfield Product Segments
the semiconductor industry. The industry grew out of
the laboratories of AT & T, but the early demand for
Analog Operational amplifiers, comparators, data conversion
the devices was driven by the growing appetite for
products, interface products, voltage regulators, and
sensors military applications (Nelson 1962; Brittain and Free-
ASICs Gate arrays, cell libraries, and progammable logic man 1980). In the United States, the increasing impor-
devices (PLDs) tance of computers gradually shifted supply to the
DSP Single-chip DSP microprocessors, microprogrammable production of digital devices, especially memory; in
devices, special function circuits, and ASIC DSP prod- Japan, the growth of consumer electronics pulled the
ucts. local industry toward specialization in analog semicon-
Discrete Diodes, transistors, power field-effect transistors
ductors (Malerba 1985). The reorientation of Japanese
(FETs), and thyristors
producers to digital production occurred only in the
GaAs Discretes (small-signal transistors and power FETs),
mid-1970s. That example of a successful transition in
optoelectronics (light-emitting devices, detectors, and
-specialization from analog to digital devices also illus-
integrated opto devices), and other applications (ana-
trates the importance of the selection environment.
log and digital)
Memory Dynamic random access memory (DRAM), static ran- The MITI policy of favoring Japanese computer mak-
dom access memory (SRAM), read only memory ers effectively pulled production toward the develop-
(ROM), erasable programmable read only memory ment of digital capabilities (Anchordoguy 1989). An-
(EPROM), and ferro-electric memory other important difference was the role played by
Micro Microprocessors, mass storage, system support, and startup companies in the United States and by estab-
key/ display chip sets lished firms in Japan.
Opto Light emitting devices (LEDs), light sensing devices, The history of technological branching of the indus-
optocouplers, and photodiodes
try can be represented by a dendrogram, as shown in
Telecom Dialers, modem, line interfaces, codec/filter, and
Figure 1. A dendrogram is a model (in contrast to a
switch arrays
theory) that "is a more or less accurate representation
of the path of evolution" (McKelvey 1982: 296).6 Our
Source: Dataquest (1988, 1990).
classification traces the divergence of the basic tech-
nology by market segmentation, as opposed to being a
strict classification by technical trait. It thus resembles
Trajectories in Semiconductors Hannan and Freeman's (1987) pragmatic approach of
Before reporting the regressions, we describe the his- identifying classes of organization by niches. (Subse-
torical trajectory in semiconductors, and then deter- quently, we validate the classification by statistical pro-
mine the pattern of relatedness and evolutionary direc- cedures.) There are obviously other ways to construct
tion among the subfields. That is, even if currently the tree, such as by the electrochemcial properties
there are two subfields, we need to provide evidence of (e.g., MOS) or by manufacturing sophistication (as
an asymmetry: experience in one subfield serves as a used by Schoonhoven et al. 1990).
platform for expansion into the other, but not vice The trunk of the tree is the basic discovery of the
versa. Simple relatedness is not sufficient for under- discrete transistor (and related discrete components)
standing the evolutionary pattern of diversification. and integrated circuit. Integrated circuits developed

Table 2 Ratio of Initial and Subsequent Entry

Analog ASIC DSP Discr GaAs Memory Micro Opto Tele

Initial .65 .46 .43 .75 .95 .83 .43 .36 .21
entry
Subsequent .35 .54 .57 .25 .05 .17 .57 .64 .79
entry

288 ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Figure 1 Evolution of Semiconductors to show that the historical evolution represents a direc-
(discrete) tional tendency that should be mapped onto the diversi-
GaA (IC) fication paths of firms.
opto- We first drew on our interviews and technical litera-
_ _ _ _ _ _ _ electronics

discrete- ture to assemble a set of priors on the relatedness

among the subfields and branches, and then turned to
transistor analog more formal methods to validate those impressions. To
\ / ~~ Tele- establish a stronger claim to directionality, we also
IC < comm
reviewed the patent histories of startup firms that
(digital) memory ASIC
reported patents to analyze the evolutionary pattern of
mwirocompohent
technological diversification in the patent record. The
DSP details of that analysis are given in Appendix 1.

(year) 1950 1960 1970 1980

Summary of Findings on Relatedness and Directionality

Our analysis of the technical literature, expert opin-
along two lines associated with analog and digital en- ions, and patent histories serves as a method of trian-
coding. Another way to distinguish the evolutionary gulation to establish patterns of relatedness at a point
branching is by use of material or, more generally, by of time and directionality over time. Table 3 summa-
tracing the evolution of materials as "co-evolving"with rizes the overall results. As noted in the discussion of
the applications. Early use of germanium gradually was the sample, we are analyzing the effects of previous
dropped in favor of silicon. An alternative material, technological experience on subsequent entries. Our
gallium arsenide, was introduced with slow acceptance. hypotheses are applied to entry into ASICs, microcom-
By the early 1970s, the industry began to divide into ponents, optoelectronics, and telecom ICs, where sub-
new subfields, first memory and then microprocessors. sequent entries occurred in significant numbers. The
The divergence of semiconductor technologies evolved principal finding is that memory is related to several
into the newer subfields of digital signal processors other subfields, both cross-sectionally and dynamically.
(DSPs, used to replace some analog devices in con- Analog experience is expected to be significant in
sumer electronics) and ASICs (used for customized entering telecom ICs. Following the strong suggestion
functions). of the technical literature and expert opinions, we
An especially interesting subfield is telecommunica- hypothesize that experience in discrete and GaAs de-
tions, for which a semiconductor device is a hybrid of vices should help firms enter optoelectronics. Patent
digital and analog technologies. Similarly, the proper- analysis does not clearly capture those relationships,
ties of gallium arsenide have proven to be useful for however. All three analyses suggest strong relatedness
optoelectronics. Those subfields are outcomes of between DSP and microcomponents, but the issue is
"convergent" technologies.7 In evolutionary biology, a directionality. Patent analysis reveals strong direction-
dendrogram (or "cladist" taxonomy) is a strictly hierar-
Table 3 Matrix of Entry Sequence (Predicted)
chical tree whose branches always diverge and never
converge again (Dawkins 1986, p. 258-259). Innova- To ASICs Micro Opto Telecom
tion, in contrast, as the outcome of recombining ele- From
ments, produces both divergent and convergent techno-
logical trajectories. Analog (?)

ASICs .....
A , E i.ES! :!!EESE: !-!... ..... ... ...
... ...... ...-:.;E
Technological Relatedness and Directionality
DSP (?)
Before testing the propositions, we validated the rela-
tionships posited in Figure 1. History provides a kind Discrete (?)
of demonstration proof by showing that a particular GaAs (?)
evolution is feasible. We are interested in a stronger Memory (?) (?) (?)
statement, namely, that the historical patterns repre- M icro . ... -- .
sent a tendency toward increasing specialization stem-
Opto
ming from technologies that embody the greatest op-
portunities. Some technologies have proven to be less Telecom .. . ........ . .

fecund; others have failed entirely. In essence, we want

ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996 289

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

ality from DSP to microcomponents but not vice versa; in shipment volumes. DENSITY, as a measure of com-
hence we hypothesize that relationship. petitive effects, is a count of the number of all firms
The method of triangulation is important because listed as making sales in a subfield. As many studies
each technique analyzes different data, yet reveals sim- have found that entry and density have a quadratic
ilar patterns. The technical literature provides a relationship, we also included the square of density.8
straightforward look at the technological evolution of
semiconductors. The multidimensional scaling relies on Method of Analysis

the opinions of experts about both technological and One of our central claims is that a firm accumulates

market relatedness. (See Appendix 2 for questionnaire.) experience by operating in one subfield that may serve
Finally, the analysis of patents provides insight into as a platform in other subfields. Because we do not

directionality, as well as technological opportunity. have direct observations on the stock of experience, we

Table 3 is a distillation of the findings of the three use a method that proxies it in two ways. First, we
approaches. record by a dummy variable whether a firm is active in

We next test whether the pattern of technological a particular subfield. Second, we analyze the condi-

relatedness and directionality is replicated in the diver- tional waiting time, or hazard, from the time a firm is

sification histories of individual firms. Even though the first recorded as existing to the time of its entry into

innovation of memory semiconductors preceded ASICs another field. We expect firms that are active in an

chronologically, the analysis looks at firms that have industry based on a platform technology to diversify
entered at a time when those subfields were contermi- more rapidly into related fields. Accordingly, we spec-
nous. It is not surprising that subsequent fields should ify the regression as estimating the hazard, or condi-
be shown to be related to historically precedent indus- tional waiting time, to diversification.
tries, but the historical ordering is hardly accidental. A discrete time hazard model was used to calculate
There is no historical necessity for startup firms in our the hazard rate of diversifying among the startup firms
sample to replicate history, other than the fact that (Allison 1984). That method defines the set of startup
there is an experiential relationship between the capa- firms at risk for each annual cross-section or time spell,

bilities gained in the platform technologies and the that is, those firms that had not previously entered a
other subfields. Nevertheless, as reported subse- subfield. Because the period of observation lasted from
quently, advancing the historical clock (that is, begin- 1977 through 1989 and we treated each year as a spell,

ning the observations at a later date so that all sub- there were 13 discrete time spells. The 13 cross-sec-
fields begin in the same calendar year) does not influ- tions were then stacked. That discrete method is simi-
ence the results. lar to continuous-time estimation by partial likelihood,
though the two models do not exactly converge to the
same specification (Cox and Oakes 1984, Allison 1984).
Regression Analysis To test the effects of explanatory variables on the
The regression analysis of the link between the diversi- hazard rate, we used a logit specification. For each
fication path and the direction of the underlying trajec- subfield, a binomial model was used to estimate the
tory is longitudinal and at the firm level, with all effects of the covariates.9 If T is an integer-valued
variables time-varying. The dependent variable is entry random variable showing the waiting time from the
into a subfield at a given time and is set to one if the inception of a startup firm to the diversification event,
firm entered and zero otherwise. The experiential ac- the hazard rate is defined as the probability of an event
quisition of the firm's technological capability is cap- at time t given no previous diversification in the sub-
tured by dummy variables for each subfield, equal to field:

one if the firm was currently active in the area. In

addition, as larger firms might have greater resources
Pt = Pr(.T= tIT t).
than smaller ones, we collected data on the size of the
startup firms, linearly interpolating the estimates for
years for which we lacked data. SIZE is measured by The logit model for estimating the effects of the covari-
the number of employees. ations on the discrete hazard is specified by:
The munificence of the selection environment is
indexed by two variables, density and growth of ship-
ments. The variable growth of a subfield (SHIPMENT pit
GROWTH) is measured by the yearly rate of changes

290 ORGANIZATION SCIENCE/Vol. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

where Pt is 'the probability of the firm diversifying into Table 4 Discrete Time Logit Regression Estimatesa

subfield i at time t, a is a constant, fB is a vector of

ASIC Micro Opto Telecom
coefficients, and X is a vector of time-varying covari-
ates. The changes in the hazard rate are restricted in
Constant - 5.20 -9.90 75.8 -93.13**
this model to arise from changes in the time-varying
(-.65) (-.57) (.61) (-2.32)
covariates; the intercept a is assumed fixed. That
Analog .190 1.36 -15.3 2.77***
restriction can be relaxed, but the inclusion of year
(.26) (1.47) (-.009) (3.82)
dummies shows that the results are not very sensitive ASIC .801 -476 .735
to such a specification. (1.16) (-.40) (1.13)
The appeal of a hazard specification is that it cap- DSP -16.9 -.371 -14.7 1.01
tures the effects of both the technological push and the (-.005) (-.28) (-.006) (.96)
environmental pull. Experience in certain subfields can Discrete -16.6 .095 1.05 .396

be expected to accumulate faster and to be more (-.005) (.06) (.76) (.39)

GaAs .094 .301 2.38** -14.8
closely related than that in other areas, and hence
(.10) (.26) (2.44) (-.007)
should lead to a more rapid rate of diversification.
Memory 1.43*** 3.73*** -.107 1.29*
Further, we would expect the timing of diversification
(2.18) (4.56) (-.09) (1.77)
to depend on the favorability of market conditions. We
Micro -.997 .652 2.40***
capture both effects by use of a hazard model with (-1.17) (.51) (3.26)
time-varying covariates that pick up the differences in Opto -16.4 -13.2 -15.0
experience accumulation and market opportunities (-.005) (-.006) (-.005)
across subfields. Telecom -15.8 -12.4 --14.3
(-.003) (-.003) (-.004)
Regression Results Shipment .026** .033** - .065 .102**

Table 4 reports the discrete time logit regression esti- growth (2.05) (2.45) (-1.25) (2.41)
Density .137 .323 - 5.34 4.92**
mates. Shipment growth significantly increases the like-
(.48) (.38) (-.57) (2.12)
lihood of entry into ASICs, microcomponents, and
Density - .001 - .006 .089 - .070**
telecommunication ICs, but is not significant in opto-
(-.77) (-.62) (.51) (-2.06)
electronics. Density measures are significant only in
Firm size .0007 .003*** -.008 -.006
the telecommunication subfield. Firm size is significant (.52) (3.70) (-1.02) (-.58)
only in microcomponents. The primary factor explain- Log- -67.547 - 58.395 - 35.062 - 46.254
ing the timing of diversification is the growth of the likelihood
market.
Table 5 is a more concise summary of the effects of at-statistics in parentheses; two-tailed test.
competing in one subfield on subsequently diversifica- *prob. < .10, **prob. < .05, ***prob. <.01.
tion. Memory emerges clearly as a technological plat-
form. The effect of experience in memory subfields is
shown to be significant and positive in increasing the
hazard of entry into the ASICs, microcomponents, and
telecommunication IC subfields. Experience with dis- technologies-are significantly related to prior entry in
crete semiconductors, however, does not seem to be a microcomponents and memory (i.e., digital technolo-
significant factor in explaining entry into optoelectron- gies) as well as in analog circuits. As expected, the
ics, whereas gallium arsenide experience proves to be superior optical property of gallium arsenide in com-
significant in entry into that subfield. DSP experience parison with silicon serves as a platform into the opto-
is not significant in entry into microcomponents. Ana- electronics subfield.
log experience is a significant platform for entering the Although all of the subfields existed at the beginning
telecommunication IC market. of our period of observation, certain products, espe-
The relationships between the prior and subsequent cially memory devices, originated much earlier. Some
entry are consistent with the broad distinction between industries only had new firm entries until a few years
analog and digital technologies. There are no crossovers after the beginning our analysis period. To validate the
in those two clusters of technologies. It is especially results, we ran the regressions starting at a later date,
interesting that entries in the subfield of telecommuni- 1983. The results summarized in Table 4 were repli-
cation devices-which embody both digital and analog cated.10

ORGANIZATION SCIENCE/VOL. 7, No. 3, May-June 1996 291

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Table 5 Matrix of Entry Sequence (Actual)a

their own markets against the head start of entering
companies from other subfields.
To ASICs Micro Opto Telecom To test the significance of that claim, we ran a
From partial likelihood analysis on the exit (hazard) rates of
startup firms that began in the four subfields of mem-
Analog ory, microcomponents, telecommunications, and
ASICs ASICs; the covariates were the same as before, but
DSP market growth represented the growth of the initial
subfield. (The results are available on request.) Given
Discrete (+)
the low number of exits, the results are poor. However,
GaAs
the result of interest is that the effect of being begun in
Memory
the ASIC industry significantly raises the hazard of
Micro exiting; beginning in the other subfields has no signifi-
Opto cant effects.
Telecom Successful ASIC firms are in a quandary. Their
success in their initial market cannot be applied easily
at ): hypothesized; [ ]: not hypothesized. for entry into other subfields, and yet entry into their
*prob. < .10, **prob. < .05, ***prob. < .01. market is easy for other firms. They illustrate the
classic dilemma of a competency trap. Not surprisingly,
as indicated in Table 6, acquisition rates of ASIC
startup firms are considerable. Such firms have
nowhere to go, and the acquisition of their subfield
Discussion
Specific Observations
experience is valuable for firms diversifying into the
industry.
Initial entry into the memory subfield serves as a
particularly broad platform by which to enter other
subfields. Yet, no subfield serves as a significant entry General Observations: Myopia versus Foresight
point into memory. In contrast, experience in ASIC The results support the explanation of firm diversifica-
production appears to be a sink, with no indication tion as the evolution of a stock of experiential knowl-
that it can be used as a way to gain entry into other edge that provides a platform by which to enter related
subfields. Industry growth has a significant effect on fields. The relatedness among technologies and mar-
attracting entry of startup companies that began in kets appears to map onto the path of diversification of
other markets. Thus, although the costs of entry into firms. That path is not entirely foreseeable, for it
ASICs appear low, the competitive pressures on star- depends on both the uncertain attractiveness of future
tups are substantial. markets and the success of the firm in its current
That inference is supported in Table 6, where the activities.
exit ratio for the total entrants (i.e., 180 startups) into The concept of technological platforms has signifi-
each subfield is calculated. ASIC firms have two times cant consequences for understanding that what a firm
the exit rate of firms making gallium arsenide devices, does lays the foundation for what it can be. In his
which have the next highest rate. ASIC startups face a analysis of the historical development of nineteenth
difficult task of developing competitive capabilities in century American industries, David (1975, p. 4) simi-

Table 6 Exit Ratio by Subfielda

Analog ASIC DSP Discr GaAs Memory Micro Opto Tele

Dissolution 031 .200 .000 .143 .182 .050 .000 .111 .000
Merger and .063 .133 .063 .000 .000 .100 .091 .000 .000
acquisition
Total .094 .333 .063 .143 .182 .150 .091 .111 .000

aTotal number of exit/total number of entry (fo

292 ORGANIZATION SCIENCE/VO1. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platfonns and Diversification

larly proposes that "choices of technique become the itable technology. Because ASIC production can be
link through which prevailing economic conditions may contracted out so that the capital investment is only in
influence the future dimensions of technological design, entry and exit costs are not high. If diversifica-
knowledge." tion options generated from the initial technology are
The "link" raises two fundamental issues about the low, a high exit rate should be expected. (See Klepper
nature of uncertainty and foresight of organizations. and Graddy (1990) for a formal treatment.)
Rumelt (1984) notes that even in a stringent model The relationship between current profitability and
whereby success is determined initially by some ran- diversification options corresponds closely to D'Aveni's
dom draw, startup firms over time differentiate them- (1994) notion of a hypercompetitive environment in
selves. He writes (p. 565): which know-how and timing are important. In such an
environment, the profitability of a given product quickly
Which activities should the entrepreneur combine? The gen-
evaporates, forcing competitors to be constantly devel-
eral answer is, those that will exhibit strongly dependent postentry
oping and exercising options for product improvements
efficiencies.
or diversification into fields related to the current
Dependent post-entry efficiency, which is synonymous experiential stock or know-how. In the absence of
with path dependence, is the cornerstone of the con- options, hypercompetition results in high exit rates, as
cept of a trajectory and has a subtle implication. Given in the ASIC subfield.
that the investment and entry an organization makes An option approach is a handy way to analyze the
today form its subsequent capabilities, current deci- case of hypercompetition when timing and know-how
sions have what can be called an "option" quality. advantages matter. D'Aveni (1994: 22) notes that "one
Entering the memory market provides a wider set of way to escape [the] cycle of competition on price and
options to enter other subfields than entry into ASICs. quality is to enter a new market or launch a new
(See Appendix 3.) product. Timing of market entry and the know-how
The option consideration runs counter to the main- that allows entry form the second arena for competitive
stream treatment of risk explored in the tradition of interaction."
Cyert and March (1963), in which organizations are In our study, the know-how is the learning by doing
described as creating buffers against uncertainty. It is that is accumulated in the source industry. We can
consistent with the distinction between exploitation think of the value of remaining in that industry as V(s),
and exploration made by March (1991) and Hedlund where s is the cash flows from current and future
and Rolander (1990). An established firm has the choice production. The value of cash flows from diversifying
to exploit its resources in its current activities or to into a new target industry, or introducing new product
explore new markets. Because of heterogeneity in their or service changes, is V(d). Normally, we treat value of
accumulated experiences or in slack resources, firms the firm as V = V(s) + V(d). However, because expe-
may differ in the extent to which they divert resources rience in producing S is useful for producing the target
from current activities to the accumulation and recom- industry product, additivity is no longer obeyed; there
bination of experience. are interactions among the projects, such that V =
An important issue raised by the argument and V(s) + V(d(s). The second term captures what Myers
results of our study, therefore, is whether organizations (1977) called a "growth option." To incorporate the
follow myopic and forward-looking policies, both in the timing dimension, we would need to add time sub-
initial choice of entry and in the subsequent path of scripts."
development. The large number of entrants into the Several subtle issues must be refined, especially in
ASIC market suggests that impediments to entry are relation to the selection environment and experimenta-
low, but the high rate of exit and low rate of diversifi- tion. For example, a forward-looking strategy would
cation suggest that such easy entry is coupled with high suggest that by diverting resources to experimentation,
risk to survival. Profitability may be higher in that an organization gains incremental experience that pro-
industry than in others, but survival rates are less vides a platform to launch into new industries. How-
favorable. ever, although such expansion may enhance a firm's
High entry and exit rates conform with a Lippman long-range opportunities, it also diverts resources and
and Rumelt (1982) model in which post-entry depen- thus may increase the hazard of failure in the short-run.
dence-that is, the option value for subsequent diver- A critical contextual variable, then, is the stringency
sification-is low. In that model, a firm pays an entry of competition, which may limit a firm's ability to divert
fee to see whether it is (randomly) assigned a prof- resources to the accumulation of new learning. To the

ORGANIZATION SCIENCE/VO1. 7, No. 3, May-June 1996 293

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platfonns and Diversification

extent that risk-taking and exploratory search require The finding that experience in memory production
slack resources, organizations that have entered com- has been historically useful for diversification is no
petitive environments are handicapped in their abilities guarantee that it will be in the future. The volume of
to develop the requisite flexibility and diversity of microprocessor production has greatly increased over
resources.'2 The findings of Carroll and Hannan (1989) the past decade. The relatedness among new flows of
on the negative effects of competitive environments on technological progress is no longer the same as the
survival during a firm's early years are consistent with relatedness that drove the accumulation of the stock of
that view. It is not surprising that incumbent semicon- technological knowledge. In principle, a firm can re-
ductor firms have won share in ASICs. Regressions not peat the exact historical path of technological learning
reported here show no effect of the number of com- by focusing on memory production and then branching
petitors or industry growth on exit rates of startup into other (but no longer new) markets. Whether fu-
firms. That finding is not surprising, for the ASIC ture market opportunities and the current set of evolv-
subfield, in which so many exits have occurred, is still ing technologies would be suited to the replication of
growing rapidly. Given the rather high exit rates, the historical lessons is unlikely. The observation is an old
findings suggest that unobserved organizational capa- one: one can enter the river once, but never the same
bilities play an important role in the evolution of new river twice.
subfields.
The radicalness of a technology is specific to firms' Acknowledgements
organizational capabilities to explore new fields given Paul Almeida's research assistance and participation made an impor-

their accumulated knowledge and incentive structure. tant contribution to the article, for which the authors thank him.
They also acknowledge the comments of the anonymous referees,
New technological opportunities often represent
and Erin Anderson, Bill Barnett, Jacques Delacroix, Richard
"organizational discontinuities" to some firms; to oth-
D'Aveni, Drew Harris, Rebecca Henderson, Dan Levinthal, Marvin
ers, they represent an opportunity for diversification
Lieberman, Bill McKelvey, Marshall Meyer, Ron Sanchez, and
and growth. Understanding the relationship between
Gordon Walker on previous drafts and of Sid Winter on the various
technological opportunity and organization capability is mutations. Carolyn Doles and Dataquest are gratefully acknowl-
one of the more intriguing challenges raised by the edged for their cooperation. Financial support for the project was
differential failure and success of startup companies in provided by a grant from AT & T under the auspices of the Reginald
diversifying and growing. H. Jones Center at the Wharton School. The authors thank the
Amos Tuck School of Dartmouth College for awarding the Whitte-

Conclusions more Prize to this article.

A subtle implication of our study is the arbitrariness of

Appendix 1: Triangulation: Relatedness and Directionality
the exact path of technological trajectories. Platform
technologies represent the coincidence of market and
RELATEDNESS: LITERATURE ANALYSIS The initial review of the
technological opportunities. They are likely to be more
literature provides some simple expectations about relatedness in the
formative in their influence during the early history of
industry. We find a very obvious distinction between discrete and
an industry, when markets are yet not well defined. In
integrated circuits and, within integrated circuits, between devices
growing new markets populated by firms with uncer- that rely on digital or analog signals.
tain but differentiated capabilities, a rule that pro- There are a few straightforward implications for the subfields.
motes searching in the environs of current practice is Optoelectronics originated through the combination of optics and
likely to lead to diversification for firms whose tech- electronics technologies. Technically, optoelectronic devices are a
nologies impinge on a wide set of market opportuni- special application of discrete semiconductor technology. Yet, opto-
ties.'3 For firms whose research experience is unre- electronics is a distinct subfield because it provides a significantly

lated to other markets, that rule is self-limiting over different technological solution-by combining optics and electron-

time. ics-to meet new (i.e., light-related) problems in the semiconductor

industry. Similarly, GaAs is considered a distinct subfield given the
As market opportunities are filled, the platform value
potential of the field and the unique demands that the use of the
of a technology diminishes. Because market opportuni-
compound places on design and wafer fabrication. Though GaAs
ties change over time, the inherent opportunities in a
chips are faster, use less power, and have greater radiation hardness
technology are historically conditioned. What has been than silicon, the material is brittle and has low heat capacity. Those
a platform in the past may not be one in the future. properties make the compound difficult to handle and result in low
The fact that history conditions opportunities is a fun- yields. The primary uses are in defense, aerospace, and telecommu-
damental limitation on projecting the past into the nications. Gallium arsenide, because its optical property is superior
future. to that of silicon, has been used heavily in optoelectronic devices.

294 ORGANIZATION SCIENCE/VO1. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Within the category of integrated circuits, analog and digital through networking personal computers, however, seem to be push-
signaling are the two primary branches of the technology. Digital ing the technological boundary of the subfield toward digital areas.
technology can be further divided into memory, microcomponent,
and application-specific integrated circuits (ASICs). The three sub- MULTIDIMENSIONAL SCALING OF EXPERT OPINIONS: RELATEDNESS
fields share a common characteristic in terms of their principal end There is danger of being too glib about the messy issue of the
market, computers. subfield classification that we adopted from industry convention.
There is fairly wide consensus that memory has acted as a Because the claim of shared learning implies a potential symmetry
platform. The preceding quotation on technology drivers ends with among subfields, it suggests the possibility that a platform switch
the conclusion that "at present, the most common 'technology drivers' such as that from DRAM to SRAM within memory could occur
are advanced computer memory devices for which high volume across subfields. Recall that the revealed technological branching
demand exists-dynamic and static random access memories represents the conjunction of technological accumulation and market
(DRAMs and SRAMs) and erasable programmable read only memo- opportunity. It is not surprising, therefore, that once technological
ries (EPROMs)" (Howell et al. 1988, p. 28). Similarly, Borrus et al. learning has accumulated, changes in market opportunity (e.g., a
(1983, p. 214) write that ". . . MOS (metal oxide silicon) memory ICs large demand for microprocessors that generates sufficient experi-
are complex circuits that require technically sophisticated design and ence effects independent of memories) can shift the revealed relat-
production capabilities. Technical sophistication here is transferable edness in diversification.
to the design and production of other complex products...." Such potential switching among subfields implies that the classifi-
Memory, like all of our subfield classifications, is an aggregation cations should be built up from a multidimensional scaling of prod-
of many products. Particular products may be more important drivers uct characteristics. We take a middle ground by maintaining the
than others. Moreover, the function may change over time. In his historical classifications, but validating a structure of relatedness
study on Intel, Burgelman (1993) noted that Intel dropped DRAM among them by a statistical analysis of a distance metric across
production when SRAM was discovered to serve as a substitute for several dimensions. The distances were derived by sampling expert
driving the learning in design and testing for both memory and opinions. A questionnaire was sent to a small number of selected
microprocessor devices. The switch from DRAM to SRAM does not industry experts. Six institutions were chosen on the basis of their
change our memory category as a critical platform for gaining experi- reputation and ease of access: two marketing research firms
ence relevant to other subfields. (Dataquest and Integrated Circuits Engineering), two industry asso-
The importance of memory devices to the microcomponents sub- ciations (Electronics Industry Association and Semiconductor Indus-
field, especially microprocessors, is evident in its origins. (Other try Association), a manufacturing company (AT&T), and an aca-
areas of microcomponents such as mass storage and graphic chip set demic institution (School of Engineering at the University of Penn-
are basically extensions of memory technologies.) The microproces- sylvania). The questionnaire was sent to a person identified through
sor was created when a Japanese company wanted to develop a personal telephone calls. Of the six, five responded (the expert from
semiconductor to provide certain logical functions to support a new the Semiconductor Industry Association did not respond).
range of calculators. It turned to a newly founded company, Intel, We asked the experts to rate technological relatedness between
that had "expertise in the design and manufacture of memory subfields on a scale of 1 (not related) to 5 (highly related). A total of
chips...." (Braun and MacDonald 1982, p. 108). Since then, the 36 combinations of pairs (i.e., 9C2 = 36) were used to estimate
current investment required to produce DRAM has moved beyond technological "distance" between any two subfields. The data were
the resources of new firms, but SRAM have grown in importance then analyzed by multidimensional scaling techniques (MDS), a
and have become a high volume product serving as a driver for many method commonly used to map perceptual (or psychological) dis-
firms (Wolf 1986, p. 619). tance among a number of objects (Green and Tull 1978: chapter 14;
A similar relationship is evident between memory and ASICs Kruskal and Wish 1978; Davies and Coxon 1983).
subfields. For example, standard cells, often called "cell libraries," We ran the MRSCAL (metric scaling) program of the MDS(X)
the most recent product segment of ASICs, are customer-specified package with the data on technological relatedness. MRSCAL is a
combinations of a number of functional cells, from logic gates to metric distance scaling program that positions a set of stimulus
memories and even complete processing units. Memories are almost objects as points in a spatial map. The input data for MRSCAL are
always included as part of the device. the lower triangle, without diagonal, of a square symmetric data
Digital signal processing (DSP) is another subfield in which digital matrix-average (dis)similarity scores of the respondents.
technologies are applied. The fundamental technological problem of The results of the scaling can be represented spatially to show the
the DSP subfield has been real-time implementation of digital tech- hidden structure. The spatial representation consists of a geometric
nologies in miniature (Mitra and Mondal 1987). Technological ad- configuration of points, which correspond to the objects analyzed.
vances in integrated circuits around memory and microprocessors The larger the dissimilarity between two objects, the farther apart
have been a major driver of that new application area (Oppenheim they are in a spatial map.
1975). Figure 2 is the spatial representation of the nine semiconductor
Finally, telecommunication integrated circuits were originally a subfields-the two-dimensional map of technological relatedness.
segment of analog integrated circuits, as the data to be processed in (See Kim (1992) for an analysis of the dimensionality and stress and
telecommunications have been mostly analog (i.e., continuous signal).a Shephard diagram of fit.) The inference from the literature is
Recent applications of telecommunication ICs to office automation clearly confirmed in the scaling derived from the expert opinions.

ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996 295

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Figure 2 Multidimensional Scaling Analysis: Metric Scaling

-l00 -90 -80 -70 -60 -50 -40 -30 -20 -30 0 30 20 30 40 50 60 70 80 90 l00

1.48 4 100

discrete

3. 19 80

0 . 8 9 opto 60
5

GaAs

0.59 40

0.30 20

Dim. 1 analog 9 0

telecom

- 0.3 0 -20

-0 .59 6 -40

DSP memory
-0 .89 7 -60
micro

2
-1.19
ASIC
-80

-1.48 -3.00

There is a distinct but overlapping division between the digital and (the nine technological subfields in the semiconductor industry). The
analog clusters. Memory, ASICs, microcomponents, and DSP form a year of patent application (commonly accepted as the year of innova-
cluster of digital technologies. Telecommunication integrated circuits tion in patent analysis) was available directly from the front page of
are between the digital and analog clusters. Discrete semiconductors the patent, but classification along technological subfields was chal-
are separated from integrated circuits (digital and analog) clusters. lenging. The patent classification system is based on technology and
Optoelectronics and gallium arsenide, as expected, form another contains about 400 primary classes and up to a 100,000 subclasses.
cluster. Such classification of the patents did not facilitate categorization
along the technological subfields we were considering.
PATENT EVIDENCE: DIRECTIONALITY14 The preceding section ex- For the purpose of the study, therefore, we examined individually
amines the relatedness between technological subfields based on the more than 750 patents belonging to the 53 firms. Two electrical
history of subfield evolution. Through the use of patent data, we can engineers read the detailed patent documents and, according to their
analyze not only the relatedness but also the "directionality" of primary field of application, classified each patent along one of the
subfields. That analysis provides further support for our proposi- nine semiconductor subfields. The more general patents and ones
tions: (1) subfields in the semiconductor industry differ in the extent that did not fit neatly into any of the subfields were placed in a
to which they facilitate technological diversification and (2) some special category, "other." Thus, we obtained the complete patent
subfields (namely, memory devices) serve as technological platforms history of the 53 firms across time and across technology.
for entry into other subfields whereas others (such as ASICs) are less The 53 firms were assigned to their subfield of initial market
likely to facilitate diversification. entry. For each category the percentage of total patents in each
The population considered was the 180 startup firms in the subfield and the percentage of firms patenting in each subfield were
semiconductor industry established between 1977 and 1989. We calculated.
examined the patenting histories of those by using the computerized Table 7 is a comparison of patenting activities of firms with initial
patent database available of LEXUS, which contains complete de- experience in different technologies. The firms belonged primarily to
tails (other than technical drawings) of all U.S. patents granted from the memory, ASICs, analog, and GaAs subfields. Discrete, telecom
1974 to date. For the purpose of the analysis, only firms that had ICs, DSP, and microcomponents were represented by only a few
been granted three or more patents by May 1993 were selected. A firms (four or fewer) and no firm in the sample initially belonged to
total of 53 firms fit into that category. optoelectronics.
For each of the 53 firms, we developed a patent profile showing The pattern of patenting activities shows that memory devices
the distribution of the patents across time and across technologies best fit the description of a platform technology. The patents of

296 ORGANIZATION SCIENCE/Vol. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Table 7 Platform Technologies-Evidence from Patents

Subfield of Initial Entry

Memory ASICs Analog GaAs DSP Telecom Discrete Micro

(15 firms) (9 firms) (11 firms) (9 firms) (4 firms) (2 firms) (2 firms) (1 firm)
Subfield of
Granted Patenta % % % % % % % %

Memory - A 24 5 2 11
- B 73 22 11 25
ASICs - A 18 63 1 4 6
- B 40 78 9 11 25
Analog - A 17 1 49 24
- B 27 11 82 44
GaAs -A 36
- B 67
DSP -A 7 12 18 10 37
- B 27 56 36 22 75
Telecom - A 5 5 15 3 100
- B 27 22 27 33 100
Discrete - A 1 67
- B 18 100
Micro -A 3 1 26 100
-B 13 11 50 100
Other - A 26 12 16 22 20 33
-B 73 56 64 78 50 100

Patents per firm 16.7 16.1 16.8 12.2 8.8 13 9 3

Technological
diversification* 2.1 1.8 1.5 2 1.8 0 1 0
Time to
diversification** 2.3 3.7 4.9 2.6 4.3 4

aA and B indicate percentages along columns,

technological subfield (columns should sum to 1
*Mean number of subfields in which patents we
**Mean number of years before first successful patent application in a new subfield.

memory firms on average are diversified across the highest number than that for memory firms.
of fields (as indicated by technological diversification) and those We checked also whether patents preceded or followed market
firms are also the fastest to patent in a field outside their own diversification. In 35 cases, startup firms filed patents in areas other
subfield. The time to diversification is 2.3 years for memory, the least than the original subfield in which they were founded. In 29 of the 35
among all subfields. The patenting activity also suggests memory cases, patents were granted prior to entry into that market (i.e.,
firms are likely to diversify into ASICs and to a lesser extent into technology came first). In only three cases was the patent obtained
analog, DSP, and telecommunications. after market entry (i.e., market came first), and the remaining three
ASIC firms have less diversified patenting records than memory patents were granted in the same year as market entry. Those trends
firms and are slower to patent outside their own field. The only likely indicate that technological knowledge is acquired prior to market
field of diversification (for ASIC firms) as indicated from patent data diversification.
is DSP. GaAs experience can aid entry into analog and DSP is Overall, the findings support the expectation that experience in
related to microcomponents (primarily microprocessors). Patent di- memory increases opportunities for technological diversification over
versification of firms with original experience in the discrete and a wide range of fields. Experience in other fields, such as ASICs, did
microcomponent subfields is narrow. not lead to a broad or rapid pattern of patenting. Whereas memory
Similar to memory firms, analog companies have a diverse patent firms may easily gain technological expertise in ASICs, the converse
history, with patent activity especially high in telecommunications is not true. Such asymmetry supports the claim of directionality in
and DSP fields. However, the patent diversification is less notable the development of a trajectory.

ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996 297

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Appendix 2: Expert Opinion Questionnaire The second feature is the growth of the target industry (e.g., ASICs).
According to the Dataquest, there are nine product subfields in the We can think of growth as proxying a stochastic process that governs
semiconductor industry. (Please see Table 1 for the list of product the price of ASICs. One representation is simply geometric Brown-
segments.) ian motion, whereby
How much are the following subfields technologically related to
each other?
p = p dt + a dz.
Not Strongly
related related
1. analog - ASIC 1 2 3 4 5 If the price of ASICs crosses a particular threshold (P*), entry is
2. analog - DSP 1 2 3 4 5 worthwhile.

3. analog - discrete 1 2 3 4 5 The problem is similar to the analysis of learning curves by Majd
4. analog - GaAs 1 2 3 4 5 and Pindyck (1989). To fit the model, production of memory devices
5. analog - memory 1 2 3 4 5 must depend on a known and certain price. Experience effects would
6. analog - micro 1 2 3 4 5 then be deterministic. Analytical solutions are possible for that case.
7. analog- opto 1 2 3 4 5 A more complicated characterization is to let the price of the
8. analog - telecom 1 2 3 4 5 currently produced product (memories) and that of the target indus-
9. ASIC - DSP 1 2 3 4 5 try (ASICs) be governed by two distinct stochastic processes. (In
10. ASIC - discrete 1 2 3 4 5 reality, they would probably share a common disturbance term.)
11. ASIC- GaAs 1 2 3 4 5 Solving for two stochastic processes is numerically tractable, but
12. ASIC - memory 1 2 3 4 5 more complicated than solving for the single process of Majd and
13. ASIC - micro 1 2 3 4 5 Pindyck. A simplifying solution is to view the firm as resource

14. ASIC - opto 1 2 3 4 5 constrained and consequently faced with the choice between produc-
15. ASIC- telecom 1 2 3 4 5 ing in the source industry or diversifying. In that case, using a single
16. DSP - discrete 1 2 3 4 5 process to describe the ratio of the prices of the two industries can
17.DSP-GaAs 1 2 3 4 5 suffice.

18. DSP - memory 1 2 3 4 5 In our empirical study, we do not analyze the value of the option
19. DSP - micro 1 2 3 4 5 but rather the hitting time distribution, or hazard rate. Leaving aside
20. DSP - opto 1 2 3 4 5 industry and other effects, the hazard model corresponding to a
21. DSP - telecom 1 2 3 4 5 Wiener process is the inverse Gaussian. To include time-varying
22. discrete - GaAs 1 2 3 4 5 covariate effects, we would set either the mean or variance or both to

23. discrete - memory 1 2 3 4 5 be a function of the covariates. We avoid the difficult specification
24. discrete - micro 1 2 3 4 5 issues associated with that parametric approach by working with the
25. discrete - opto 1 2 3 4 5 semiparametric discrete hazard model. Thus, the underlying option
26. discrete - telecom 1 2 3 4 5 model is not transparent, because we work with the hitting time
27. GaAs - memory 1 2 3 4 5 distribution and we do not specify that distribution parametrically.
28. GaAs - micro 1 2 3 4 5 To return to the Majd and Pindyck model, we implicitly follow the
29. GaAs - opto 1 2 3 4 5 case of a single stochastic process for the price of the target industry,
30. GaAs - telecom 1 2 3 4 5 which we proxied by shipment growth. We assumed that experience
31. memory- micro 1 2 3 4 5 accumulates in some way from the time of entry in the source
32. memory- opto 1 2 3 4 5 industry. The experience effect is captured by time-varying dummy
33. memory- telecom 1 2 3 4 5 variables that serve as the state variables. The other principal state
34. micro - opto 1 2 3 4 5 variable is price. When price, proxied by the growth rate (with other
35. micro - telecom 1 2 3 4 5 covariates held constant), crosses some threshold, the firm enters the
36. opto - telecom 1 2 3 4 5 target industry. The other covariate effects (i.e., density) are added
to the reduced-form equation, but are derived (ad hoc) from a
structural model in which price is partly endogenous to market
Appendix 3: Hypercompetition and Growth Options structure.
As noted in the text, the attribution of technological accumulation An important issue, given that characterization, is when the firm
implies that a firm acquires an option to diversify into related fields. should diversify. That question is the same as asking whether the
That option has two features. One is a modified experience effect, hitting time is when the option is exercised. A formal way to
whereby the accumulated experience in one field (i) such as memory understand hypercompetition in terms of timing and know-how is in
lowers the production costs of another field (j), say ASICs. More a dynamic context where difficult-to-imitate assets are accumulated
formally, we can write by experience, and entry is the hitting time when market growth, or
price, hits a critical threshold.
To add more strategic content to the formulation, we could
C(Qj,t) = ce8 - hQir,. consider competitive interactions, as highlighted by D'Aveni, and the

298 ORGANIZATION SCIENCE/Vol. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platfomns and Diversification

choice to allocate resources to learning. Competitive interactions in References

an option context are especially complex, because price becomes Allison, P. D. (1984), Event Histo?y Analysis: Regression for Longitudi-
endogenous. Kulatilaka and Perotti (1994) provide an analysis of the nal Data, Beverly Hills, CA: Sage Publications.
related question of whether to maintain the wait option or early Almeida, P. (1992), "Note on Directionality in the Patent Histories
exercise of entry. Learning is analyzed in Kogut and Kulatilaka of Semiconductor Start-ups," mimeo, Wharton School.
(1994). (Paper available from the authors upon request.)
Anchordoguy, M. (1989), Computers Inc.: Japan's Challenge to IBM,
Cambridge, MA: Harvard University Press.

Endnotes Arthur, W. B. (1985), "Competing Technologies and Lock-in by

See D'Aveni (1994) on hypercompetition and chapters 2 and 4 for a Historical Events: The Dynamics of Allocation under Increasing
discussion of the conversion of know-how to stronghold advantages. Returns," Center for Economic Policy Research, Paper 43, Stan-
See also Eisenhart (1990) on high velocity environments. ford University.

Barley, S. R. (1990), "The Alignment of Technology and Structure

2We thank an anonymous reviewer for consolidating this argument.
3See the operationalization of new knowledge in Schoonhoven et al. through Roles and Networks," Administrative Science Quarterly,
(1990). 35, 61-103.

4Not only technology, but also the applicability of other capabilitiesBijker, W. E. and T. Pinch (1987), "How the Sociology of Science
such as marketing and distribution, may drive diversification to the and Sociology of Technology Benefit Each Other," in W. E.
new industry. Bijker, T. P. Hughes, and T. Pinch (Eds.), The Social Construc-

5We thank Naren Udayagiri for bringing this quote to our attention. tion of Technological Systems, Cambridge, MA: MIT Press.

6See Dawkins (1986: 258ff.) for a discussion and Barley (1990) for Borrus, M., J. E. Millstein, and J. Zysman (1983), "Trade and
another use of a dendrogram. Development in the Semiconductor Industry: Japanese Chal-

7David and Bunn (1990) call the innovations that bridge technologi- lenge and American Response," in J. Zysman and L. Tyson
cal branches "gateway technologies"; Sahal (1981) has called such (Eds.), American Industry in International Competition, Ithaca,

convergence "creative symbiosis." NY: Cornell University Press.

8See, for example, Hannan and Freeman (1987) and Carroll and Braun, E. and S. MacDonald (1982), Revolution in Miniature, 2nd ed.,

Hannan (1989). Theoretical issues are associated with the use of the New York: Cambridge University Press.

density squared measure (Klepper and Graddy 1990), primarily con- Brittain, J. W. and J. H. Freeman (1980), "Organizational Prolifera-
cerning misspecification due to omission of demand effects. We tion and Density Dependence Selection," in The Organizational
include demand effects in our model. Also, density squared is highly Life Cycle, J. R. Kimberly, R. H. Miles et al. (Eds.), San
correlated with concentration, which is an alternative though still Francisco: Jossey-Bass.
problematic measure. The density squared measures does not affect Burgelman, R. (1993), "A Process Model of Internal Corporate
any other estimates and it is included for control and to enhance Venturing in the Diversified Major Firm," Administrative Science
comparability with other studies. Quarterly, 28, 223-244.
9A multinomial model cannot be implemented because it would Carroll, G. R. and M. T. Hannan (1989), "Density Dependence in
entail using a subfield as both a dependent and an explanatory the Evolution of Populations of Newspaper Organizations,"
variable. American Sociological Review, 54, 524-541.
10The only change was a significant relationship between analog and Cohen, W. and D. Levinthal (1990), "Absorptive Capacity: A New
microprocessors. Both of those subfields predate the start of the Perspective on Learning and Innovation," Administrative Science
longer period (used for Table 4) and were not of concern in investi- Quarterly, 35, 128-152.
gating the 1983 truncated sample. Contractor, F. J. (1981), International Technology Licensing, Lexing-
That characterization also suggests a way to define hypercompeti- ton, MA: Lexington Books.
tion. Let us define a hypercompetition index as IHP = V(s, d)/V(s). Cox, D. R. and D. Oakes (1984), Analysis of Survival Data, London:
A hypercompetitive environment is one in which the index tends Chapman and Hall.
toward a high value over time, that is, when current competition has
Cyert, R. M. and J. G. March (1963), A Behavioral Theory of the
eroded profits but the accumulated and proprietary learning is
Firm, Englewood Cliffs, NJ: Prentice Hall.
valuable to entry into new growth markets. There are degenerative
Dataquest (1988), A Decade of Semiconductor Companies, San Jose,
cases when both current operations and future opportunities are
CA.
equally poor because of competition; then it would be more reason-
able to define the numerator as the "representative industry" or to (1990), A Decade of Semiconductor Start-ups, San Jose, CA.

redefine the index as Tobin's Q. D'Aveni, R. (1994), Hypercompetitive: Managing the Dynamics of
Strategic Maneuvering, New York: The Free Press.
12See Singh (1986) for one study showing a positive relationship
between risk-taking and slack resources. David, P. A. (1975), Technical Choice, Innovation, and Economic
Growth, Cambridge, UK: Cambridge University Press.
13See Kauffman's (1993: 76-95) discussion on rugged landscapes with
"generatively entrenched" genes. and J. A. Bunn (1990), "Gateway Technologies and the Evolu-
14This section is based on research by Paul Almeida. See Almeida tionary Dynamics of Network Industries: Lessons from Electric-
(1992) for details. (Paper available from the authors upon request.) ity Supply History," Evolving Technology and Market Structure, A

ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996 299

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Heertje and M. Perlman (Eds.), Ann Arbor: University of Michi- Lippman, S. and R. Rumelt (1982), "Uncertain Imitability: An Anal-
gan Press. ysis of Interfirm Differences in Efficiency under Competition,"
Davies, P. M. and A. P. M. Coxon (1983), MDS(X) User Manual: Bell Journal of Economics, 13, 418-453.
The MDS(X) Series of Multidimensional Scaling Programs, Uni- Majd, S. and R. S. Pindyck (1989), "The Learning Curve and
versity of Edinburgh. Optimal Production under Uncertainty," Rand Journal of Eco-
Dawkins, R. (1986), The Blind Watchmaker, New York: W. W. nomics, 20, 331-343.

Norton. Malerba, F. (1985), The Semiconductor Business: The Economies of

Dosi, G. (1982), "Technological Paradigms and Technological Tra- Rapid Growth and Decline, Madison: University of Wisconsin
jectories," Research Policy, 11, 147-162. Press.

Eisenhardt, K. (1990), "Speed and Strategic Choice: How Managers March, J. G. (1991), "Exploration and Exploitation in Organizational
Accelerate Decision Making," California Management Review, 32 Learning," Organization Science, 2, 71-87.

(3), 39-54. McKelvey, B. (1982), Organizational Systematics, Berkeley: University

Farrel, J. and G. Saloner (1985), "Standardization, Compatibility, of California Press.

and Innovation," Rand Journal of Economics, 16, 70-83. Mitra, S. K. and K. Mondal (1987), "Scanning the Issue," Proceedings
Freeman, C. (1987), Technology Policy and Economic Performance: of the IEEE, 75 (9), 1139-1140.

Lessons from Japan, London: Pinter Publisher. Montgomery, C. and S. Hariharan (1991), "Diversified Expansion by
Green, P. E. and D. S. Tull (1978), Research for Marketing Decisions, Large Established Firms," Journal of Economic Behavior and

4th ed., Englewood Cliffs, NJ: Prentice Hall. Organization, 15, 71-89.

Hannan, M. T. and J. Freeman (1987), "The Ecology of Organiza- Mowery, D. C. and N. Rosenberg (1981), "Technical Change in the
tional Founding: American Labor Unions, 1836-1985," Ameri- Commercial Aircraft Industry, 1925-1975," Technology Forecast-

can Journal of Sociology, 94, 910-943. ing and Social Change, 20, 347-358.

Hedlund, G. and D. Rolander (1990), "Action in Heterachies: New Myers, S. (1977), "Determinants of Corporate Borrowing," Journal of

Approaches to Managing the MNC," in Managing the Global Financial Economics, 5, 147-175.

Firm, C. A. Bartlett, Y. Doz, and G. Hedlund (Eds.), London: Nelson, R. R. (1962), "The Link between Science and Innovation:
Routledge. The Case of the Transistor," National Bureau of Economic

Helfat, C. (1994), "Evolutionary Trajectories in Petroleum Firm Research.

R & D," Management Science, 40, 1720-1747. and S. G. Winter (1977), "In Search of Useful Theory of
Henderson, R. M. and K. B. Clark (1990), "Architectural Innovation: Innovation," Research Policy, 6, 36-76.

The Reconfiguration of Existing Product Technologies and the and (1982), An Evolutionary Theory of Economic Change,
Failure of Established Firms," Administrative Science Quarterly, Cambridge, MA: Harvard University Press.
35, 9-30. Oppenheim, A. V. (1975), "Scanning the Issue," Proceedings of the
Howell, T. R., W. A. Noeliert, J. H. MacLaughlin, and W. A. Wolff IEEE, 63 (4), 548-549.
(1988), The Microelectronics Race: The Impact of Government Pavitt, K. (1987), "International Patterns of Technological Accumu-
Policy on International Competition, Boulder, CO: Westview lation," in Strategies in Global Competition, N. Hood and J. E.
Press. Vahlne (Eds.), London: Croom Helm.
Kauffman, S. A. (1993), The Origins of Order: Self-organization and Penrose, E. (1959), The Theory of the Growth of the Firm, New York:
Selection in Evolution, New York: Oxford University Press. John Wiley & Sons.
Kim, D.-J. (1992), "Technological Accumulation, Sequential Entry, Rumelt, R. (1974), Strategy, Structure, and Economic Performance,
and Exit of Start-up Companies in the Global Semiconductor Boston, MA: Harvard Business School.
Industry," unpublished Ph.D. dissertation, Wharton School, Uni- (1984), "Towards a Strategic Theory of the Firm," in Competi-
versity of Pennsylvania. tive Strategic Management, R. B. Lamb (Ed.), Englewood Cliffs,
Klepper, S. and E. Graddy (1990), "The Evolution of New Industries NJ: Prentice Hall.
and the Determinants of Market Structure," Rand Journal of Sahal, D. (1981), Patterns of Technological Innovation, Reading, MA:
Economics, 21, 27-44. Addison-Wesley.
Kogut, B. and N. Kulatilaka (1994), "What is a Capability? History, Schoonhoven, C. B., K. Eisenhardt, and K. Lyman (1990), "Speeding
Complementarities, and Organizational Discontinuity," mimeo. Products to Market: Waiting Time to First Product Introduction
and U. Zander (1992), "Knowledge of the Firm, Combinative in New Firms," Administrative Science Quarterly, 35, 177-207.
Capabilities, and the Replication of Technology," Organization Schumpeter, J. (1968), The Theory of Economic Development, Cam-
Science, 3 (3), 383-397. bridge, MA: Harvard University Press (first published in 1911).
Kulatilaka, N. and E. Perotti (1994), "Strategic Investment Timing Silverberg, G., G. Dosi, and L. Orsenigo (1988), "Innovation, Diver-
Under Uncertainty," discussion paper no. 145, London School of sity and Diffusion: A Self-Organisation Model," Economic Jour-
Economics. nal, 98, 1032-1054.
Kruskal, J. B. and M. Wish (1978), Multidimensional Scaling, Quanti- Singh, J. V. (1986), "Performance, Slack, and Risk Taking in Organi-
tative Applications in the Social Sciences, 07-011, Newbury Park, zational Decision Making," Academy of Management Journal, 29
CA: Sage Publications. (3), 562-585.

300 ORGANIZATION SCIENCE/Vol. 7, No. 3, May-June 1996

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
DONG-JAE KIM AND BRUCE KOGUT Technological Platforms and Diversification

Teece, D. J. (1976), "The Multinational Corporation and the Resource Udayagiri, N. (1993), "Learning Curves: Cross Product Spillovers
Cost of International Technology Transfer, Cambridge, MA: and Firm Differences," unpublished Ph.D. dissertation, Carlson
Ballinger. School of Management, University of Minnesota.
, R. Rumelt, G. Dosi, and S. Winter (1994), "Understanding User, A. T. (1971), "Technological Change and Capital Formation,"
Corporate Coherence: Theory and Evidence," Journal of Eco- in Economics of Technological Change, N. Rosenberg (Ed.),
nomic Behavior and Organization, 23 (1), 1-30. Harmondsworth: Penguin.
Tushman, M. L. and P. Anderson (1986), "Technological Discontinu-
Wolf, S. (1986), Silicon Processing for the VLSI Era, Vol 2: Process
ities and Organizational Environments," Administrative Science
Integration, Sunset Beach: CA: Lattice Press.
Quarterly, 31, 439-465.

Accepted by Richard D'Aveni, acting as special editor; received March 1994. This paper has been with the authors for two revisions.

ORGANIZATION SCIENCE/VOl. 7, No. 3, May-June 1996 301

This content downloaded from

202.40.157.7 on Sun, 20 Aug 2023 09:32:10 +00:00
All use subject to https://about.jstor.org/terms
Received: 31 May 2021 Revised: 29 October 2022 Accepted: 11 March 2023
DOI: 10.1111/jpim.12681

ORIGINAL ARTICLE

Where digital meets physical innovation: Reverse salients

and the unrealized dreams of 3D printing

Thierry Rayna 1 | Joel West 2,3

1

i³-CRG, Ecole Polytechnique, CNRS,
Institut polytechnique de Paris, Palaiseau, Abstract
France For more than three decades, enthusiasts have predicted that direct
2
Hildegard College, Costa Mesa, manufacturing enabled by 3D printing would inevitably supplant traditional
California, USA
3
manufacturing methods. Alas, for nearly as long, these utopian predictions
Riggs School of Applied Life Sciences,
Keck Graduate Institute, Claremont,
have failed to materialize. One reason is a flawed assumption that hybrid
United States digital-physical systems such as 3D printing would advance as rapidly as
purely digital innovations enabled by Moore's law. Instead, like other exam-
Correspondence


Thierry Rayna, i³-CRG, Ecole ples of cyber-physical systems (CPSs), technological progress in 3D printing
Polytechnique, CNRS, Institut faces inherent limitations that are emblematic of the differences between
polytechnique de Paris, Palaiseau, France.
CPSs and purely digital innovations. As with any complex CPS, improved
Email: [email protected]
performance of a 3D printing system has been limited by that of its key
Special Issue Guest Editors: Michael components—the sort of limiting problem previously defined as a reverse
Stanko and Aric Rindfleisch
salient. Unlike previously studied technologies, several reverse salients for
3D printing performance have neither resolved nor signs of resolving soon.
Here we analyze these key reverse salients, and show how they have ham-
pered the suitability of 3D printing for direct manufacturing and other pre-
dicted applications. We contrast predicted versus actual capabilities for 3D
printing-enabled transformation in six key areas: product innovation, mass
customization, home fabrication, distributed manufacturing, supply chain
optimization and business model innovation. From this, we suggest opportu-
nities for greater realism in future 3D printing research, as well as broader
implications for our understanding of CPSs and reverse salients.

KEYWORDS
3D printing, cyber-physical systems, digital innovation, innovation, reverse salients

1 | INTRODUCTION synthesizer,” a 23rd-century technology that allowed

food to be manufactured on demand for crew members
In September 1966, viewers of the American science light years from home. In 1987, the next Star Trek
fiction series Star Trek were introduced to the “food series featured a voice-controlled “replicator,” a 24th-
century technology that could make almost anything.
Thierry Rayna and Joel West contributed equally to this study. These thus popularized the dream of on-demand

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2023 The Authors. Journal of Product Innovation Management published by Wiley Periodicals LLC on behalf of Product Development & Management Association.

530 wileyonlinelibrary.com/journal/jpim J Prod Innov Manag. 2023;40:530–553.

 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 531

computer-controlled production of arbitrary physical

objects.1 Practitioner points
Also in 1987, in a Los Angeles suburb 20 miles from
Star Trek's Hollywood studio, a new firm named • 3D printing points out the risk of planning
3D Systems, Inc. produced and sold the world's first addi- around utopian predictions of technology
tive manufacturing system. Based on a technology called performance.
stereolithography—patented by founder Chuck Hull— • We show how persistent reverse salients can
these systems could create complex 3D shapes by use of a slow the performance and adoption of a hybrid
computer-controlled laser that solidified liquid resin. digital + physical technology.
DTM Corporation was formed that same year to commer- • We offer a framework for identifying which
cialize selective laser sintering, a UT Austin invention by systems and systems components are most
Carl Deckard that created 3D objects by fusing metal likely to face these persistent performance
powder.2 These two firms thus pioneered what eventually limitations.
became known as 3D printing.
As in their sci-fi archetypes, such products offered
the promise to render arbitrary digital designs into a transformation made possible by the adoption of digital
tangible form. For more than 30 years, these complex computing and communications enabled by large-scale
and expensive systems have been predicted to inevitably circuit integration, rapid improvements in computing
transform manufacturing industries (e.g., Ben-Ner & performance and the malleability of software (Schaller,
Siemsen, 2017; Ghaffar & Mullett, 2021; Li, 2013; 1997; West, 2008).
Peltz, 1987; Petrick & Simpson, 2013). For example, in a We thus consider the impact of the limits of digital
2011 story entitled “Could 3D printing change the and physical technologies that are inherent for digital
world?” a leading U.S. foreign policy think tank breath- systems that serve physical beings who are living in a
lessly predicted: material world. We do so for a broad class of digital-
physical hybrid innovation known as cyber-physical sys-
3D Printing/Additive Manufacturing (AM) is tems (CPSs) that integrate digital and physical compo-
a revolutionary emerging technology that nents in a single system (Lee, 2008; Rajkumar
could up-end the last two centuries of et al., 2010). The study of CPSs often emphasizes adding
approaches to design and manufacturing digital intelligence and connectivity to mature physical
with profound geopolitical, economic, social, technologies, but here we study the case of 3D printing,
demographic, environmental, and security in which digital technologies are combined with imma-
implications (Campbell et al., 2011, p. 1). ture physical technologies.
We begin the article by reviewing how theories of
Two years later, President Barack Obama predicted in his both digital innovation and CPSs explain innovation tra-
2013 State of the Union address that 3D printing “has the jectories for systems that combine digital and physical
potential to revolutionize the way we make almost components. We then consider how these trajectories are
everything.” often limited by a critical technological bottleneck that
Instead, the reality of 3D printing has fallen short of Thomas Hughes (1983, 1987) defines as a reverse salient.
these early visions, as technological performance and This article examines 3D printing as a specific case study
thus adoption have consistently failed to meet predictions for understanding the reverse salients of CPSs. Using more
during these years. Such hype is not unique to 3D print- than 80 articles from scientific journals on digital manufactur-
ing, as many forms of digital innovation have attracted ing and other applications of 3D printing, we trace the pro-
technologically utopian visions to mobilize public sup- gress and limitations of both the digital and physical
port and adoption (Kling, 1996; Turner, 2010). However, components of this technology over the past 35 years, and
purely digital innovations have fulfilled such visions identify the key reverse salients for 3D printing. From this,
during this same period, as witnessed by the global we offer broader implications and research opportunities for
3D printing, CPSs and the study of reverse salients.
1
The first known mention of additive manufacturing in science fiction
came in a 1947 short story in Astounding magazine by Eric Frank
Russell, which referred to “atom fed to atom like brick after brick to
2 | L I T E R A T UR E R E V I E W
build a house” (Nerlich, 2005).
2
3D Systems acquired DTM in 2001, the first of many such acquisitions.
Deckard died in 2019, while Hull remains executive vice president of 3D In recent decades, our world has been transformed by
Systems. ubiquitous digital innovation. One widely cited definition
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
532 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

of such innovation is “the carrying out of new combina- physical, one option is that the physical enables the digi-
tions of digital and physical components to produce novel tal. The most dramatic example of ongoing improve-
products” (Yoo et al., 2010, p. 725). However, subsequent ments in semiconductor manufacturing under Moore's
researchers concluded “What is missing to date is an law have made predictable and rapid improvements in
understanding of how digital transformation manifests computing performance (Moore, 2006; Schaller, 1997)—
itself in industries in which the core products are primar- in turn enabling advances in digital innovation (Autio
ily physical” (Hanelt et al., 2015, p. 1313). We therefore et al., 2021), However, in this article, we focus on the
are interested in this question: How do technological lim- converse case: how the digital enables the physical. An
itations (specifically reverse salients) constrain the tech- examination of digitally enabled technologies of the late
nological trajectory of a system that requires both digital 20th and early 21st century suggests three different pat-
and physical innovations? We also consider what hap- terns for how the digital enables the physical (Table 1):
pens when such limits are either not recognized, or are
assumed to be temporary. 1. Digital represents the physical, where the digital good
We thus review three prior bodies of research that is an intermediate representation of a good that will
relate to digital + physical innovation. The first is how be consumed in physical form. This includes objects
digital innovation considers systems combining digital printed on demand, whether a book or a spare part
and physical components, while the second is research (Jiang et al., 2017; Osterwalder & Pigneur, 2004).
on CPSs. We extend this with prior research on reverse Information goods such as text, pictures and
salients in physical and digital systems to suggest how audiovisual material are a special case, because the
reverse salients could impact the performance trajectory value is the information contained in the object
of a CPS such as 3D printing. rather than the object itself (Negroponte, 1996;
Rayna, 2008).
2. Digital coordinates the physical: digital technologies
2.1 | The promise of integrating digital are used to coordinate activities in the physical world,
and physical innovation whether e-mail, a project plan or an ecosystem map
(Boland et al., 2007; Markus, 1994; Randhawa
Digital technology removes the existing constraints of et al., 2020). For example, to design and make physi-
physical innovation and makes possible a great degree of cal objects, the open-source hardware movement has
redefining and reconfiguring the nature of the innovation used many of the same collaboration processes, social
(Rindfleisch et al., 2017; Yoo et al., 2012). The malleabil- norms and IP rules as those used for producing open
ity of digital technologies thus enables experimentation source software, a purely digital good (Raasch
and improvisation in a way not possible with physical et al., 2009; West & Kuk, 2016).
innovation (Nylén & Holmström, 2015), which has pow-
erful implications for the nature of innovation itself.
First, these reduced constraints—plus the use of a com-
mon digital interface for a wide range of problem T A B L E 1 Taxonomy of digital innovations that depend on the
domains—mean that the pace of innovation has acceler- physical world.
ated (Yoo et al., 2010).
Category Example
Second, digital innovations are often not governed by
Digital represents Print-on-demand 2D objects
the well-known S-curves limiting performance trajecto-
the physical (Osterwalder & Pigneur, 2004) and
ries of physical technologies (Priestly et al., 2020). Even if
3D objects (Jiang et al., 2017);
such S-curves exist for specific domains for key enabling Entertainment and other information
technologies—such as quantum computing (Hill & goods (Negroponte, 1996;
Rothaermel, 2003) or artificial intelligence (AI) (Iansiti & Rayna, 2008); innovation as data
Lakhani, 2020)—advances in general purpose enabling (Rindfleisch et al., 2017)
technologies will accelerate societal innovation more Digital coordinates E-mail (Markus, 1994); boundary objects
broadly than previous S-curves related to product- or the physical (Boland et al., 2007; Randhawa et al.,
industry-specific technologies such as aircraft or wind 2011); open source hardware (Raasch
turbines. et al., 2009)
While digital technologies also impact the nature of Digital controls Manufacturing (Åstebro, 2002) and
physical innovation, that intersection remains a rarely the physical transportation systems (Zhang
explored frontier (Hanelt et al., 2015; Lee & Berente, et al., 2011); robots (Fujita, 2001;
Sturgeon, 2019)
2012; Sturgeon, 2019). When the digital meets the
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 533

3. Digital controls the physical: digital technologies con- mention issues related to digital technologies, regulation
trol activities in the physical world—such as complex and legal issues, or the need to find new business models.
manufacturing systems (Zhang et al., 2011), self- Rarely considered issues are arising from the physical
driving cars (Lyytinen, 2022), and personal or indus- aspects of the CPS itself.
trial robots (Fujita, 2001; Sturgeon, 2019). The focus on the cyber part of a CPS has often been
for the control of updated versions of existing
While all three patterns apply to 3D printing, here we physical objects, such as home appliances, thermostats,
focus on the technological challenges of how the digital automobiles or factory equipment (Borgia, 2014; Hanelt
controls the physical. By its nature, the final results of et al., 2015; Lee et al., 2015). Despite this focus on mature
such this sort of system will be delivered (and con- technologies and equipment, there are examples of CPSs
strained) by the physical processes that are digitally that rely on physical components that are yet immature.
controlled. An emerging example is the so-called metaverse, which
requires affordable, lightweight, durable and high-
definition virtual reality headsets. Another example is 3D
2.2 | Cyber-physical systems printing, which has remained an emerging technology
for decades, and can thus help us understand the role
The most commonly studied examples of innovations similar limitations can have for other CPSs.
where the digital controls the physical are CPSs, first
defined as “integrations of computation with physical
processes” (Lee, 2008, p. 363). The name captures how 2.3 | Reverse salients in physical
the “cyber-world of computations and communications” and CPSs
integrated with “real-world physical activities” (Lien
et al., 2012). A more recent definition states that CPS Innovation in physical technologies often follows a punc-
“comprise interacting digital, analog, physical, and tuated equilibrium model, which assumes a long period
human components engineered for function through of incremental innovation within a given technological
integrated physics and logic” (NIST, 2020). paradigm, interrupted by infrequent discontinuities that
Examples of CPS research have included the Inter- start a new paradigm (Dosi, 1982; Garcia & Calantone,
net of Things (IoT) (Baars et al., 2014; Mahmood & 2002; Nelson & Winter, 1982). Technological progress
Mubarik, 2020) and Industry 4.0 (Greer et al., 2019; within a given paradigm will follow a specific trajectory
Lee et al., 2015). While innovation is often mentioned (Dosi, 1982; Nelson & Winter, 1982) that is only clear
in the academic literature related to CPS, the innova- after the fact. The revealed trajectory depends on a com-
tion processes are rarely studied. Those who do eso bination of market, technical and organizational factors
xplicitly, typically raise the issue of CPS as opportuni- (Bijker & Law, 1994, p. 18; Kim & Kogut, 1996, p. 285).
ties for Open Innovation (Mubarak & Petraite, 2020; Complex innovations are often delivered as systems
Zheng et al., 2019), service and product innovation that integrate a range of components and technologies.
(Li, 2018; Li et al., 2021; Mubarak & Petraite, 2020; While such component reuse offers powerful economies
Zheng et al., 2018), and business model innovation of scale, it provides challenges for the firm both in attain-
(Li et al., 2018; Mahmood & Mubarik, 2020). ing economic success, and also for managing the com-
However, the innovation potential of CPS identified plexity and interdependencies of the various components
in the literature chiefly relates to progress made and (Davies & Brady, 2000; Majidpour, 2016; Prencipe
brought about by the digital side, as with studies of AI, et al., 2003; West, 2006).
IoT or blockchain (Bongo et al., 2020; Li, 2018; Li et al., Within these systems, the technological progress on one
2021). In these studies, the physical aspects of the CPS or more of these components will lag the rest of the system
appear to have been largely overlooked. and thus be the limiting factor of improving overall system
This emphasis on the digital side is prominent in performance. By analogy to a military front, Hughes (1983,
research that bridges theories of digital innovation and 1987) terms such a performance-limiting component a
CPSs, which emphasize the new capabilities and relaxing reverse salient, which Lehenkari & Miettinen (2002, p. 111)
of design constraints made possible by digital control of later summarized as “when one system component—
existing systems (Hendler, 2019; Herterich et al., 2016; technical, organizational or economic—has fallen behind
Lyytinen, 2022; Sandberg et al., 2020). In fact, articles or is out of phase with the others, and impedes the further
investigating either success factors or challenges and bar- development of the system.”
riers to CPS adoption (e.g., Bongo et al., 2020; Leitão Here our interest is in the former case, specifically that
et al., 2020; Sinclair et al., 2019) appear to chiefly subset of technological bottlenecks for components that
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
534 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

limit the performance improvement of the entire system. changes to data exchange interfaces that constrained
Hughes (1983, 1987) suggests that the image of a “bottle- performance.
neck” is a less accurate metaphor for such limits than a Other reverse salients can only be resolved through
salient within a broad technological front. Some have trea- changes to sociotechnical institutions. For example,
ted the overlapping concepts of bottleneck and reverse when compatibility standards both provide interoperabil-
salient as being related issues in technological change ity but constrain innovation, improvements to the shared
(Baldwin, 2015; Windrum et al., 2017) while others have standard can become the limiting factor for improving
used the terms synonymously (e.g., Davies, 1996). performance, such as in digital telecommunications
For Hughes, the solution to such a salient comes from (Dedehayir & Hornsby, 2008; West, 2008). When a
entrepreneurs, inventors and other scientists who define reverse salient is deeply embedded in society, overcoming
“the reverse salient as a set of ‘critical problems,’” and that salient is often more complex to overcome and thus
then work to correct those problems (Hughes, 1983, requires a more comprehensive approach to resolve
p. 80). This definition is fueled by the assumption that (Mathews et al., 2018; Shen, 2019).
“such problems are believed to be solvable. The belief Overall, studies of specific reverse salients have tended
that a reverse salient might be corrected increases dra- to emphasize successfully overcoming the salient. Such
matically by its articulation in a set of critical problems” research includes salients that are “solved” (Christiansen &
(Mulder & Knot, 2001, p. 266). While studies by Hughes Buen, 2002; Windrum et al., 2017) or “resolved” (Sandberg
and many others typically emphasize absolute perfor- et al., 2020) to achieve the system's “potential” (Daim
mance, the critical problem can also be one of efficiency et al., 2014; Dedehayir & Mäkinen, 2011b; Sandberg et al.,
or cost effectiveness, particularly for industries that 2020); this may reflect a long-remarked tendency toward
require economies of scale and scope (Chandler, 1990; success bias in studies of innovation (Coopersmith, 2020;
Davies, 1996). Rogers & Shoemaker, 1971). Thus there is a gap in research
The original theory was developed from Hughes' stud- that considers salients that are not resolved or overcome.
ies of the electrical power infrastructure, but the concept
subsequently demonstrated its relevance in a wide array
of sectors and applications, mainly railroads and rail- 3 | REVERSE SALIENTS IN 3D
ways, aviation, electric vehicles, ICTs, production, tele- PRIN TING
phone, and wireless services (Dedehayir, 2009). In
particular, subsequent studies of physical systems have We are interested in how the digital and physical
examined renewable power generation (Christiansen & components of a CPS—and the interaction between these
Buen, 2002; Hoppmann et al., 2014) and airplane engines components—serve as reverse salients for that system's
(Bowles & Dawson, 1998). While many studies have technological trajectory. We do so in the context of a spe-
focused on a specific salient, a few have examined the cific CPS: 3D printing. Studying 3D printing is an
impact of multiple salients on technological progress extreme case of CPS due to the perfect integration
(Dedehayir & Hornsby, 2008; Sandberg et al., 2020; required between the digital and physical to realize the
Windrum et al., 2017). longstanding promise of a one-to-one correspondence
More recent studies have examined reverse salients of between the digital and physical worlds. Such a “bits to
digital innovation. Since the 1980s, CPU speed has been atoms” solution would fulfill the science fiction dream of
the limiting factor for most general- and special-purpose materializing any digital object as a physical one (Millard
computing systems (Daim et al., 2014; Dedehayir & et al., 2018; Nyman & Sarlin, 2014; Söderberg & Daoud,
Mäkinen, 2011a), while specialized applications have 2012). While originally the best-known subset of a broader
been limited by graphic performance (Dedehayir & category of technologies termed additive manufacturing,
Mäkineif, 2008) or portability (Windrum et al., 2017). In the term 3D printing has since been commonly used to
fact, the PC industry suggests that different types of refer to the entire category (e.g., Wohlers, 2005).
reverse salient may emerge over time, not only between Consistent with other reverse salient studies
different hardware components (Dedehayir & Mäkinen, (e.g., Lehenkari & Miettinen, 2002; Lenfle & Söderlund,
2011a), but also between hardware and software compo- 2022; Sandberg et al., 2020), we identify reverse salients
nents (Dedehayir & Mäkinen, 2011b), making identifying using a longitudinal case study that includes economic
and forecasting reverse salients particularly critical and and technological factors that shaped the evolution of
difficult (Han & Shin, 2014). In the Sandberg et al. (2020) this technology over the past 35 years. Much like these
study of digitalization of previously physical-only process earlier studies, we used a combination of (a) research on
control systems—a rare study of reverse salients in the history of the science and innovation in this field;
CPSs—tight coupling of system components limited (b) scientific literature, in this case from manufacturing
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 535

and materials science journals; (c) economic and business material (such as metal or wood) is cut down to final shape
analyses of 3D printing technology and applications; and using cutting devices such as a saw, lathe or mill. While
(d) comprehensive industry reports—particularly the the CNC process worked with materials that had been
Wohlers Report—about industry statistics and trends. Our used for centuries, as in the pre-CNC era it remained lim-
analysis was informed by previous interviews from our ited by the mechanics of the cutting devices. Initial CNC
respective earlier studies of 3D printing. adoption was also limited by the quality and availability of
The predicted rapid adoption of 3D printing digital Computer Aided-Design (CAD) software, as well as
technologies has failed to materialize due to how they dif- the employee skills necessary to use this software to create
fer from other digital technologies. Indeed, the optimistic digital designs (Åstebro, 2002; Tilbury, 2019).
predictions for 3D printing all implicitly assumed that it Thus nascent additive manufacturing technologies of
would follow the same trends, adoption and use patterns the 1980s could build upon the CAD software developed
as purely digital technologies. But as a CPS, 3D printing for CNC applications. Early 3D printers were first con-
systems require a complex integration of a broad range of trolled by dedicated CAD workstations and later using
both digital and physical technologies. Because improve- personal computers (Jacobs, 1992; Shane, 2000). They ini-
ments in physical component technologies have not kept tially faced the same limitations of CAD software and
up with the digital components, they act as reverse designer skills, but digitally enabled advances in CAD
salients for the overall performance—in terms of speed, software improved the usability of these tools (de Jong &
quality, cost and degree of automation—of how 3D print- de Bruijn, 2013).
ing can create physical artifacts. Table 2 summarizes six
reverse salients of 3D printing, of which five are tied to
the physical side of 3D printing. 3.1.2 | Consumer design tools
First, we examine the digital and physical reverse
salients limiting improvements in 3D printing perfor- The original CAD software was solely intended for profes-
mance, and we then consider the effects of those limits sionals, usually with a drafting or engineering background.
upon the societal and economic transformations Even with improvements in usability for professionals in
predicted to occur through widespread adoption of the early 21st century, further improvements are still
3D printing. needed to make 3D designing as usable for ordinary con-
sumers as 2D design tools (Rayna et al., 2015). For exam-
ple, Digital Forming, a London startup founded in 2008 to
3.1 | Digital components of 3D printing address this problem, promised that “our software har-
nesses the power of mass customization”; instead, the
3.1.1 | Design tools company was declared insolvent in 2019.

To design physical objects, 3D printing build upon one of

the earliest CPSs, Computer Numerical Control. With the 3.1.3 | Rendering layers
invention of the microprocessor, starting in 1972, CNC sys-
tems provided digital control of the mature and reliable Compared to CNC and other subtractive manufacturing,
subtractive model of manufacturing, where a raw block of 3D printing allows the design of arbitrary shapes made by

TABLE 2 Limits and reverse salients in 3D printing.

Category Limitation Current status Salient

Digital Professional design tools Mature
Consumer design tools Slowly improving X
Rendering layers Mature
Distribution of digital designs Mature
Physical Replicability: repeatability and reproducibility Ongoing challenges X
Indirect replicability: digitization Slowly improving X
Costs Slowly improving X
Time Slowly improving X
Skills and knowledge Ongoing challenges X
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
536 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

adding one layer upon top of another. However, it required 3.2.1 | Direct replicability of artifacts
invention of entirely new material science technologies for
fabricating physical objects, as well as digital algorithms to Replicability is a serious and persistent challenge for 3D
translate a 3D digital design into a series of 2D layers, printing systems. Aggregating molecules of a particular
printed one at a time. While the basic layer printing issues material layer by layer never happens the exact same way
were addressed within a decade, repeatability requires twice. These inconsistencies in printing the same digital
developing precision solutions for each new manufacturing object arise even with similar printer technologies, and
material (Gibson et al., 2014; Jacobs, 1992). sometimes with two successive prints from the same
printer and materials (Dowling et al., 2020; Kim
et al., 2018). Efforts to accurately predict the output (a 3D
3.1.4 | Digital distribution print preview) depend on physical phenomena taking
place at molecular level and today require computations
The digital designs for 3D printing are digital artifacts, by a supercomputer (Goh et al., 2020; Sahini et al., 2020).
which by their nature are stored and replicated without a The direct replicability of 3D printing assumed in
loss of information (Quah, 2003; Rayna, 2008). Since the prior forecasts require both repeatability and reproduc-
advent of digitalization in the late 20th century, the stor- ibility (Khorasani et al., 2021; Mussatto et al., 2022; Salmi
age, duplication and eventual dissemination of informa- et al., 2013). Repeatability means creating identical
tion goods has moved from the analog to the digital objects using the same operator, environment, and equip-
and from the physical to the electronic (Rayna & ment (e.g., 3D printer, software, materials) in a short
Striukova, 2021; Shapiro & Varian, 1999). Online com- interval of time. Reproducibility means obtaining the
munities were created to share or sell these 3D printable same results with different operators, materials, and
designs, of which the best known and most often studied equipment. The latter is still a dream, but even the for-
is Thingiverse (Claussen & Halbinger, 2021; Kyriakou mer remains unsolved (Dowling et al., 2020).
et al., 2017; Naik et al., 2021; Stanko & Allen, 2022; A final failure of utopian forecasts for materializing
West & Kuk, 2016). physical objects is that 3D printed objects typically
While such digital designs can be cheaply, reliably require manual post-processing, such as sanding, polish-
and quickly replicated, the same is not true for the physi- ing or even painting; sometimes these steps can only be
cal object 3D printed from such design. In principle, this done by hand (Sgarbossa et al., 2021; Thürer et al., 2021).
distinction applies to the materialization of other digital Some designs also require post-processing assembly.
goods such as a photo or sound recording, but such infor- Beyond the increase in time and cost, this nonautomated
mation goods can be viewed or played on a computer post-processing reduces the consistency of replicating 3D
without being materialized. printed artifacts.

3.2 | How the physical limits digital 3.2.2 | Digitalization and indirect
innovation replicability

Printing arbitrary 3D objects required developing new Whether by CAD design or digitization, digital files used
materials and material science. During the industry's for 3D printing typically only describe the outer shape of
first 15 years, 10 new patent families (by 10 different the object. The internal structure (including wall thick-
inventors) were introduced, each creating a different ness) needed to instantiate the object is generated prior
materials technological trajectory and seeking to over- to each print, depending on the software, the printer, the
come the limitations of other 3D printing technologies; materials and user settings. While solid objects are the
however, all had significant limitations and none cap- norm with traditional fabrication processes such as injec-
tured a dominant market position (West & Kuk, 2016). tion molding and milling, the opposite is true for 3D
The proliferation of technologies, patents and firms con- printing, where printing solid objects is slow, expensive
tinued during the initial decades of the 21st century, and failure prone (Rouf et al., 2022).
with a common core of patents shaping the entire indus- For printing digital objects, indirect replicability
try until 2008 (Wang et al., 2020). Despite this high rate would mean that copies of copies are not only identical
of innovation and new entry, these technologies have to each other, but also to the original (Quah, 2003;
failed to overcome the cost and speed advantages that Rayna, 2008). Unlike digitizing media streams or 2D
high-volume manufacturing provides for most applica- objects, commercial and consumer 3D scanning tools are
tions (Jiang et al., 2017). decades behind 2D counterparts (Conkle et al., 2018):
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 537

widely available 3D scanners can copy the outer layer of technologies. Just as physical limits mean that semicon-
the object, but not its internal structure, leaving unre- ductor manufacturing faces inextricable challenges at the
solved how the object's structure should be replicated nanoscale (Khan et al., 2018), the lack of progress in
(Javaid et al., 2021). While the photograph-scanning 3D printing speed for at least some of the technology
model has been effective for digitizing what objects look branches is due to physical constraints that will be diffi-
like, it will not tell designers how to replicate those cult to resolve, in part due to the unpredictable nature of
aspects of an object's internal structure that determine its the melting/congealing processes required by most 3D
strength, weight and other key properties.3 Even with a printing technologies (Böckin & Tillman, 2019).
3D scan that produces an accurate copy of the original
artifact, the information embedded in this scanned blue-
print would most likely be radically differ from the origi- 3.2.5 | Human skill and knowledge
nal blueprint due to inherently missing information.
Replication of digital objects has been enabled by auto-
mated processes that minimize the required human
3.2.3 | Costs knowledge and skills. Extrapolated to 3D printing, this
brought predictions that consumers would be able to
In addition to high quality, purely digital artifacts can be manufacture objects at home instead of purchasing them
replicated at a negligible cost by reusing digital storage in a store, as soon as 3D printers became widely available
for other purposes. However, 3D printing requires raw (Bogers et al., 2016; Jiang et al., 2017; Rayna &
materials that cannot be reused and are difficult to recy- Striukova, 2015). However, after decades 3D printing still
cle (Ford & Despeisse, 2016). Not only does 3D printing remains a long way from providing fully automated fabrica-
an object have a significant cost, but that cost is tion and “zero skill manufacturing” (Eyers & Potter, 2017).
irrecoverable. Because of the lack of consistency and reliability, the use
Unlike smartphones and other general-purpose com- of 3D printers requires specific craft-like knowledge for both
puting devices, 3D printing also requires specialized up effective design and manufacturing (Despeisse & Minshall,
front capital investments. Not only does 3D printing 2017; Pradel et al., 2018). Different settings are required for
require a 3D printer, but today such printer is only suit- each combination of printer and material—and often for
able for printing a particular type of object with a specific specific objects. As noted in Rouhani (2018, p. 17):
material such as metal, ceramics, or plastic. In contrast to
the multifunctionality of digital devices, the idea of a Mastering FDM 3D printing requires a lot of
single 3D printer that can automatically manufacture tinkering and patience. There are many print
arbitrary objects remains a utopian fantasy for the fore- quality issues common to FDM such as
seeable future. warping, step lines, stringing, under/over
extrusion, layer shifting, and many, many
more. While there are solutions to all of
3.2.4 | Time these problems, it can take multiple attempts
in order to get the desired finished look.
Often overlooked is that the replication of 3D printed FDM machines have a lot of moving parts
objects also requires significant time. Replicating a digital and settings which need to be adjusted
artifact is usually a matter of seconds. 3D printing time according to the geometry of the part, the
depends on the shape, size, geometry and material, but environment the printer is in, and the
typically a small object will require 1–3 h, a medium expected print quality.
sized object 12–48 h, while the largest may take up to a
week (Avinson, 2018). Some improvements have been made—for example,
While print speed for some materials has improved in self-calibrating printers—but 3D printing is unlikely to
recent years, overall speed of 3D printing technologies become “plug and play” in the foreseeable future.
has not followed a similar pattern as other digital Instead, the modest successes of 3D printing have been
enabled by tacit knowledge developed and applied by
3
In principle, the issue can be solved with a medical grade CT or MRI skilled artisans. Just as in 19th century machine tools
machine that can render (and thus replicate) a single item made of
and 20th-century CNC machines, the 21st-century opera-
homogeneous materials such as a replacement bone implant
(Randhawa et al., 2021). However, even this technology is unable to
tion of additive manufacturing systems requires skilled
automatically detect and replicate the structure (let alone materials) of a workers who know their machine and can tell what will
complex assembled product. or will not work. And while digital knowledge is reliably
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
538 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

and costlessly replicated, such tacit knowledge is inher- manufactured by any subtractive (e.g., milling) or
ently neither (Nonaka & Takeuchi, 2007) and is also diffi- transformative (e.g., injection molding) manufacturing
cult to transfer: solving one 3D printing problem does not technology.
directly solve another, although experience provides a
starting point for diagnosing and addressing problems. These factors are tied to the application of 3D printing,
whether building prototypes (Rapid Prototyping), casts,
molds and jigs (Rapid Tooling), end-user parts or objects
3.3 | Effect of 3D printing reverse (Direct Manufacturing), or for Distributed Manufacturing
salients (Rayna & Striukova, 2016).

Due to these limits, after three decades of optimistic

predictions, additive manufacturing has failed to displace 4 | A S S E S S I N G PR E D I C T I O N S
potential niche applications addressed by CNC O F 3 D P R I N T I N G - E N A B LE D
machines—let alone the broader market for high volume TR ANSFORMATIONS
formative manufacturing technologies such as injection
molding or casting. As one frustrated potential user said Using the framework a for understanding the reverse
in 2016: “At this stage, AM machines are a waste of salients of 3D printing developed in Section 3, to reassess
money due to immature technology. These machines extant knowledge and generate new insights for under-
have very high operation and maintenance costs. There- standing the phenomenon we conducted a problematiz-
fore, we are not investing any money on AM technology, ing review of research on 3D printing applications
and the bulk of our income is from CNC machining” (as detailed in Appendix A). From the corpus of 84 jour-
(Wohlers, 2017, p. 170). nal articles shown in Appendix B, we identified six major
Meanwhile, 3D printing failed to realize the promise themes for predicted societal or economic impacts of 3D
of large-scale distributed replication that is today possible printing (Table 3). For each theme, we show how the
for the digital representations but not for the physical impact of 3D printing in that area has been limited by
objects created from those representations—which would one or more of the previously mentioned reverse salients
be necessary achieve predicted transformational out- in 3D printing performance.
comes. This reveals a sharp divide between the physical
side and the digital side, because the overall performance
of the technology is limited by the reverse salients of the 4.1 | Product innovation
physical side.
Given these inherent physical limitations, 3D printing Product innovation is one of the most frequently cited
is only competitive under one or more of the following transformational impact of 3D printing. In practice, this
conditions: has meant Rapid Prototyping (RP), once the only applica-
tion of 3D printing (Hull et al., 1995), and still its most
• Short lead time: despite its comparatively slow speed, prevalent use (Sculpteo, 2020; Wohlers, 2020), because
3D printing has an advantage when time constraints limitations outldeined in Section 3 do not act as reverse
are very tight, as it does not require tools to be built salients for making prototypes.
prior to manufacturing. This applies only when a small RP has had critical benefits for product innovation
number of units are needed, after which the faster pro- (Cooper, 2021; Haleem et al., 2020; Luchs et al., 2015;
duction speed of injection molding will make up for Rayna & Striukova, 2021), especially when done collec-
the time required to build tools (Achillas et al., 2015; tively (Bogers & Horst, 2014). However, these benefits are
Blakey-Milner et al., 2021; Chaudhuri et al., 2021). being eroded by advances in virtual reality (VR), aug-
• Small batches: the lack of tooling cost makes 3D print- mented reality (AR), and digital twin technologies
ing particularly attractive when small numbers of units (Coutts et al., 2019).
are produced. However, the (global) lack of economies Product innovation can also be accelerated by Rapid
of scale for 3D printing compared to traditional Tooling (Fontana et al., 2019; Wohlers, 2020)—printing
manufacturing (Baumers & Holweg, 2019), makes it molds, casts or jigs used in conventional high-volume
less competitive (ceteris paribus) when more than a manufacturing techniques (Radstok, 1999; Rayna &
few dozen to hundred units have to be manufactured Striukova, 2016). This combines advantages of estab-
(Minguella-Canela et al., 2019) lished manufacturing processes (cost, speed, and mate-
• High complexity: 3D printers enable the manufacture rials) and 3D printing (cheaper and more complex molds
of designs of such complexity that they could not be and tools). For instance, 3D printed tire molds enabled
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 539

TABLE 3 Societal impacts of 3D printing predicted by prior research.

Societal impact Prior research Main themes

Product Weller et al. (2015), Stanko (2016), Flath et al. (2017), Disruptive innovation; innovation from data; rapid
innovation Murmura and Bravi (2017), Rindfleisch et al. and iterative prototyping done collectively; real-
(2017), Steenhuis and Pretorius (2017), Kwak time customer feedback; complex tooling to
et al. (2018), Marzi et al. (2018), Candi and enable product innovation; design freedom; 4D
Beltagui (2019), Steenhuis et al. (2020), designs; new product development; no-assembly
Rayna and Striukova (2021b), Ukobitz and products; process innovation; product remixes;
Faullant (2021) technological development; open innovation;
open source
Mass Petrick and Simpson (2013), Rayna et al. (2015), Weller Customer and consumer involvement in design and
customization/ et al. (2015), Bogers et al. (2016), Rayna and Striukova manufacturing; economies of one; knowledge
co-creation (2016), Oettmeier and Hofmann (2016), Steenhuis and reuse; large-scale product customization and
Pretorius (2016), Attaran (2017), Khorram Niaki and co-creation; platforms; pricing; value creation
Nonino (2017), Kyriakou et al. (2017), Aquino et al.
(2018), Shukla et al. (2018), Chaney et al. (2021),
Lacroix et al. (2021)
Prosumption/ Fox (2014), Appleyard (2015), Rayna et al. (2015), Bogers DIY; consumer behavior and adoption; prosumption
home et al. (2016), Kothman and Faber (2016), Rayna and (consumers involved in production process); home
fabrication Striukova (2016), Rogers et al. (2016), Steenhuis and fabrication with 3D printers; intellectual property;
Pretorius (2016), West and Kuk (2016), Wang et al. open design; online design and manufacturing
(2016), Yoo et al. (2016), Despeisse et al. (2017), Flath communities; move from e-commerce to file-
et al. (2017), Jiang et al. (2017), Halassi et al. (2019), commerce; piracy; social manufacturing; user
Kleer and Piller (2019), Naghshineh et al. (2021), Rayna innovation
and Striukova (2021a)
Distributed Garrett (2014), Gress and Kalafsky (2015), Ford and Advanced manufacturing; circular economy;
manufacturing/ Despeisse (2016), Huang et al. (2016), Laplume developing countries; digital encapsulation;
global value et al. (2016), Sasson and Johnson (2016), Despeisse production ecosystems; global factory; industry
chains et al. (2017), Durão et al. (2017), Garmulewicz et al. 4.0; internationalization; value chain
(2018), Hannibal and Knight (2018), Rehnberg and reconfiguration; resource efficiency; product
Ponte (2018), Xu et al. (2018), Kleer and Piller life cycle; geographic span and density;
(2019), Li et al. (2019), Sauerwein et al. (2019), economies of scale and scope; long tail; make-
Beltagui et al. (2020a, 2020b), Hannibal (2020), to-order: manufacturing supercenters;
Rinaldi et al. (2022) organizational structure; restructuring; spare
parts; sustainability
Logistics/supply Baumers et al. (2015), Bogers et al. (2016), Glocalized production; barriers; facility location;
chains Holmström et al. (2016) Sandström (2016), hybrid manufacturing; intellectual property rights;
Oettmeier and Hofmann (2016), Rogers et al. integration; inventories; logistics management;
(2016), Durach et al. (2017), Rylands et al. (2016), SMEs; supply chain innovation; supply chain
Li et al. (2017), Ryan et al. (2017), Schniederjans flexibility supply chain management components
(2017), Chan et al. (2018), Ghadge et al. (2018), and processes; supply chain performance; supply
Eyers et al. (2018), Martinsuo and Luomaranta chain strategy; supply chain reconfiguration;
(2018), Halassi et al. (2019), Luomaranta and responsiveness; remanufacturing; reverse logistics;
Martinsuo (2019), Tziantopoulos et al. (2019), value streams
Zanoni et al. (2019), Delic and Eyers (2020),
Ghobadian et al. (2020), Strong et al. (2020),
Choudhary et al. (2021), Engelseth et al. (2021),
Hecker (2021)
Enabling new Baumers et al. (2015), Prause (2015), Bogers et al. (2016), Business model dynamics; business ecosystems;
business Ford and Despeisse (2016), Rayna and Striukova (2016), capabilities; digital entrepreneurship; disruptive
models Rogers et al. (2016), Steenhuis and Pretorius (2016), innovation; industrial sustainability; nascent
Holzmann et al. (2017), Khorram Niaki and Nonino industries; platforms; SMEs; technology adoption;
(2017), Kwak et al. (2018), Rong et al. (2018), Xu et al. user entrepreneurship; value creation; value
(2018), Holzmann et al. (2020b), Rayna and Striukova capture; value chain reconfiguration
(2021b)

Note: All references shown in this table are provided in Appendix B.

 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
540 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Michelin to develop its CrossClimate/All Season product technologies (Rayna & Striukova, 2021), this also largely
line (Molitch-Hou, 2017). failed to materialize. Despite an early enthusiasm in the
However, this prevents reaping the full benefits of 3D 2010s, consumer 3D printing faces multiple reverse
printing for product innovation. As emphasized explicitly salients—unreliability, relative high cost, post-processing,
(Rayna & Striukova, 2021) or implicitly (Bogers skills and competencies—that have hindered professional
et al., 2016; Ford & Despeisse, 2016; Kleer & Piler, 2019; direct manufacturing applications. Many experts consider
Rindfleisch et al., 2017) in the literature, the benefits only that, aside from small maker (hobbyist) communities
happen when actual products (or a significant part of (Mavri et al., 2021), consumer 3D printing was indeed a
them) are directly manufactured (DM) with 3D printers. fad (Blueprint, 2020; Wohlers, 2020).
This provides digital freedom for physical products: pre- Consumer 3D printing uptake was a key requirement
assembled designs (Haleem et al., 2020), 4D printed prod- for the predicted emergence of transformational con-
ucts (Lee et al., 2017), or products with complex internal sumer communities that would jointly innovate, remix
structure that weight a fraction of usual products (Feng and tinker, share and distribute products and artifacts
et al., 2018; Wohlers, 2020). (Bogers et al., 2016; Jiang et al., 2017; Rayna et al., 2015;
Unfortunately, the reverse salients are also most preva- Rayna & Striukova, 2016; Rayna & Striukova, 2021), as
lent with DM: risk of low replicability, more complex (and has happened with other digital technologies. Such com-
costly) the post-processing (because of the complex struc- munities have emerged, but have had a limited impact,
ture), and requiring new skills (Rylands et al., 2016; since so few people have access to a 3D printer. The
Thomas-Seale et al., 2018). Consequently, DM has majority of 3D printing platforms identified by Rayna
remained confined to a few niche markets—medical/dental et al. (2015) have disappeared or become 3D printing ser-
prosthetics, aerospace and defense (Sculpteo, 2020; vice bureaus (e.g., 3DHub, Sculpteo, i.Materialize). Even
Wohlers, 2020). Other envisaged applications have largely on the popular Thingiverse online community, objects
been abandoned, preventing a key transformation of prod- regularly appear but the top 20 most fabricated ones are
uct innovation predicted by prior research: the real-time more than 5 years old, and contrary to expectations
integration of customer feedback to continuously improve (Stanko & Allen, 2022), have led to few remixes and are
products (Rayna & Striukova, 2016; Rayna & Striukova, seldom remixes themselves.
2021; Walsh et al., 2017; West & Bogers, 2014). Likewise, the seminal RepRap Open-Source Hardware
project that gave rise to MakerBot, Ultimaker, and dozens
of other startups (de Bruijn, 2010; Greul et al., 2018; Hu &
4.2 | Consumer involvement: Sun, 2022; Stanko, 2020; Stanko & Allen, 2022; West &
Customization, co-creation, and Kuk, 2016) has proved unable to move beyond (the rela-
prosumption tively unsophisticated) plastic filament printing technol-
ogy. Limited consumer adoption has dashed widespread
The literature also widely predicted the involvement of hopes firms might leverage 3D printing user communities
consumers and other customers in the production pro- (de Jong & de Bruijn, 2013; Rayna et al., 2015; Rayna &
cess, either in design (mass customization, co-creation) Striukova, 2016; Rindfleisch et al., 2017).
or manufacturing (prosumption, home fabrication). Also unrealized is a predicted shift from e-commerce
Because of higher costs, mass customization via DM to “file-commerce” (products purchased online to be
requires a higher willingness to pay for customized prod- printed at home) and 3D printing that would dramati-
ucts (Franke & Piller, 2004). Yet, actual willingness to cally change how products are developed, distributed,
pay appear to vary greatly (Kalantari et al., 2021), so the and acquired (Jiang et al., 2017). Consequently, the once
reverse salient for DM costs make mass customization predicted IP crisis from consumer piracy of objects
mainly viable in the medical sector (Wohlers, 2020). (Birtchnell et al., 2018; Jiang et al., 2017; Mendis
Mass customization also requires specific skills for et al., 2019; Rayna & Striukova, 2016), has also failed to
both designers and users. After failed efforts to provide materialize.
easy-to-use interfaces, mass customization has generally
been restricted to pseudo-customization (choice within a
range of pre-set attributes) where 3D printing faces 4.3 | Distributed manufacturing and
cost and time disadvantages compared to regular global value chains
manufacturing.
A more radical predicted impact was that consumers Unlike other manufacturing technologies with significant
would manufacture products with their own 3D printers. economies of scale, 3D printing is characterized by a con-
While in line with the success of other digital stant unit cost, which provides far fewer reasons to
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 541

concentrate production geographically or temporally. component are key benefits of 3D printing identified in
Thus, a key impact predicted by the literature is a shift the literature (Beltagui et al., 2020a); a famous example is
toward distributed manufacturing and, consequently, a the claim made by GE to have simplified a jet engine noz-
redefinition of global value chains. zle from 20 parts to one, and, in another case 855 parts to
Digital encapsulation (Holmström et al., 2016) just 12 3D printed parts (GE, 2020). On-demand produc-
through home 3D printing (Despeisse & Minshall, 2017; tion can reduce costs through reduced waste, as well as
Jiang et al., 2017; Rayna & Striukova, 2016) would be the lower assembly, storage and transportation costs, hereby
most extreme form of distributed manufacturing, but is saving fuel (only raw materials transported) and reducing
impractical without widespread consumer 3D printing supply chain and logistics costs (since fewer parts are
uptake. Even RepRap printers, originally designed to be involved) (Ford & Despeisse, 2016).
self-replicable—by having a 3D printer make many of the Literature has hypothesized benefits of on-demand
next printer's parts (Rayna et al., 2023; West & production for sustainability (Ford & Despeisse, 2016;
Kuk, 2016)—are generally purchased as mass-produced Gebler et al., 2014; Rouf et al., 2022), chiefly focusing on
kits or fully assembled printers. spare part production. Yet, the use of 3D printing to pro-
Distributed manufacturing is theoretically possible: duce spare parts remains rate and claims of superiority of
firms could share production facilities and use service 3D printing for spare part manufacturing (Li et al., 2017)
bureaus, while large companies could internally distrib- appear to be limited to specific niches (Frandsen
ute manufacturing across different locations, which et al., 2020) where reverse salients are not as relevant;
would improve sustainability by reducing transport even with production costs four times higher, less than
costs (Ford & Despeisse, 2016; Gebler et al., 2014). Yet, 10% of spare parts are worth 3D printing (Heinen &
we were unable to find in any significant application of Hoberg, 2019). Post-processing further worsens the 3D
this concept, beside the well-publicized (but brief) pro- printing cost disadvantage (Heinen & Hoberg, 2019),
duction of medical equipment during the first months while the environmental improvement over traditional
of the COVID-19 pandemic (Budinoff et al., 2021; manufacturing appears moderate to negligible (Böckin &
Frazer, 2022). Tillman, 2019).
Recent literature suggests that even in best-case scenar- Lack of tooling (Westerweel et al., 2018) is predicted
ios, the alleged superiority of distributed manufacturing to enable shorter lead-times, but this has little impact
may not stand scrutiny. Li et al. (2019) found that, contrary overall (Knofius et al., 2019), because of the reverse
to the existing perception, distributed manufacturing does salient (comparatively slow speed of and higher
not always guarantee a quick response, and that pooling manufacturing cost) makes that, unless unique parts are
effect makes that centralized manufacturing performs bet- involved. Otherwise, 3D printing's lead-time advantage
ter when the demand rate is relatively high. Furthermore, rapidly wears off (Muir & Haddud, 2017).
Rinaldi et al. (2022) found the benefits of distributed In regard to logistics and supply chains, 3D printing's
manufacturing rather limited even when average demand ability to manufacture shapes that embed several parts
is rather low, as centralized manufacturing provides a bet- into one (consolidation) has also been emphasized. Yet,
ter capacity utilization. this often leads to a higher total cost (Knofius et al.,
This contrast between predictions and the reality can 2019)—consolidation worsens the impacts of wear or fail-
be largely attributed to the reverse salients. When physi- ure, while greater design complexity means higher
cal properties (structural or aesthetical) of a product post-processing costs (Knofius et al., 2019; Sgarbossa
matter, distributed manufacturing requires a very tight et al., 2021).
control of the manufacturing process, which makes 3D Overall, the benefits of 3D printing on supply chain
printing a QA nightmare. In a world where transport and and logistics appear chiefly limited to niches where the
logistics are rapid, efficient, and cheap, the (mainly) cen- reverse salients do not matter as much: inherently low
tralized manufacturing model is hard to beat. The physi- volumes, no post-processing, and parts that do not wear
cal limitations of 3D printing stand in the way of the out, for example, highly specialized low-volume applica-
long-predicted global value chain reconfiguration. tions, such as industrial equipment, manufacturing
at sea, or remote supply depots for combat ships
(Kenney, 2013)—or in space, as demonstrated in 2014 by
4.4 | Logistics and supply chains NASA using an FDM 3D printer to produce spare parts
on the International Space Station (Prater et al., 2016), in
On-demand production and ability of manufacture com- which case the cost of getting cargo to outer space out-
plex designs to consolidate several parts into a single weigh the limitations of 3D printing.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
542 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

4.5 | Enabling business model 4.6 | Reevaluating societal impacts of 3D

innovation printing

Beside investigating business models for the 3D printing The literature shows a prevalent belief that 3D printing
industry (e.g., Holzmann et al., 2020), literature has pro- represents the next stage of digital innovation, blurring the
jected that 3D printing would radically reconfigure busi- boundaries between digital and physical, hereby opening
ness models in physical sectors, just like what has an entire new realm of innovation (Aryan et al., 2021;
happened in the digital realm, whether by choice or Nasiri et al., 2020; Rindfleisch et al., 2017). This brought
necessity due to consumer piracy of objects (Appleyard, predictions that the physical world would witness a free-
2015; Birtchnell et al., 2018; Depoorter, 2014; Mendis dom of creation similar to the digital one, enabling new
et al., 2019). 3D printing would not only change business modes of digital innovation that would “digitize physical
models—leading to more consumer-centric business goods much like the first digital revolution digitized infor-
models involving consumers in the manufacturing supply mation goods” (Rindfleisch et al., 2017, p. 681) and trans-
chains (Bogers et al., 2016), with, over time, a prolifera- form product development and innovation, value chain,
tion of “business models that involve increased levels inventory management, customization and business model
of customer self-service” (Bogers et al., 2016, p. 12), innovation.
but business model innovation in itself (Rayna & Aside from very specific niches, the transformative
Striukova, 2016). impacts of 3D printing identified in the literature have
Yet, beside occasional business model upgrades, it is largely failed to materialize, because of the physical limi-
hard to find cases of radical business model reconfigura- tations of 3D printing that act as reverse salients. These
tion through 3D printing. The RepRap Open-Source physical reverse salients have prevented a widespread
Hardware project has enabled shifts between open adoption of 3D printing by firms and individuals—a pre-
and closed business models (Stanko, 2020; West & requisite to most of transformative impact predictions, as
Kuk, 2016), as well as business models blending these it would enable firms to gather early customer product
two modes (Appleyard & Chesbrough, 2017; Rayna et al., feedback (West & Bogers, 2014), harvest data through 3D
2023). Even in this case, only startups have carried out printing to better understand user needs (Rindfleisch
bold business model innovation (Greul et al., 2018), with et al., 2017) and leverage users and user communities to
incumbent perpetuating essentially the same business innovate derivatively through 3D printing (de Jong & de
model. Bruijn, 2013; Rindfleisch et al., 2017). Such adoption
Expectations of transformative impact on business would lead to a potential large-scale shift to online reposi-
models were largely based on the premise of a large cus- tories for product design (Jiang et al., 2017) and bypass
tomer adoption, and involved mass customization, dis- traditional manufacturing via digital encapsulation
tributed manufacturing and reconfigured global supply (Holmström et al., 2016). In turn, this would force rethink-
chains, or on-demand production. Because of the ing of the role of brands (Rindfleisch & O'Hern, 2015;
reverse salients, all have been confined to narrow Wang et al., 2016). Widespread consumer adoption would
niches. It is not hard to understand why the predicted lead to the rise of intermediaries and platforms that would,
radical business model shifts have not materialized. The on the positive side, facilitate innovation and, on the nega-
related literature generally ignores the reverse salients tive side, bring a predicted IP doomsday when IP protec-
and instead emphasizes the benefits brought about by tion of physical goods would suddenly become as
the digital side of 3D printing, with the implicit assump- challenging as for other digital goods (Jiang et al., 2017).
tion that 3D printed artifacts behave just like purely dig- As industry experts now admit (Wohlers, 2020), wide-
ital artifacts. spread adoption is unlikely to arise until reverse salients
Even in cases when 3D printing was largely adopted, have been resolved: with decades of hindsight, predic-
business models may largely remain the same. The main- tions and expectations were simply unrealistic. Conse-
stream uptake of 3D printing in the hearing aid industry quently, discussions surrounding what is the correct
in the 1990s has still not brought radical business model business model for products marketed through 3D
reconfiguration (Sandström, 2015). Despite widespread printing—such as pricing per print, subscription or open
adoption, the very same networks of firms are involved access model (Bogers et al., 2016)—remain largely
in the value creation, with no radical shift in value net- irrelevant. Even predictions that did not require a large
works, such as consumer involvement or other changes consumer adoption have largely remained niche applica-
described in Bogers et al. (2016) and Rayna and Striukova tions, such as 3D printing mass customized medical
(2016). implants (Randhawa et al., 2020).
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 543

The same reverse salients have also impeded the pre- materialize due to persistent limitations of 3D printing
dicted expansion of distributed manufacturing, and the performance. The traditional response in prior research
advent of dispersed 3D printing micro-production centers has been that those limitations would soon be solved:
located close to customers (Luchs et al., 2015), which, in
turn, would have distributed innovation beyond a single By [comparing] direct digital manufacturing
firm. Consequently, questions of best practices for collabo- with traditional tool-based manufacturing, we
rative innovation with distributed production—particularly show that direct digital manufacturing so far
the role of digital manufacturing in B2B collaboration lags behind by several orders of magnitude
between established firms and startups and SMEs, also have compared to traditional manufacturing
been pushed into the future, except for niches such as methods. Yet we also find that direct digital
defense, aerospace and industrial equipment. manufacturing clearly is on an improvement
This consistent failure to meet expectations has led to a trajectory that eventually will see it being able
reevaluation some of the most optimistic predictions of 3D to compete with traditional manufacturing on
printing use. This can be seen in the content of the influen- a unit cost basis (Holmström et al., 2016, p. 1).
tial Wohlers Report—a yearly report by industry experts
and key source of information about the 3D printing As we have shown, the manufacturing (i.e., physical)
industry. Between the Wohlers (2012) and Wohlers (2020) aspects of 3D printing have been the reverse salient for its
annual reports, most (9 of 17) of the predicted of 3D print- use and adoption for three decades: while closing this “sev-
ing applications—avatars and figurines; consumer-created eral orders of magnitude” gap has long been a priority, on
products; fashion products; furniture, home, and office the current trajectory the eventual ability to compete with
accessories; gifts, trophies, and memorials; marketing and traditional manufacturing remains decades away.
advertising; museum displays; musical instruments; and For future research, refractory reverse salients of 3D
personal 3D printers—were dropped. Applications com- printing require a new research approach. Rather than
mon across both periods are essentially those already pre- assume an imminent (but still unfulfilled) “hockey-stick”
dicted 25 years ago: aerospace, architectural modeling, art inflection in performance or adoption, researchers should
and jewelry, automotive, machinery, medical applications, instead focus on what is known to be technologically
prototyping, tooling. feasible. By doing so, they will produce more realistic
Industry experts appear to have finally realized that (and relevant) analyses that considers the impact of these
the physical reverse salients of 3D printing were not tem- limitations, and how persistent reverse salients can be cir-
porary impediments, but were instead intrinsic to 3D cumvented with more attainable technological solutions.
printing: building objects layer by layer usually requires To ground research in reality, two different paths lie
repeated but unique reactions at molecular level, leading ahead for 3D printing researchers. The first requires revi-
to complexity, unpredictability, comparatively high siting 3D printing predictions for each of the six themes
energy usage and slow speed. In fact, it can be argued of Table 3, and, accounting for the actual limitations,
that the very limitations of 3D printing lie in its very investigate which predictions remain realistic. In particu-
strength. As a result, these reverse salients are unlikely to lar, this means moving away from the focus on direct
be resolved any time soon. manufacturing, which will remain a niche use.
Mainstream uses of 3D printing remain prototyping
and tooling, so an immediate research opportunity is
5 | FUTURE RESEARCH how product design best practices can leverage both. This
is particularly important at a time when virtual prototyp-
Given the enduring reverse salients of 3D printing, how ing technologies—such as augmented or virtual reality—
does that change future research? Here we consider the are rapidly increasing in performance and use. These
implications of this for the future of digital manufactur- newer technologies may undercut prior 3D printing suc-
ing research, and more broadly for our understanding of cesses, as digital substitutes evolve more quickly than
digital innovation, CPSs, and reverse salients, concluding physical-limited processes. Likewise, fast moving high-
with managerial implications of this research. flexibility robotic automation may reduce the appeal of
rapid tooling.
A second research path would study how other digital
5.1 | Implications for 3D printing and technologies might help overcome the physical limita-
digital manufacturing tions of 3D printing. For instance, combining 3D printing
with other digital manufacturing technologies (e.g., CNC,
As analyzed in Section 4, research on 3D printing has laser cutting) could alleviate the speed and post-
brought high expectations that have repeatedly failed to processing issues that plague 3D printing, while using AI
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
544 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

in design may diminish the complexity and time needed • Business model innovation: Besides direct manufactur-
to customize 3D printed items. Augmented reality could ing and home fabrication, could rapid prototyping and
finally provide the intuitive and easy-to-use user interface tooling enable reconfigured or entirely new business
(equivalent the mouse and windows for PCs) that is models? Are there disruptive (Christensen, 1997) busi-
needed to fully leverage 3D printing technologies. ness models for direct or local manufacturing
Overall, both research directions are examples of how analogous to how lower quality MP3 music disrupted
to reduce the impact of persistent reverse salients on music consumption in the 1990s?
improving the performance and economic impact of 3D
printing. Future research and commercial applications
cannot ignore these reverse salients, but instead must 5.2 | Implications for CPSs and digital
consider those trajectories that either minimize or over- innovation
come their impact.
Considering the main themes identified in the litera- Considering the differences of the digital and physical
ture (Table 3), opportunities for further research include: models suggests that we need a more nuanced model of
innovation and technological change in today's digital
• Product innovation: Assuming direct manufacturing world, particularly with regard to reverse salients. In
will remain a niche application, how can 3D printed Table 4, we compare the rate of innovation and reverse
tooling enable complex molds and gigs that can be salients for all possible combinations of digital and physi-
used as a source of product innovation? How can com- cal innovations: the two extremes of purely digital or
bining 3D printing with virtual prototyping technolo- physical systems, and four possible configurations of
gies be leveraged to facilitate user feedback in near CPSs. We contrast the emerging technologies (where dra-
real-time? matic performance improvements are still possible) with
• Consumer involvement: If consumer adoption of 3D mature technologies that have approached the limits of
printing remains low, how can combining these tech- their technological trajectory.
nologies enable large-scale co-creation with con- The Purely Digital configuration corresponds to the
sumers? Could digital technologies supplant 3D digital innovation research of the past decade. In most
printed physical prototypes, thereby enabling online cases, the malleability and rapidly improving perfor-
communities to utilize digital twins that provide a true mance of digital systems have enabled dramatic progress
correspondence between the digital artifacts and their in such systems. A notable exception is artificial intelli-
physical materialization? Could mass-customization gence, where the rate of progress has been “patchy and
serve intermediate-sized global markets with similar unpredictable” (Stone et al., 2016, p. 4).
tastes, desires, or needs with 3D printing be used to Among CPSs, we identify four configurations.
decrease tooling cost to make economical micro-batch (1) Digital-Limited have mature physical technologies,
production? but still require dramatic improvements in the digital
• Distributed manufacturing and global value chains: technologies. (2) Emerging Digital and Physical, where
Assuming the limitations of direct manufacturing can- both modes are significant limiting factors, but neither
not be overcome, to which extent other digital technology has reached maturity. (3) Mature Digital and
manufacturing technologies (such as laser cutting or Physical technologies that are combined for an immedi-
CNC) could be combined with 3D printing in inte- ate wave of combinatorial (often digital) innovation
grated and universal manufacturing units in order to (Sandberg et al., 2020). However, once the combination
deliver the promise of distributed manufacturing, and, reaches maturity, subsequent improvements are slow and
consequently, enable more sustainable global value incremental. (4) Physical-Limited, where the rate of digi-
chain reconfigurations? tal innovation is much faster than the physical and thus
• Logistics and supply chains: Could integrated digital overall performance is limited by the physical.
manufacturing units be leveraged to deliver pseudo Finally, the Purely Physical systems are the familiar
on-demand manufacturing, possibly with the help of domain of innovation studies of the past 40 years—such as
rapid tooling enabling to produce just-in-time small railroads, power systems airplanes, automobiles and
batches production using traditional manufacturing? manufacturing processes (Anderson & Tushman, 1990;
Likewise, could rapid tooling meet very tight lead Hughes, 1983; Utterback, 1994). Today, most mechanical
times by utilizing local integrated digital manufactur- innovations have some digital component, but purely phys-
ing units? Also could hybrid production strategies ical innovation can be found in materials science—such as
manufacture parts less prone to wearing out through photovoltaic cells and modern battery materials—where
3D printing, while consumable parts remain mass the rate of progress remains slow despite billions of dollars
produced? in annual R&D spending.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 545

TABLE 4 Impact of reverse salients in digital and physical innovation.

Reverse Rate of
Domain Digital Physical salient innovation Example
Purely digital Emerging N/a Digital Typically high Artificial intelligence (Haenlein &
Kaplan, 2019)
CPS (digital-limited) Emerging Mature Digital Typically high Facial recognition, edge computing, IoT
CPS (digital and physical Emergent Emergent Both Slow Virtual reality (Zhan et al., 2020),
limited) autonomous vehicles (Bagloee
et al., 2016)
CPS (mature digital and Mature Mature Both Initially high, CNC (Ernst, 1997), Internet-enabled
physical) then slow appliances (Boutellier & Heinzen, 2014)
CPS (physical-limited) Mature Emerging Physical Slow 3D printers, wind turbines (Beiter
et al., 2021; Veers et al., 2019)
Purely physical N/a Emerging Physical Slow Photovoltaic cells, battery materials (Chu
et al., 2017)

Research on CPS innovation so far has emphasized the digital constraints on system performance, what is
digital-limited case, but future research should examine needed is a more nuanced discussion of the interaction of
CPSs that are physical-limited. Since CPS performance can this digital innovation with the inevitable physical
be limited by either digital or physical components, if constraints.
researchers overestimate the rate of improvement for the Additional research should examine other hybrid
physical components, this leads to incorrect predictions of CPSs (beyond digital manufacturing) where the perfor-
the innovation trajectory of a CPS. Future research is also mance drivers are more physical than digital. Such stud-
needed on the interdependence of component innovation ies should consider two possible technological
for CPS and their respective innovation rates and techno- trajectories. The first is when physical components limit
logical bottlenecks. the rate of innovation for digital components or the over-
At the same time, theories of digital innovation (and all system. The second is when improvements in digital
innovation more generally) have rarely considered what components work around the limitations of slowly
happens when the digital meets the physical. Thus improving physical ones, as appears to be happening
research on digital innovation should consider the rarely with self-driving cars and virtual reality displays. Theo-
explored middle ground of Table 4, between traditional ries and measures for reverse salients could help explain
(physical) and digital innovation theories. The case of 3D both technological trajectory. Finally, future research
printing reminds us that in the end, digital innovation should examine reverse salients where progress depends
must serve material beings living in a material world. on a complex interplay of digital and physical innovation;
Cyber-physical and other systems that span both domains such studies could link to the rich literature on
inherently face material constraints not applicable to co-evolutionary innovation processes (e.g., Garud &
purely digital innovations, but even the latter will eventu- Rappa, 1994; Geels, 2005; van den Ende et al., 2012).
ally face material constraints when they interface with
the material world.
Future research on digital innovation is thus at a 5.3 | Implications for reverse salients
crossroads as to how to address this tangible reality. The
easiest way would be to define the scope of the theory While Hughes' concept of a reverse salient continues to
and its predictions as bounded by purely digital influence innovation research (Kapoor & McGrath, 2014;
innovations—emphasizing the boundaries of previous Lenfle & Söderlund, 2022; Mäkitie et al., 2022; Mathews
theoretical findings and perhaps the crossover points to et al., 2018; Sandberg et al., 2020; Schubert et al., 2013),
adjacent theories. The path less traveled would be to that research continues to have a singular blind spot, by
adapt and extend theories of digital innovation to explain assuming reverse salients can and will be overcome. While
digitally-enabled physical innovation, as we have done innovation studies examine successful technologies—or
with 3D printing and others have done with automobiles occasionally failed projects within a successful technology
(Lee & Berente, 2012). While previous studies of digital- (Law & Callon, 1992)—few reverse salients studies have
physical (and CPS) systems studied have emphasized examined when the salient is not overcome.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
546 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

This could be just a methodological issue: Hughes predictions. These suggest multiple opportunities to study
(1983) did not publish his study of electric power in 1890, persistent reverse salients that shape technological trajec-
just as West (2008) did not publish his study of deep tories, problem recognition and commercialization over
space communications in 1965 and Lenfle and Söderlund decades.
(2022) did not publish their study of numerical weather
prediction in 1990. In 2000, researchers would have con-
cluded photovoltaic solar energy equipment was a tech- 5.4 | Managerial implications
nology that failed to meet its goals for decades
(West, 2014), but today such equipment generates more As Hughes emphasized, reverse salients can provide great
than $100 billion in global revenue and a double-digit market opportunities. However, 3D printing points to
compound growth rate. what happens when the time horizon for resolving them
Future research should thus consider this fundamental is beyond the time horizon of private managers. While
question: What happens to a technological trajectory when utopian predictions sell analyst reports, consulting ser-
salients are not overcome? As noted earlier, 3D printing vices and stock offerings, the longstanding salients in 3D
suggests that two possible options are to target market printing performance provide exemplars of how man-
niches or to compensate by innovating in other compo- agers can identify inherent (and difficult to resolve) tech-
nents of the system. But in both cases, the salient remains. nological limits. Although rarely identified in prior
How might this be studied? While Coopersmith research, multiple simultaneous reverse salients make it
(2020) suggests that one explanation is the lack of data even less likely that the technology will make immediate
for failed technologies, a well-documented example of and rapid progress.
utopian predictions that are unrealized due to persistent At the same time, not all hybrid digital-physical sys-
reverse salients would include artificial intelligence. tems are the same. We offer a framework for identifying
Beginning with the field's origins at a 1956 Dartmouth which systems are most capable of rapid progress and
Summer conference that included Claude Shannon, which ones are not. At the same time, digital innovation
Herb Simon and Marvin Minksy, AI missed key predic- promises rapid and malleable innovation that can deliver
tions for decades. For example, in 1973 AI experts pre- rapid initial improvements in CPSs until the physical lim-
dicted (Nilson, 2009, p. 515) the commercial availability itations are reached. Together, these suggest caution for
of automatic medical diagnosis (in 1980), automated managers investing in CPS innovation where physical
voice-to-text (in 1992) and self-driving vehicles for city innovation is required to realize the hypothesized benefit
streets (in 2000) but the actual wait was at least 50% of that system.
longer. For companies involved in 3D printing, whether as
Empirical research on reverse salients could thus providers or customers, a key implication is to take the
develop a more complete framework for the co-evolution physical limitations of 3D printing as they are, without
of salients, critical problems and technological trajecto- expecting that they will eventually disappear. Instead of
ries. Prior research has noted the difference between considering uses where 3D printing would have a major
salients that are quickly recognized and those that are impact if limitations were eliminated, they should instead
not. Future research could consider both salients that focus on uses where 3D printing already can make a dif-
take decades to resolve and those not yet resolved as of ference. Such analysis should leverage convergence with
the current writing. Even among successful innovations, other digital technologies, whether on the digital
such decades-long trajectories are more common than (AI, AR) or the physical side (robotics, CNC, IoT) to fully
often assumed: The 1987 Knowledge Navigator video that unleash the potential of 3D printing as it now is, not as
made Apple CEO John Sculley the butt of endless jokes we dream it could be.
was largely realized 24 years later when Apple added the
Siri voice assistant to the iPhone. ACKNOWLEDGMENTS
Another complex systemic innovation that has been We are grateful to the reviewers and guest editor Michael
the subject of several salient studies is the transition to Stanko for their guidance and support throughout the
renewable energy, as when Mäkitie et al. (2022) exam- development of this article; we particularly appreciate
ined how reverse salients in three RE transportation tech- the reviewer who suggested a possible connection to the
nologies influenced investment decisions. Beyond AI and influential work of Thomas Hughes.
RE, data are readily available on reverse salients for other
key technologies such as fusion power production, F U N D I N G IN F O R M A T I O N
humanoid robots or interplanetary exploration; more This research was in part funded through the Chair Tech-
recently, small-scale nuclear power plants and flying tax- nology for Change of the Institut Polytechnique de Paris,
ies are growing closer to realizing decades-long utopian as supported by the patronage of Accenture.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 547

CONFLICT OF INTEREST STATEMENT Beiter, P., A. Cooperman, E. Lantz, T. Stehly, M. Shields, R. Wiser,
The authors declare no conflicts of interest. T. Telsnig, L. Kitzing, V. Berkhout, and Y. Kikuchi. 2021.
“Wind Power Costs Driven by Innovation and Experience With
Further Reductions on the Horizon.” WIREs Energy and Envi-
E TH IC S ST A T EME N T
ronment 10(5): e398. Portico. https://doi.org/10.1002/wene.398.
The authors have read and agreed to the Committee on Ben‐Ner, A., and E. Siemsen. 2017. “Decentralization and Localiza-
Publication Ethics (COPE) international standards for tion of Production.” California Management Review 59(2): 5–23.
authors. https://doi.org/10.1177/0008125617695284.
Bijker, W. E., and J. Law, eds. 1994. Shaping Technology/Building
ORCID Society: Studies in Sociotechnical Change. Cambridge, MA: MIT
Thierry Rayna https://orcid.org/0000-0003-0144-6031 Press.
Birtchnell, T., A. Daly, T. Rayna, and L. Striukova. 2018. 3D Print-
Joel West https://orcid.org/0000-0002-7958-5229
ing and Intellectual Property Futures (Independent Report No.
2018/03). Newport, UK: UK Intellectual Property Office.
R EF E RE N C E S Blakey-Milner, B., P. Gradl, G. Snedden, M. Brooks, J. Pitot, E.
Achillas, C., D. Aidonis, E. Iakovou, M. Thymianidis, and D. Lopez, M. Leary, F. Berto, and A. du Plessis. 2021. “Metal
Tzetzis. 2015. “A Methodological Framework for the Inclusion Additive Manufacturing in Aerospace: A Review.” Materials &
of Modern Additive Manufacturing into the Production Portfo- Design 209: 110008. https://doi.org/10.1016/j.matdes.2021.
lio of a Focused Factory.” Journal of Manufacturing Systems 37: 110008.
328–39. https://doi.org/10.1016/j.jmsy.2014.07.014. Blueprint. 2020. “CES 2020 Shows Us What the End of the Consumer
Appleyard, M. M., and H. W. Chesbrough. 2017. “The Dynamics of 3D Printing Hype Cycle Looks Like.” 3DPrint.ComjThe Voice of
Open Strategy: From Adoption to Reversion.” Long Range Plan- 3D Printing/Additive Manufacturing. January 24, 2020. https://
ning 50(3): 310–21. https://doi.org/10.1016/j.lrp.2016.07.004. 3dprint.com/262622/ces-2020-shows-us-what-the-end-of-the-
Aryan, V., J. Bertling, and C. Liedtke. 2021. “Topology, Typology, consumer-3d-printing-hype-cycle-looks-like/.
and Dynamics of Commons-Based Peer Production: On Plat- Bogers, M., R. Hadar, and A. Bilberg. 2016. “Additive Manufactur-
forms, Actors, and Innovation in the Maker Movement.” Crea- ing for Consumer-Centric Business Models: Implications for
tivity and Innovation Management 30(1): 63–79. https://doi.org/ Supply Chains in Consumer Goods Manufacturing.” Technolog-
10.1111/caim.12392. ical Forecasting and Social Change 102: 225–39. https://doi.org/
Åstebro, T. 2002. “Noncapital Investment Costs and the Adoption 10.1016/j.techfore.2015.07.024.
of CAD and CNC in U.S. Metalworking Industries.” RAND Bogers, M., and W. Horst. 2014. “Collaborative Prototyping:
Journal of Economics 33(4): 672–88. https://doi.org/10.2307/ Cross-Fertilization of Knowledge in Prototype-Driven Problem
3087480. Solving.” Journal of Product Innovation Management 31(4):
Autio, E., R. Mudambi, and Y. Yoo. 2021. “Digitalization and Glob- 744–64. https://doi.org/10.1111/jpim.12121.
alization in a Turbulent World: Centrifugal and Centripetal Boland, R. J., K. Lyytinen, and Y. Yoo. 2007. “Wakes of Innovation
Forces.” Global Strategy Journal 11(1): 3–16. https://doi.org/10. in Project Networks: The Case of Digital 3‐D Representations
1002/gsj.1396. in Architecture, Engineering, and Construction.” Organization
Avinson. 2018. “How Long Does 3D Printing Take?” September Science 18(4): 631–47. https://doi.org/10.1287/orsc.1070.0304.
20, 2018. https://3space.com/blog/how-long-does-3d-printing- Borgia, E. 2014. “The Internet of Things Vision: Key Features,
take/. Applications and Open Issues.” Computer Communications 54:
Böckin, D., and A.-M. Tillman. 2019. “Environmental Assessment 1–31. https://doi.org/10.1016/j.comcom.2014.09.008.
of Additive Manufacturing in the Automotive Industry.” Jour- Boutellier, R., and M. Heinzen. 2014. Growth Through Innovation:
nal of Cleaner Production 226: 977–87. https://doi.org/10.1016/ Managing the Technology-Driven Enterprise. Cham, Switzer-
j.jclepro.2019.04.086. land: Springer Science & Business Media.
Baars, H., C. Felden, P. Gluchowski, A. Hilbert, H.-G. Kemper, and Bowles, M. D., and V. P. Dawson. 1998. “The Advanced Turboprop
S. Olbrich. 2014. “Shaping the Next Incarnation of Business Project: Radical Innovation in a Conservative Environment.” In
Intelligence.” Business & Information Systems Engineering 6(1): From Engineering Science to Big Science—The NACA and NASA
11–6. https://doi.org/10.1007/s12599-013-0307-z. Collier Trophy Research Project Winners, edited by Pamela Etter
Bagloee, S. A., M. Tavana, M. Asadi, and T. Oliver. 2016. “Autono- Mack, 321–43. Honolulu: University Press of the Pacific.
mous Vehicles: Challenges, Opportunities, and Future Implica- Budinoff, H. D., J. Bushra, and M. Shafae. 2021. “Community-
tions for Transportation Policies.” Journal of Modern Driven PPE Production Using Additive Manufacturing During
Transportation 24(4): 284–303. https://doi.org/10.1007/s40534- the COVID-19 Pandemic: Survey and Lessons Learned.” Jour-
016-0117-3. nal of Manufacturing Systems 60: 799–810. https://doi.org/10.
Baldwin, C. Y. 2015. “Bottlenecks, Modules, and Dynamic Architec- 1016/j.jmsy.2021.07.010.
tural Capabilities.” In The Oxford Handbook of Dynamic Capa- Campbell, T., C. Williams, O. Ivanova, and B. Garrett. 2011. Could
bilities, Online ed., edited by David J. Teece and Sohvi Heaton. 3D Printing Change the World? Washington, DC: Atlantic
Oxford: Oxford University Press. https://doi.org/10.1093/ Council.
oxfordhb/9780199678914.013.011. Chandler, A. D. 1990. Scale and Scope: The Dynamics of Industrial
Baumers, M., and M. Holweg. 2019. “On the Economics of Additive Capitalism. Cambridge, MA: Belknap Press.
Manufacturing: Experimental Findings.” Journal of Operations Chaudhuri, A., H. A. Gerlich, J. Jayaram, A. Ghadge, J. Shack,
Management 65(8): 794–809. https://doi.org/10.1002/joom. B. H. Brix, L. H. Hoffbeck, and N. Ulriksen. 2021. “Selecting
1053. Spare Parts Suitable for Additive Manufacturing: A Design
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
548 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Science Approach.” Production Planning & Control 32(8): 670– Dedehayir, O., and S. J. Mäkineif. 2008. “Dynamics of Reverse
87. https://doi.org/10.1080/09537287.2020.1751890. Salience as Technological Performance Gap: An Empirical
Christiansen, A. C., and J. Buen. 2002. “Managing Environmental Study of the Personal Computer Technology System.” Journal
Innovation in the Energy Sector: The Case of Photovoltaic and of Technology Management & Innovation 3(3): 55–66. https://
Wave Power Development in Norway.” International Journal of doi.org/10.4067/S0718-27242008000100006.
Innovation Management 6(3): 233–56. https://doi.org/10.1142/ Dedehayir, O., and S. J. Mäkinen. 2011b. “Determining Reverse
s1363919602000586. Salient Types and Evolutionary Dynamics of Technology Sys-
Chu, S., Y. Cui, and N. Liu. 2017. “The Path Towards Sustainable tems With Performance Disparities.” Technology Analysis &
Energy.” Nature Materials 16(1): 16–22. https://doi.org/10. Strategic Management 23(10): 1095–114. https://doi.org/10.
1038/nmat4834. 1080/09537325.2011.621308.
Claussen, J., and M. A. Halbinger. 2021. “The Role of Pre- Depoorter, B. 2014. “Intellectual Property Infringements & 3D Printing:
Innovation Platform Activity for Diffusion Success: Evidence Decentralized Piracy.” Hastings Law Journal 65(6): 1483–504.
From Consumer Innovations on a 3D Printing Platform.” https://repository.uclawsf.edu/hastings_law_journal/vol65/iss6/2.
Research Policy 50(8): 103943. https://doi.org/10.1016/j.respol. Despeisse, M., and T. Minshall. 2017. “Skills and Education for
2020.103943. Additive Manufacturing: A Review of Emerging Issues.” In
Conkle, J., P. S. Suchdev, E. Alexander, R. Flores-Ayala, Advances in Production Management Systems. The Path to Intel-
U. Ramakrishnan, and R. Martorell. 2018. “Accuracy and Reli- ligent, Collaborative and Sustainable Manufacturing, edited by
ability of a Low-Cost, Handheld 3D Imaging System for Child H. Lödding, R. Riedel, K.-D. Thoben, G. von Cieminski, and D.
Anthropometry.” PLoS One 13(10): e0205320. https://doi.org/ Kiritsis, 289–97. Cham, Switzerland: Springer International
10.1371/journal.pone.0205320. Publishing. https://doi.org/10.1007/978-3-319-66923-6_34.
Cooper, R. G. 2021. “Accelerating Innovation: Some Lessons From Dosi, G. 1982. “Technological Paradigms and Technological Trajec-
the Pandemic.” Journal of Product Innovation Management tories.” Research Policy 11(3): 147–62. https://doi.org/10.1016/
38(2): 221–32. https://doi.org/10.1111/jpim.12565. 0048-7333(82)90016-6.
Coopersmith, J. 2020. “Spectacular Flops: Game‐Changing Technol- Dowling, L., J. Kennedy, S. O'Shaughnessy, and D. Trimble. 2020.
ogies That Failed by Michael Brian Schiffer.” Technology and “A Review of Critical Repeatability and Reproducibility Issues
Culture 61(4): 1235–6. https://doi.org/10.1353/tech.2020.0134. in Powder Bed Fusion.” Materials & Design 186: 108346.
Coutts, E. R., A. Wodehouse, and J. Robertson. 2019. “A Compari- https://doi.org/10.1016/j.matdes.2019.108346.
son of Contemporary Prototyping Methods.” Proceedings of the Ernst, H. 1997. Small Business Economics 9(4): 361–81. https://doi.
Design Society: International Conference on Engineering Design org/10.1023/a:1007921808138.
1(1): 1313–22. https://doi.org/10.1017/dsi.2019.137. Eyers, D. R., and A. T. Potter. 2017. “Industrial Additive
Daim, T., J. Justice, L. Hogaboam, S. J. Mäkinen, and O. Dedehayir. Manufacturing: A Manufacturing Systems Perspective.” Com-
2014. “Identifying and Forecasting the Reverse Salient in Video puters in Industry 92–93: 208–18. https://doi.org/10.1016/j.
Game Consoles: A Performance Gap Ratio Comparative Analy- compind.2017.08.002.
sis.” Technological Forecasting and Social Change 82: 177–89. Feng, J., J. Fu, Z. Lin, C. Shang, and B. Li. 2018. “A Review of the
https://doi.org/10.1016/j.techfore.2013.06.007. Design Methods of Complex Topology Structures for 3D Print-
Davies, A. 1996. “Innovation in Large Technical Systems: The Case ing.” Visual Computing for Industry, Biomedicine, and Art 1(1):
of Telecommunications.” Industrial and Corporate Change 5(4): 5. https://doi.org/10.1186/s42492-018-0004-3.
1143–80. https://doi.org/10.1093/icc/5.4.1143. Fontana, F., C. Klahn, and M. Meboldt. 2019. “Value-Driven Clus-
Davies, A., and T. Brady. 2000. “Organisational Capabilities and tering of Industrial Additive Manufacturing Applications.”
Learning in Complex Product Systems: Towards Repeatable Journal of Manufacturing Technology Management 30(2): 366–
Solutions.” Research Policy 29(7–8): 931–53. https://doi.org/10. 90. https://doi.org/10.1108/JMTM-06-2018-0167.
1016/s0048-7333(00)00113-x. Ford, S., and M. Despeisse. 2016. “Additive Manufacturing and Sus-
de Bruijn, E. 2010. On the Viability of the Open Source Development tainability: An Exploratory Study of the Advantages and Chal-
Model for the Design of Physical Objects: Lessons Learned From lenges.” Journal of Cleaner Production 137: 1573–87. https://
the RepRap Project. Unpublished MSc Thesis. Tilburg: Tilburg doi.org/10.1016/j.jclepro.2016.04.150.
University. Frandsen, C. S., M. M. Nielsen, A. Chaudhuri, J. Jayaram, and K.
de Jong, J. P., and E. de Bruijn. 2013. “Innovation Lessons From Govindan. 2020. “In Search for Classification and
3-D Printing.” MIT Sloan Management Review 54(2): 43–52. Selection of Spare Parts Suitable for Additive Manufacturing:
Dedehayir, O., and S. J. Mäkinen. 2011a. “Measuring Industry Clock- A Literature Review.” International Journal of Production
speed in the Systemic Industry Context.” Technovation 31(12): Research 58(4): 970–96. https://doi.org/10.1080/00207543.
627–37. https://doi.org/10.1016/j.technovation.2011.07.008. 2019.1605226.
Dedehayir, O. 2009. “Bibliometric Study of the Reverse Salient Con- Franke, N., and F. Piller. 2004. “Value Creation by Toolkits for User
cept.” Journal of Industrial Engineering and Management 2(3): Innovation and Design: The Case of the Watch Market.” Jour-
3. https://doi.org/10.3926/jiem.v2n3.p569-591. nal of Product Innovation Management 21(6): 401–15. https://
Dedehayir, O., and A. Hornsby. 2008. “Staying at the Technology doi.org/10.1111/j.0737-6782.2004.00094.x.
Forefront by Spotting the Reverse Salient: The Case of Digital Frazer, J. S. 2022. “The Role of Distributed Manufacturing and 3D
Video Broadcasting.” In 2008 4th IEEE International Conference Printing in Development of Personal Protective Equipment
on Management of Innovation and Technology 189–94. https:// Against COVID-19.” In Emerging Applications of 3D Printing
doi.org/10.1109/icmit.2008.4654360. During CoVID 19 Pandemic, edited by K. Sandhu, S. Singh, C.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 549

Prakash, N. R. Sharma, and K. Subburaj, 15–34. Singapore: Operations Management 65(8): 810–26. https://doi.org/10.1002/
Springer. https://doi.org/10.1007/978-981-33-6703-6_2. joom.1054.
Fujita, M. 2001. “AIBO: Toward the Era of Digital Creatures.” The Hendler, S. 2019. “Digital‐Physical Product Development: A
International Journal of Robotics Research 20(10): 781–94. Qualitative Analysis.” European Journal of Innovation Man-
https://doi.org/10.1177/02783640122068092. agement 22(2): 315–34. https://doi.org/10.1108/ejim-01-2018-
Garcia, R., and R. Calantone. 2002. “A Critical Look at Technologi- 0026.
cal Innovation Typology and Innovativeness Terminology: A Herterich, M. M., T. Buehnen, F. Uebernickel, and W. Brenner.
Literature Review.” Journal of Product Innovation Management 2016. “A Taxonomy of Industrial Service Systems Enabled by
19(2): 110–32. https://doi.org/10.1111/1540-5885.1920110. Digital Product Innovation.” In Proceedings of the 49th Hawaii
Garud, R., and M. A. Rappa. 1994. “A Socio‐Cognitive Model of International Conference on System Sciences (HICSS-49) 1236–
Technology Evolution: The Case of Cochlear Implants.” Orga- 45. https://doi.org/10.1109/HICSS.2016.157.
nization Science 5(3): 344–62. https://doi.org/10.1287/orsc.5. Hill, C. W. L., and F. T. Rothaermel. 2003. “The Performance of
3.344. Incumbent Firms in the Face of Radical Technological Innova-
GE. 2020. Better Performance From Fewer Parts. GE.com. https:// tion.” The Academy of Management Review 28(2): 257. https://
web.archive.org/web/20200522182434/https://www.ge.com/ doi.org/10.2307/30040712.
additive/additive-manufacturing/industries/aviation-aerospace. Holmström, J., M. Holweg, S. H. Khajavi, and J. Partanen. 2016.
Gebler, M., A. J. S. Uiterkamp, and C. Visser. 2014. “A Global Sus- “The Direct Digital Manufacturing (R)evolution: Definition of a
tainability Perspective on 3D Printing Technologies.” Energy Research Agenda.” Operations Management Research 9(1–2): 1–
Policy 74: 158–67. https://doi.org/10.1016/j.enpol.2014.08.033. 10. https://doi.org/10.1007/s12063-016-0106-z.
Geels, F. W. 2005. Technological Transitions and System Innova- Holzmann, P., R. J. Breitenecker, and E. J. Schwarz. 2020. “Business
tions: A Co-Evolutionary and Socio-Technical Analysis. Chelten- Model Patterns for 3D Printer Manufacturers.” Journal of
ham, UK: Elgar. Manufacturing Technology Management 31(6): 1281–300.
Ghaffar, S., and P. Mullett. 2021. “Discussion: Commentary: 3D https://doi.org/10.1108/JMTM-09-2018-0313.
Printing Set to Transform the Construction Industry.” Proceed- Hoppmann, J., J. Huenteler, and B. Girod. 2014. “Compulsive Pol-
ings of the Institution of Civil Engineers—Structures and Build- icy‐Making—The Evolution of the German Feed‐In Tariff Sys-
ings 174(4): 336–7. https://doi.org/10.1680/jstbu.21.00024. tem for Solar Photovoltaic Power.” Research Policy 43(8): 1422–
Gibson, I., D. Rosen, B. Stucker, and M. Khorasani. 2014. Additive 41. https://doi.org/10.1016/j.respol.2014.01.014.
Manufacturing Technologies. New York: Springer. Hu, B., and Z. Sun. 2022. “Managing Self-Replicating Innovative
Goh, G. D., S. L. Sing, and W. Y. Yeong. 2020. “A Review on Goods.” Management Science 68(1): 399–419. https://doi.org/10.
Machine Learning in 3D Printing: Applications, Potential, and 1287/mnsc.2020.3936.
Challenges.” Artificial Intelligence Review 54(1): 63–94. https:// Hughes, T. P. 1983. Networks of Power: Electrification in Western
doi.org/10.1007/s10462-020-09876-9. Society, 1880–1930. Baltimore: Johns Hopkins University Press.
Greer, C., M. Burns, D. Wollman, and E. Griffor. 2019. Cyber- Hughes, T. P. 1987. “The Evolution of Large Technological Sys-
Physical Systems and Internet of Things. Gaithersburg, MD: tems.” In The Social Construction of Technological Systems, edi-
National Institute of Standards and Technology. ted by W. E. Bijker, T. P. Hughes, and T. J. Pinch, 51–82.
Greul, A., J. West, and S. Bock. 2018. “Open at Birth? Why New Cambridge, MA: MIT Press.
Firms Do (or Don’t) Use Open Innovation.” Strategic Entrepre- Hull, C., M. Feygin, Y. Baron, R. Sanders, E. Sachs, A. Lightman,
neurship Journal 12(3): 392–420. Portico. https://doi.org/10. and T. Wohlers. 1995. “Rapid Prototyping: Current Technology
1002/sej.1282. and Future Potential.” Rapid Prototyping Journal 1(1): 11–9.
Haenlein, M., and A. Kaplan. 2019. “A Brief History of Artificial https://doi.org/10.1108/13552549510732026.
Intelligence: On the Past, Present, and Future of Artificial Intel- Iansiti, M., and K. R. Lakhani. 2020. Competing in the Age of AI:
ligence.” California Management Review 61(4): 5–14. https:// Strategy and Leadership When Algorithms and Networks Run
doi.org/10.1177/0008125619864925. the World. Boston: Harvard Business Review Press.
Haleem, A., M. Javaid, and S. Rab. 2020. “Impact of Additive Jacobs, P. F. 1992. Rapid Prototyping & Manufacturing: Fundamen-
Manufacturing in Different Areas of Industry 4.0.” Interna- tals of Stereolithography. Dearborn, MI: Society of Manufactur-
tional Journal of Logistics Systems and Management 37(2): 239. ing Engineers.
https://doi.org/10.1504/IJLSM.2020.110578. Javaid, M., A. Haleem, R. Pratap Singh, and R. Suman. 2021.
Han, K., and J. Shin. 2014. “A Systematic Way of Identifying and “Industrial Perspectives of 3D Scanning: Features, Roles and
Forecasting Technological Reverse Salients Using QFD, Biblio- It's Analytical Applications.” Sensors International 2: 100114.
metrics, and Trend Impact Analysis: A Carbon Nanotube Bio- https://doi.org/10.1016/j.sintl.2021.100114.
sensor Case.” Technovation 34(9): 559–70. https://doi.org/10. Jiang, R., R. Kleer, and F. T. Piller. 2017. “Predicting the Future of
1016/j.technovation.2014.05.009. Additive Manufacturing: A Delphi Study on Economic and
Hanelt, A., E. Piccinini, R. W. Gregory, B. Hildebrandt, and L. M. Societal Implications of 3D Printing for 2030.” Technological
Kolbe. 2015, March. “Digital Transformation of Primarily Physical Forecasting and Social Change 117: 84–97. https://doi.org/10.
Industries-Exploring the Impact of Digital Trends on Business 1016/j.techfore.2017.01.006.
Models of Automobile Manufacturers.” Proceedings der 12. Inter- Kalantari, H. D., L. W. Johnson, and C. R. Perera. 2021. “The Effect
nationalen Tagung Wirtschaftsinformatik: 1313–27. of Gender on Willingness to Pay for Mass Customised Running
Heinen, J. J., and K. Hoberg. 2019. “Assessing the Potential of Addi- Shoes.” Journal of Global Fashion Marketing 12(2): 161–75.
tive Manufacturing for the Provision of Spare Parts.” Journal of https://doi.org/10.1080/20932685.2020.1867878.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
550 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Kapoor, R., and P. J. McGrath. 2014. “Unmasking the Interplay Open Journal of the Industrial Electronics Society 1: 298–310.
Between Technology Evolution and R&D Collaboration: Evi- https://doi.org/10.1109/OJIES.2020.3031660.
dence From the Global Semiconductor Manufacturing Indus- Lenfle, S., and J. Söderlund. 2022. “Project‐Oriented Agency and
try, 1990–2010.” Research Policy 43(3): 555–69. https://doi.org/ Regeneration in Socio‐Technical Transition: Insights From the
10.1016/j.respol.2013.08.002. Case of Numerical Weather Prediction (1978–2015).” Research
Kenney, M. E. 2013. Cost Reduction Through the Use of Additive Policy 51(3): 104455. https://doi.org/10.1016/j.respol.2021.
Manufacturing (3D Printing) and Collaborative Product Life Cycle 104455.
Management Technologies to Enhance the Navy's Maintenance Li, L. 2018. “China's Manufacturing Locus in 2025: With a Compar-
Programs. Doctoral dissertation. Monterey, CA: Naval Postgradu- ison of “Made-in-China 2025” and “Industry 4.0”.” Technologi-
ate School. https://apps.dtic.mil/sti/citations/ADA591795. cal Forecasting and Social Change 135: 66–74. https://doi.org/
Khan, H. N., D. A. Hounshell, and E. R. H. Fuchs. 2018. “Science 10.1016/j.techfore.2017.05.028.
and Research Policy at the End of Moore's Law.” Nature Elec- Li, S. 2013. A Desktop Factory; 3-D Printers, Which Can Make Plas-
tronics 1(1): 14–21. https://doi.org/10.1038/s41928-017-0005-9. tic Objects, Are Creeping Into the Home. Los Angeles Times,
Khorasani, M., A. Ghasemi, B. Rolfe, and I. Gibson. 2021. “Additive July 28, p. B1.
Manufacturing a Powerful Tool for the Aerospace Industry.” Lien, Shao-Yu, Shin-Ming Cheng, Sung-Yin Shih, and Kwang-
Rapid Prototyping Journal 28(1): 87–100. https://doi.org/10. Cheng Chen. 2012. “Radio Resource Management for QoS
1108/RPJ-01-2021-0009. Guarantees in Cyber-Physical Systems.” IEEE Transactions on
Kim, D.‐J., and B. Kogut. 1996. “Technological Platforms and Diver- Parallel and Distributed Systems 23(9): 1752–61. https://doi.org/
sification.” Organization Science 7(3): 283–301. https://doi.org/ 10.1109/TPDS.2012.151.
10.1287/orsc.7.3.283. Luchs, M. G., K. S. Swan, and M. E. H. Creusen. 2015. “Perspective:
Kim, H., Y. Lin, and T.-L. B. Tseng. 2018. “A Review on Quality A Review of Marketing Research on Product Design With
Control in Additive Manufacturing.” Rapid Prototyping Journal Directions for Future Research.” Journal of Product Innovation
24(3): 645–69. https://doi.org/10.1108/RPJ-03-2017-0048. Management 33(3): 320–41. Portico. https://doi.org/10.1111/
Kling, R. 1996. “Hopes and Horrors: Technological Utopianism and jpim.12276.
Anti-Utopianism in Narratives of Computerization.” In Com- Lyytinen, K. 2022. “Innovation Logics in the Digital Era: A Sys-
puterization and Controversy: Value Conflicts and Social temic Review of the Emerging Digital Innovation Regime.”
Choices, edited by R. Kling, 40–58. San Diego: Academic Press. Innovation: Organization and Management 24(1): 13–34.
Knofius, N., M. C. van der Heijden, and W. H. M. Zijm. 2019. “Con- https://doi.org/10.1080/14479338.2021.1938579.
solidating Spare Parts for Asset Maintenance With Additive Mäkitie, T., J. Hanson, M. Steen, T. Hansen, and A. D. Andersen.
Manufacturing.” International Journal of Production Economics 2022. “Complementarity Formation Mechanisms in Technol-
208: 269–80. https://doi.org/10.1016/j.ijpe.2018.11.007. ogy Value Chains.” Research Policy 51(7): 104559. https://doi.
Law, J., and M. Callon. 1992. “The Life and Death of an Aircraft: A org/10.1016/j.respol.2022.104559.
Network Analysis of Technical Change.” In Shaping Mahmood, T., and M. S. Mubarik. 2020. “Balancing Innovation and
Technology/Building Society: Studies in Sociotechnical Change, Exploitation in the Fourth Industrial Revolution: Role of Intel-
edited by W. E. Bijker and J. Law, 21–52. Cambridge, MA: MIT lectual Capital and Technology Absorptive Capacity.” Techno-
Press. logical Forecasting and Social Change 160: 120248. https://doi.
Lee, E. A. 2008. “Cyber Physical Systems: Design Challenges.” In org/10.1016/j.techfore.2020.120248.
2008 11th IEEE International Symposium on Object and Compo- Majidpour, M. 2016. “Technological Catch‐Up in Complex Product
nent‐Oriented Real‐Time Distributed Computing (ISORC). Systems.” Journal of Engineering and Technology Management
https://doi.org/10.1109/isorc.2008.25. 41: 92–105. https://doi.org/10.1016/j.jengtecman.2016.07.003.
Lee, J., B. Bagheri, and H.‐A. Kao. 2015. “A Cyber‐Physical Systems Markus, M. L. 1994. “Electronic Mail as the Medium of Managerial
Architecture for Industry 4.0‐Based Manufacturing Systems.” Choice.” Organization Science 5(4): 502–27. https://doi.org/10.
Manufacturing Letters 3: 18–23. https://doi.org/10.1016/j. 1287/orsc.5.4.502.
mfglet.2014.12.001. Mathews, J. A., H. Tan, and M. C. Hu. 2018. “Moving to a Circular
Lee, J., and N. Berente. 2012. “Digital Innovation and the Division Economy in China: Transforming Industrial Parks into Eco-
of Innovative Labor: Digital Controls in the Automotive Indus- Industrial Parks.” California Management Review 60(3): 157–81.
try.” Organization Science 23(5): 1428–47. https://doi.org/10. https://doi.org/10.1177/0008125617752692.
1287/orsc.1110.0707. Mavri, M., E. Fronimaki, and A. Kadrefi. 2021. “Survey Analysis for
Lee, J., H.-C. Kim, J.-W. Choi, and I. H. Lee. 2017. “A Review on the Adoption of 3D Printing Technology: Consumers' Perspec-
3D Printed Smart Devices for 4D Printing.” International Jour- tive.” Journal of Science and Technology Policy Management 14
nal of Precision Engineering and Manufacturing-Green Technol- (2): 353–85. https://doi.org/10.1108/JSTPM-02-2020-0023.
ogy 4(3): 373–83. https://doi.org/10.1007/s40684-017-0042-x. Mendis, D. K., M. A. Lemley, and M. Rimmer, eds. 2019. 3D Print-
Lehenkari, J., and R. Miettinen. 2002. “Standardisation in the ing and Beyond: Intellectual Property and Regulation. Chelten-
Construction of a Large Technological System—The Case of the ham, UK: Elgar.
Nordic Mobile Telephone System.” Telecommunications Policy Millard, J., M. N. Sorivelle, S. Deljanin, E. Unterfrauner, and
26(3–4): 109–27. https://doi.org/10.1016/s0308-5961(02)00004-6. C. Voigt. 2018. “Is the Maker Movement Contributing to
Leitão, P., F. Pires, S. Karnouskos, and A. W. Colombo. 2020. “Quo Sustainability?” Sustainability 10(7): 7. https://doi.org/10.3390/
Vadis Industry 4.0? Position, Trends, and Challenges.” IEEE su10072212.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 551

Minguella-Canela, J., S. M. Planas, J. R. G. Ayats, and M. A. S. Peltz, J. F. . 1987. Sylmar Firm Puts Hot Idea for Model-Making
L
opez. 2019. “Study and Comparison of the Different Costs' Machine on the Market Via Canada. Los Angeles Times,
Schema Associated to Geometry, Material and Processing October 11, 1987. p. OC-D13.
Between 3D Printing, Injection Molding and Machining Petrick, I. J., and T. W. Simpson. 2013. “3D Printing Disrupts
Manufacturing Technologies.” Procedia Manufacturing 41: Manufacturing: How Economies of One Create New Rules of
280–7. https://doi.org/10.1016/j.promfg.2019.09.010. Competition.” Research‐Technology Management 56(6): 12–6.
Molitch-Hou, M. 2017. “Michelin and Fives' AddUp: From Making https://doi.org/10.5437/08956308x5606193.
Tires to Making Metal 3D Printers.” January 30, 2017. https:// Pradel, P., Z. Zhu, R. Bibb, and J. Moultrie. 2018. “Investigation of
www.engineering.com/story/michelin-and-fives-addup-from- Design for Additive Manufacturing in Professional Design Prac-
making-tires-to-making-metal-3d-printers. tice.” Journal of Engineering Design 29(4–5): 165–200. https://
Moore, G. E. 2006. “Moore's Law at 40.” In Understanding Moore's doi.org/10.1080/09544828.2018.1454589.
Law: Four Decades of Innovation, edited by D. C. Brock, 67–84. Prater, T. J., Q. A. Bean, R. D. Beshears, T. D. Rolin, N. J.
Philadelphia: CHC Press. Werkheiser, E. A. Ordonez, R. M. Ryan, and F. E. Ledbetter,
Mubarak, M. F., and M. Petraite. 2020. “Industry 4.0 Technologies, III. 2016. “Summary Report on Phase I Results From the 3D
Digital Trust and Technological Orientation: What Matters in Printing in Zero G Technology Demonstration Mission.” https://
Open Innovation?” Technological Forecasting and Social Change ntrs.nasa.gov/citations/20160008972.
161: 120332. https://doi.org/10.1016/j.techfore.2020.120332. Prencipe, A., A. Davies, and M. Hobday, eds. 2003. The Business of
Muir, M., and A. Haddud. 2017. “Additive Manufacturing in the Systems Integration. Oxford: Oxford.
Mechanical Engineering and Medical Industries Spare Parts Sup- Quah, D. 2003. “Digital Goods and the New Economy.” In New
ply Chain.” Journal of Manufacturing Technology Management Economy Handbook, edited by D. Jones, 289–321. San Diego:
29(2): 372–97. https://doi.org/10.1108/JMTM-01-2017-0004. Academic Press.
Mulder, K., and M. Knot. 2001. “PVC Plastic: A History of Systems Raasch, C., C. Herstatt, and K. Balka. 2009. “On the Open Design of
Development and Entrenchment.” Technology in Society 23(2): Tangible Goods.” R&D Management 39(4): 382–93. https://doi.
265–86. https://doi.org/10.1016/s0160-791x(01)00013-6. org/10.1111/j.1467-9310.2009.00567.x.
Mussatto, A., R. Groarke, R. K. Vijayaraghavan, C. Hughes, M. A. Radstok, E. 1999. “Rapid Tooling.” Rapid Prototyping Journal 5(4):
Obeidi, M. N. Dogu, M. A. Yalçin, P. J. McNally, Y. Delaure, 164–9. https://doi.org/10.1108/13552549910295488.
and D. Brabazon. 2022. “Assessing Dependency of Part Proper- Rajkumar, R. ( Raj), I. Lee, L. Sha, and J. Stankovic. 2010. “Cyber‐
ties on the Printing Location in Laser-Powder Bed Fusion Metal Physical Systems.” In Proceedings of the 47th Design Automation
Additive Manufacturing.” Materials Today Communications 30: Conference. https://doi.org/10.1145/1837274.1837461.
103209. https://doi.org/10.1016/j.mtcomm.2022.103209. Randhawa, K., J. West, K. Skellern, and E. Josserand. 2020. “Evolv-
Naik, H. S., A. Fritzsche, and K. M. Moeslein. 2021. “Modularity in ing a Value Chain to an Open Innovation Ecosystem: Cognitive
Making: Simplifying Solution Space for User Innovation.” R&D Engagement of Stakeholders in Customizing Medical
Management 51(1): 57–72. https://doi.org/10.1111/radm.12427. Implants.” California Management Review 63(2): 101–34.
Nasiri, M., M. Saunila, J. Ukko, T. Rantala, and H. Rantanen. 2020. https://doi.org/10.1177/0008125620974435.
“Shaping Digital Innovation Via Digital-Related Capabilities.” Rayna, T., and L. Striukova. 2015. “Open Innovation 2.0: Is Co‐crea-
Information Systems Frontiers. https://doi.org/10.1007/s10796- tion the Ultimate Challenge?” International Journal of Technol-
020-10089-2. ogy Management 69(1): 38. https://doi.org/10.1504/ijtm.2015.
Negroponte, N. 1996. Being Digital. New York: Vintage. 071030.
Nelson, R. R., and S. G. Winter. 1982. An Evolutionary Theory of Rayna, T., and L. Striukova. 2016. “From Rapid Prototyping to Home
Economic Change. Cambridge, MA: Belknap Press. Fabrication: How 3D Printing is Changing Business Model Inno-
Nerlich, B. 2005. “From Nautilus to Nanobo(a)ts: the visual con- vation.” Technological Forecasting and Social Change 102: 214–
struction of nanoscience.” AZojono: Journal of Nanotechnology. 24. https://doi.org/10.1016/j.techfore.2015.07.023.
https://doi.org/10.2240/azojono0109. Rayna, T., and L. Striukova. 2021. “Involving Consumers: The Role
NIST. 2020. “Cyber-Physical Systems, National Institute of Standards of Digital Technologies in Promoting ‘Prosumption’ and User
and Technology.” U.S. Department of Commerce, March Innovation.” Journal of the Knowledge Economy 12(1): 218–37.
16, 2020. https://web.archive.org/web/20200316224708/https:// https://doi.org/10.1007/s13132-016-0390-8.
www.nist.gov/el/cyber-physical-systems. Rayna, T. 2008. “Understanding the Challenges of the Digital Econ-
Nonaka, I., and H. Takeuchi. 2007. “The Knowledge-Creating Com- omy: The Nature of Digital Goods.” Communications and Strat-
pany.” Harvard Business Review 85(7/8): 162. egies 71: 13–6.
Nylén, D., and J. Holmström. 2015. “Digital Innovation Strategy: A Rayna, T., L. Striukova, and J. Darlington. 2015. “Co-Creation and
Framework for Diagnosing and Improving Digital Product and User Innovation: The Role of Online 3D Printing Platforms.”
Service Innovation.” Business Horizons 58(1): 57–67. https:// Journal of Engineering and Technology Management 37: 90–102.
doi.org/10.1016/j.bushor.2014.09.001. https://doi.org/10.1016/j.jengtecman.2015.07.002.
Nyman, H. J., and P. Sarlin. 2014. “From Bits to Atoms: 3D Printing Rindfleisch, A., M. O'Hern, and V. Sachdev. 2017. “The Digital
in the Context of Supply Chain Strategies.” In Proceedings of Revolution, 3D Printing, and Innovation as Data.” Journal of
the 47th Hawaii International Conference on System Sciences Product Innovation Management 34(5): 681–90. Portico. https://
4190–9. https://doi.org/10.1109/HICSS.2014.518. doi.org/10.1111/jpim.12402.
Osterwalder, A., and Y. Pigneur. 2004. “An Ontology for E-Business Rindfleisch, A., and M. O’Hern. 2015. “Brand Remixing: 3D Print-
Models.” In Value Creation From E-Business Models, edited by ing the Nokia Case.” Review of Marketing Research: 53–81.
W. Currie, 65–97. Oxford: Butterworth-Heinemann. https://doi.org/10.1108/s1548-643520150000012003.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
552 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Rogers, E. M., and F. F. Shoemaker. 1971. Communication of Inno- Stanko, M. A. 2020. “Building an Understanding of How Winning
vations: A Cross-Cultural Approach. New York: FreePress. Products Emerge When Open and Proprietary Products Coex-
Rouf, S., A. Raina, M. Irfan Ul Haq, N. Naveed, S. Jeganmohan, ist: Evidence From the RepRap Community.” Creativity and
and A. Farzana Kichloo. 2022. “3D Printed Parts and Mechani- Innovation Management 29(3): 398–412. https://doi.org/10.
cal Properties: Influencing Parameters, Sustainability Aspects, 1111/caim.12376.
Global Market Scenario, Challenges and Applications.” Stanko, M. A., and B. J. Allen. 2022. “Disentangling the Collective
Advanced Industrial and Engineering Polymer Research. 5: Motivations for User Innovation in a 3D Printing Community.”
143–58. https://doi.org/10.1016/j.aiepr.2022.02.001. Technovation 111: 102387. https://doi.org/10.1016/j.
Rouhani, C. . 2018. “Three BIG Challenges for Using FDM 3D Print- technovation.2021.102387.
ing at Scale.” Sculpteo.com. August 2, 2018. https://www. Stone, P., R. Brooks, E. Brynjolfsson, R. Calo, O. Etzioni, G. Hager,
sculpteo.com/blog/2018/08/02/three-big-challenges-for-using- J. Hirschberg, et al. 2016. Artificial Intelligence and Life in 2030:
fdm-3d-printing-at-scale/. The One Hundred Year Study on Artificial Intelligence. Stanford,
Söderberg, J., and A. Daoud. 2012. “Atoms Want to Be Free Too! CA: Stanford University.
Expanding the Critique of Intellectual Property to Physical Sturgeon, T. J. 2019. “Upgrading Strategies for the Digital Econ-
Goods.” TripleC: Communication, Capitalism & Critique. omy.” Global Strategy Journal 11(1): 34–57. Portico. https://doi.
Open Access Journal for a Global Sustainable Information org/10.1002/gsj.1364.
Society 10(1): 66–76. https://doi.org/10.31269/triplec. Thürer, M., Y. Huang, and M. Stevenson. 2021. “Workload Control
v10i1.288. in Additive Manufacturing Shops Where Post-Processing is a
Sahini, D. K., J. Ghose, S. K. Jha, A. Behera, and A. Mandal. 2020. Constraint: An Assessment by Simulation.” International Jour-
“Optimization and Simulation of Additive Manufacturing Pro- nal of Production Research 59(14): 4268–86. https://doi.org/10.
cesses: Challenges and Opportunities—A Review.” In Additive 1080/00207543.2020.1761038.
Manufacturing Applications for Metals and Composites, edited Thomas-Seale, L. E. J., J. C. Kirkman-Brown, M. M. Attallah, D. M.
by K. R. Balasubramanian and V. Senthilkumar, 187–209. Her- Espino, and D. E. T. Shepherd. 2018. “The Barriers to the Pro-
shey, Penn: IGI Global. gression of Additive Manufacture: Perspectives From UK
Salmi, M., K.-S. Paloheimo, J. Tuomi, J. Wolff, and A. Mäkitie. Industry.” International Journal of Production Economics 198:
2013. “Accuracy of Medical Models Made by Additive 104–18. https://doi.org/10.1016/j.ijpe.2018.02.003.
Manufacturing (Rapid Manufacturing).” Journal of Cranio- Tilbury, D. M. 2019. “Cyber‐Physical Manufacturing Systems.”
Maxillofacial Surgery 41(7): 603–9. https://doi.org/10.1016/j. Annual Review of Control, Robotics, and Autonomous Systems
jcms.2012.11.041. 2(1): 427–43. https://doi.org/10.1146/annurev-control-053018-
Sandberg, J., J. Holmström, and K. Lyytinen. 2020. “Digitization 023652.
and Phase Transitions in Platform Organizing Logics: Evidence Turner, F. 2010. From Counterculture to Cyberculture: Stewart
from the Process Automation Industry.” MIS Quarterly 44(1): Brand, the Whole Earth Network, and the Rise of Digital Utopi-
129–53. https://doi.org/10.25300/MISQ/2020/14520. anism. Chicago: University of Chicago Press.
Schaller, R. R. 1997. “Moore's Law: Past, Present and Future.” IEEE van den Ende, J., G. van de Kaa, S. den Uijl, and H. J. de Vries.
Spectrum 34(6): 52–9. https://doi.org/10.1109/6.591665. 2012. “The Paradox of Standard Flexibility: The Effects of Co‐
Schubert, C., J. Sydow, and A. Windeler. 2013. “The Means of Man- evolution Between Standard and Interorganizational Network.”
aging Momentum: Bridging Technological Paths and Organisa- Organization Studies 33(5–6): 705–36. https://doi.org/10.1177/
tional Fields.” Research Policy 42(8): 1389–405. https://doi.org/ 0170840612443625.
10.1016/j.respol.2013.04.004. Veers, P., K. Dykes, E. Lantz, S. Barth, C. L. Bottasso, O. Carlson,
Sculpteo. 2020. The State of 3D Printing 2020. San Francisco: A. Clifton, et al. 2019. “Grand Challenges in the Science of
Sculpteo. Wind Energy.” Science 366(6464): eaau2027.
Sgarbossa, F., M. Peron, F. Lolli, and E. Balugani. 2021. “Conven- Walsh, G. S., J. Przychodzen, and W. Przychodzen. 2017. “Support-
tional or Additive Manufacturing for Spare Parts Management: ing the SME Commercialization Process: The Case of
An Extensive Comparison for Poisson Demand.” International 3D Printing Platforms.” Small Enterprise Research 24(3):
Journal of Production Economics 233: 107993. https://doi.org/ 257–73. https://doi.org/10.1080/13215906.2017.1396490.
10.1016/j.ijpe.2020.107993. Wang, L., S. Jiang, and S. Zhang. 2020. “Mapping Technological
Shane, S. 2000. “Prior Knowledge and the Discovery of Entrepre- Trajectories and Exploring Knowledge Sources: A Case Study
neurial Opportunities.” Organization Science 11(4): 448–69. of 3D Printing Technologies.” Technological Forecasting and
https://doi.org/10.1287/orsc.11.4.448.14602. Social Change 161: 120251. https://doi.org/10.1016/j.techfore.
Shapiro, C., and H. R. Varian. 1999. Information Rules: A Strategic 2020.120251.
Guide to the Network Economy 2163. Boston, MA: Harvard Busi- Wang, Q., X. Sun, S. Cobb, G. Lawson, and S. Sharples. 2016. “3D
ness School Press. Printing System: An Innovation for Small-Scale Manufacturing
Shen, X. 2019. “Building ‘Mass’ and ‘Momentum’—A Latecomer in Home Settings?—Early Adopters of 3D Printing Systems in
Country's Passage to Large Technological Systems—The Case China.” International Journal of Production Research 54(20):
of China.” Technology Analysis & Strategic Management 31(5): 6017–32. https://doi.org/10.1080/00207543.2016.1154211.
503–16. https://doi.org/10.1080/09537325.2018.1519184. West, J. 2008. “Commercializing Open Science: Deep Space Com-
Stanko, M. A. 2016. “Toward a Theory of Remixing in Online Inno- munications as the Lead Market for Shannon Theory, 1960–
vation Communities.” Information Systems Research 27(4): 73.” Journal of Management Studies 45(8): 1506–32. https://doi.
773–91. https://doi.org/10.1287/isre.2016.0650. org/10.1111/j.1467-6486.2008.00807.x.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 553

West, J. 2014. “Too Little, Too Early: California's Transient Advan- Paradigm With Lifecycle Personalization Concerns.” Interna-
tage in the Photovoltaic Solar Industry.” The Journal of Tech- tional Journal of Production Research 57(8): 2571–84. https://
nology Transfer 39(3): 487–501. https://doi.org/10.1007/s10961- doi.org/10.1080/00207543.2018.1530475.
012-9291-6.
West, J., and G. Kuk. 2016. “The Complementarity of Openness:
How MakerBot Leveraged Thingiverse in 3D Printing.” Techno- AUTHOR BIOGRAPHIES
logical Forecasting and Social Change 102: 169–81. https://doi.
org/10.1016/j.techfore.2015.07.025. Thierry Rayna is Chaired Professor of Technology
West, J., and M. Bogers. 2014. “Leveraging External Sources of
for Change at the École Polytechnique and a Fellow
Innovation: A Review of Research on Open Innovation.” Jour-
nal of Product Innovation Management 31(4): 814–31. Portico.
of the CNRS Innovation Interdisciplinary Institute
https://doi.org/10.1111/jpim.12125. (i³), France. His research investigates the changes
West, J. 2006. “Does Appropriability Enable or Retard Open Inno- brought about by digital technologies (such as 3D
vation?” In Open Innovation: Researching a New Paradigm, edi- printing, artificial intelligence, blockchain, cloud, IoT,
ted by H. Chesbrough, W. Vanhaverbeke, and J. West, 109–33. platforms) on business strategy, organization, and
Oxford: Oxford University Press. public policy. In particular, his research focuses on the
Westerweel, B., R. J. I. Basten, and G.-J. van Houtum. 2018. “Tradi-
issues of business model innovation, ecosystem gover-
tional or Additive Manufacturing? Assessing Component
nance, and impact (environmental, social, economic),
Design Options through Lifecycle Cost Analysis.” European
Journal of Operational Research 270(2): 570–85. https://doi.org/ in relation to digital technologies. Prof. Rayna has a
10.1016/j.ejor.2018.04.015. long track record of collaboration with companies
Windrum, P., K. Frenken, and L. Green. 2017. “The Importance of (small and large) in various industries “disrupted” by
Ergonomic Design in Product Innovation. Lessons From the digital technologies, and policymakers.
Development of the Portable Computer.” Industrial and Corpo-
rate Change 26(6): 953–71. https://doi.org/10.1093/icc/dtx006. Joel West is Professor of Entrepreneurship at Hilde-
Wohlers, T. 2005. “New Trends and Developments in Additive Fab- gard College, and Professor Emeritus at the Keck
rication.” In Virtual Modeling and Rapid Manufacturing: Graduate Institute. His research has examined the
Advanced Research in Virtual and Rapid Prototyping, edited by impact of technological evolution upon industry struc-
P. B
artolo. London: Taylor & Francis. ture in personal computers, mobile phones, digital
Wohlers. 2012. Wohlers Report 2012. Ft Collins, CO: Wohlers communications, solar energy, and 3D printing. His
Associates.
has also pioneered the understanding of how firms
Wohlers. 2017. Wohlers Report 2017. Ft Collins, CO: Wohlers
Associates.
utilize strategic openness to build competitive advan-
Wohlers. 2020. Wohlers Report 2020. Ft Collins, CO: Wohlers tage through open platforms, open standards, open
Associates. communities, and open consortia. He is the co‐editor
Yoo, Y., O. Henfridsson, and K. Lyytinen. 2010. “Research Com- of Open Innovation: Researching a New Paradigm
mentary—The New Organizing Logic of Digital Innovation: An (2006), New Frontiers of Open Innovation (2014) and
Agenda for Information Systems Research.” Information Sys- the Oxford Handbook of Open Innovation (2023) and
tems Research 21(4): 724–35. https://doi.org/10.1287/isre.1100.
cofounder of the World Open Innovation Conference.
0322.
Yoo, Y., R. J. Boland, K. Lyytinen, and A. Majchrzak. 2012. “Orga-
nizing for Innovation in the Digitized World.” Organization Sci-
ence 23(5): 1398–408. https://doi.org/10.1287/orsc.1120.0771.
Zhan, T., K. Yin, J. Xiong, Z. He, and S.‐T. Wu. 2020. “Augmented SU PP O R TI N G I N F O RMA TI O N
Reality and Virtual Reality Displays: Perspectives and Chal- Additional supporting information can be found online
lenges.” IScience 23(8): 101397. https://doi.org/10.1016/j.isci. in the Supporting Information section at the end of this
2020.101397. article.
Zhang, J., F.‐Y. Wang, K. Wang, W.‐H. Lin, X. Xu, and C. Chen.
2011. “Data‐Driven Intelligent Transportation Systems: A Sur-
vey.” IEEE Transactions on Intelligent Transportation Systems How to cite this article: Rayna, Thierry, and
12(4): 1624–39. https://doi.org/10.1109/tits.2011.2158001. Joel West. 2023. “Where Digital Meets Physical
Zheng, P., T.-J. Lin, C.-H. Chen, and X. Xu. 2018. “A Systematic
Innovation: Reverse Salients and the Unrealized
Design Approach for Service Innovation of Smart Product-
Dreams of 3D Printing.” Journal of Product
Service Systems.” Journal of Cleaner Production 201: 657–67.
https://doi.org/10.1016/j.jclepro.2018.08.101. Innovation Management 40(4): 530–553. https://
Zheng, P., Y. Lin, C.-H. Chen, and X. Xu. 2019. “Smart, Connected doi.org/10.1111/jpim.12681
Open Architecture Product: An IT-Driven Co-Creation
Received: 31 May 2021 Revised: 29 October 2022 Accepted: 11 March 2023
DOI: 10.1111/jpim.12681

ORIGINAL ARTICLE

Where digital meets physical innovation: Reverse salients

and the unrealized dreams of 3D printing

Thierry Rayna 1 | Joel West 2,3

1

i³-CRG, Ecole Polytechnique, CNRS,
Institut polytechnique de Paris, Palaiseau, Abstract
France For more than three decades, enthusiasts have predicted that direct
2
Hildegard College, Costa Mesa, manufacturing enabled by 3D printing would inevitably supplant traditional
California, USA
3
manufacturing methods. Alas, for nearly as long, these utopian predictions
Riggs School of Applied Life Sciences,
Keck Graduate Institute, Claremont,
have failed to materialize. One reason is a flawed assumption that hybrid
United States digital-physical systems such as 3D printing would advance as rapidly as
purely digital innovations enabled by Moore's law. Instead, like other exam-
Correspondence


Thierry Rayna, i³-CRG, Ecole ples of cyber-physical systems (CPSs), technological progress in 3D printing
Polytechnique, CNRS, Institut faces inherent limitations that are emblematic of the differences between
polytechnique de Paris, Palaiseau, France.
CPSs and purely digital innovations. As with any complex CPS, improved
Email: [email protected]
performance of a 3D printing system has been limited by that of its key
Special Issue Guest Editors: Michael components—the sort of limiting problem previously defined as a reverse
Stanko and Aric Rindfleisch
salient. Unlike previously studied technologies, several reverse salients for
3D printing performance have neither resolved nor signs of resolving soon.
Here we analyze these key reverse salients, and show how they have ham-
pered the suitability of 3D printing for direct manufacturing and other pre-
dicted applications. We contrast predicted versus actual capabilities for 3D
printing-enabled transformation in six key areas: product innovation, mass
customization, home fabrication, distributed manufacturing, supply chain
optimization and business model innovation. From this, we suggest opportu-
nities for greater realism in future 3D printing research, as well as broader
implications for our understanding of CPSs and reverse salients.

KEYWORDS
3D printing, cyber-physical systems, digital innovation, innovation, reverse salients

1 | INTRODUCTION synthesizer,” a 23rd-century technology that allowed

food to be manufactured on demand for crew members
In September 1966, viewers of the American science light years from home. In 1987, the next Star Trek
fiction series Star Trek were introduced to the “food series featured a voice-controlled “replicator,” a 24th-
century technology that could make almost anything.
Thierry Rayna and Joel West contributed equally to this study. These thus popularized the dream of on-demand

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2023 The Authors. Journal of Product Innovation Management published by Wiley Periodicals LLC on behalf of Product Development & Management Association.

530 wileyonlinelibrary.com/journal/jpim J Prod Innov Manag. 2023;40:530–553.

 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 531

computer-controlled production of arbitrary physical

objects.1 Practitioner points
Also in 1987, in a Los Angeles suburb 20 miles from
Star Trek's Hollywood studio, a new firm named • 3D printing points out the risk of planning
3D Systems, Inc. produced and sold the world's first addi- around utopian predictions of technology
tive manufacturing system. Based on a technology called performance.
stereolithography—patented by founder Chuck Hull— • We show how persistent reverse salients can
these systems could create complex 3D shapes by use of a slow the performance and adoption of a hybrid
computer-controlled laser that solidified liquid resin. digital + physical technology.
DTM Corporation was formed that same year to commer- • We offer a framework for identifying which
cialize selective laser sintering, a UT Austin invention by systems and systems components are most
Carl Deckard that created 3D objects by fusing metal likely to face these persistent performance
powder.2 These two firms thus pioneered what eventually limitations.
became known as 3D printing.
As in their sci-fi archetypes, such products offered
the promise to render arbitrary digital designs into a transformation made possible by the adoption of digital
tangible form. For more than 30 years, these complex computing and communications enabled by large-scale
and expensive systems have been predicted to inevitably circuit integration, rapid improvements in computing
transform manufacturing industries (e.g., Ben-Ner & performance and the malleability of software (Schaller,
Siemsen, 2017; Ghaffar & Mullett, 2021; Li, 2013; 1997; West, 2008).
Peltz, 1987; Petrick & Simpson, 2013). For example, in a We thus consider the impact of the limits of digital
2011 story entitled “Could 3D printing change the and physical technologies that are inherent for digital
world?” a leading U.S. foreign policy think tank breath- systems that serve physical beings who are living in a
lessly predicted: material world. We do so for a broad class of digital-
physical hybrid innovation known as cyber-physical sys-
3D Printing/Additive Manufacturing (AM) is tems (CPSs) that integrate digital and physical compo-
a revolutionary emerging technology that nents in a single system (Lee, 2008; Rajkumar
could up-end the last two centuries of et al., 2010). The study of CPSs often emphasizes adding
approaches to design and manufacturing digital intelligence and connectivity to mature physical
with profound geopolitical, economic, social, technologies, but here we study the case of 3D printing,
demographic, environmental, and security in which digital technologies are combined with imma-
implications (Campbell et al., 2011, p. 1). ture physical technologies.
We begin the article by reviewing how theories of
Two years later, President Barack Obama predicted in his both digital innovation and CPSs explain innovation tra-
2013 State of the Union address that 3D printing “has the jectories for systems that combine digital and physical
potential to revolutionize the way we make almost components. We then consider how these trajectories are
everything.” often limited by a critical technological bottleneck that
Instead, the reality of 3D printing has fallen short of Thomas Hughes (1983, 1987) defines as a reverse salient.
these early visions, as technological performance and This article examines 3D printing as a specific case study
thus adoption have consistently failed to meet predictions for understanding the reverse salients of CPSs. Using more
during these years. Such hype is not unique to 3D print- than 80 articles from scientific journals on digital manufactur-
ing, as many forms of digital innovation have attracted ing and other applications of 3D printing, we trace the pro-
technologically utopian visions to mobilize public sup- gress and limitations of both the digital and physical
port and adoption (Kling, 1996; Turner, 2010). However, components of this technology over the past 35 years, and
purely digital innovations have fulfilled such visions identify the key reverse salients for 3D printing. From this,
during this same period, as witnessed by the global we offer broader implications and research opportunities for
3D printing, CPSs and the study of reverse salients.
1
The first known mention of additive manufacturing in science fiction
came in a 1947 short story in Astounding magazine by Eric Frank
Russell, which referred to “atom fed to atom like brick after brick to
2 | L I T E R A T UR E R E V I E W
build a house” (Nerlich, 2005).
2
3D Systems acquired DTM in 2001, the first of many such acquisitions.
Deckard died in 2019, while Hull remains executive vice president of 3D In recent decades, our world has been transformed by
Systems. ubiquitous digital innovation. One widely cited definition
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
532 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

of such innovation is “the carrying out of new combina- physical, one option is that the physical enables the digi-
tions of digital and physical components to produce novel tal. The most dramatic example of ongoing improve-
products” (Yoo et al., 2010, p. 725). However, subsequent ments in semiconductor manufacturing under Moore's
researchers concluded “What is missing to date is an law have made predictable and rapid improvements in
understanding of how digital transformation manifests computing performance (Moore, 2006; Schaller, 1997)—
itself in industries in which the core products are primar- in turn enabling advances in digital innovation (Autio
ily physical” (Hanelt et al., 2015, p. 1313). We therefore et al., 2021), However, in this article, we focus on the
are interested in this question: How do technological lim- converse case: how the digital enables the physical. An
itations (specifically reverse salients) constrain the tech- examination of digitally enabled technologies of the late
nological trajectory of a system that requires both digital 20th and early 21st century suggests three different pat-
and physical innovations? We also consider what hap- terns for how the digital enables the physical (Table 1):
pens when such limits are either not recognized, or are
assumed to be temporary. 1. Digital represents the physical, where the digital good
We thus review three prior bodies of research that is an intermediate representation of a good that will
relate to digital + physical innovation. The first is how be consumed in physical form. This includes objects
digital innovation considers systems combining digital printed on demand, whether a book or a spare part
and physical components, while the second is research (Jiang et al., 2017; Osterwalder & Pigneur, 2004).
on CPSs. We extend this with prior research on reverse Information goods such as text, pictures and
salients in physical and digital systems to suggest how audiovisual material are a special case, because the
reverse salients could impact the performance trajectory value is the information contained in the object
of a CPS such as 3D printing. rather than the object itself (Negroponte, 1996;
Rayna, 2008).
2. Digital coordinates the physical: digital technologies
2.1 | The promise of integrating digital are used to coordinate activities in the physical world,
and physical innovation whether e-mail, a project plan or an ecosystem map
(Boland et al., 2007; Markus, 1994; Randhawa
Digital technology removes the existing constraints of et al., 2020). For example, to design and make physi-
physical innovation and makes possible a great degree of cal objects, the open-source hardware movement has
redefining and reconfiguring the nature of the innovation used many of the same collaboration processes, social
(Rindfleisch et al., 2017; Yoo et al., 2012). The malleabil- norms and IP rules as those used for producing open
ity of digital technologies thus enables experimentation source software, a purely digital good (Raasch
and improvisation in a way not possible with physical et al., 2009; West & Kuk, 2016).
innovation (Nylén & Holmström, 2015), which has pow-
erful implications for the nature of innovation itself.
First, these reduced constraints—plus the use of a com-
mon digital interface for a wide range of problem T A B L E 1 Taxonomy of digital innovations that depend on the
domains—mean that the pace of innovation has acceler- physical world.
ated (Yoo et al., 2010).
Category Example
Second, digital innovations are often not governed by
Digital represents Print-on-demand 2D objects
the well-known S-curves limiting performance trajecto-
the physical (Osterwalder & Pigneur, 2004) and
ries of physical technologies (Priestly et al., 2020). Even if
3D objects (Jiang et al., 2017);
such S-curves exist for specific domains for key enabling Entertainment and other information
technologies—such as quantum computing (Hill & goods (Negroponte, 1996;
Rothaermel, 2003) or artificial intelligence (AI) (Iansiti & Rayna, 2008); innovation as data
Lakhani, 2020)—advances in general purpose enabling (Rindfleisch et al., 2017)
technologies will accelerate societal innovation more Digital coordinates E-mail (Markus, 1994); boundary objects
broadly than previous S-curves related to product- or the physical (Boland et al., 2007; Randhawa et al.,
industry-specific technologies such as aircraft or wind 2011); open source hardware (Raasch
turbines. et al., 2009)
While digital technologies also impact the nature of Digital controls Manufacturing (Åstebro, 2002) and
physical innovation, that intersection remains a rarely the physical transportation systems (Zhang
explored frontier (Hanelt et al., 2015; Lee & Berente, et al., 2011); robots (Fujita, 2001;
Sturgeon, 2019)
2012; Sturgeon, 2019). When the digital meets the
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 533

3. Digital controls the physical: digital technologies con- mention issues related to digital technologies, regulation
trol activities in the physical world—such as complex and legal issues, or the need to find new business models.
manufacturing systems (Zhang et al., 2011), self- Rarely considered issues are arising from the physical
driving cars (Lyytinen, 2022), and personal or indus- aspects of the CPS itself.
trial robots (Fujita, 2001; Sturgeon, 2019). The focus on the cyber part of a CPS has often been
for the control of updated versions of existing
While all three patterns apply to 3D printing, here we physical objects, such as home appliances, thermostats,
focus on the technological challenges of how the digital automobiles or factory equipment (Borgia, 2014; Hanelt
controls the physical. By its nature, the final results of et al., 2015; Lee et al., 2015). Despite this focus on mature
such this sort of system will be delivered (and con- technologies and equipment, there are examples of CPSs
strained) by the physical processes that are digitally that rely on physical components that are yet immature.
controlled. An emerging example is the so-called metaverse, which
requires affordable, lightweight, durable and high-
definition virtual reality headsets. Another example is 3D
2.2 | Cyber-physical systems printing, which has remained an emerging technology
for decades, and can thus help us understand the role
The most commonly studied examples of innovations similar limitations can have for other CPSs.
where the digital controls the physical are CPSs, first
defined as “integrations of computation with physical
processes” (Lee, 2008, p. 363). The name captures how 2.3 | Reverse salients in physical
the “cyber-world of computations and communications” and CPSs
integrated with “real-world physical activities” (Lien
et al., 2012). A more recent definition states that CPS Innovation in physical technologies often follows a punc-
“comprise interacting digital, analog, physical, and tuated equilibrium model, which assumes a long period
human components engineered for function through of incremental innovation within a given technological
integrated physics and logic” (NIST, 2020). paradigm, interrupted by infrequent discontinuities that
Examples of CPS research have included the Inter- start a new paradigm (Dosi, 1982; Garcia & Calantone,
net of Things (IoT) (Baars et al., 2014; Mahmood & 2002; Nelson & Winter, 1982). Technological progress
Mubarik, 2020) and Industry 4.0 (Greer et al., 2019; within a given paradigm will follow a specific trajectory
Lee et al., 2015). While innovation is often mentioned (Dosi, 1982; Nelson & Winter, 1982) that is only clear
in the academic literature related to CPS, the innova- after the fact. The revealed trajectory depends on a com-
tion processes are rarely studied. Those who do eso bination of market, technical and organizational factors
xplicitly, typically raise the issue of CPS as opportuni- (Bijker & Law, 1994, p. 18; Kim & Kogut, 1996, p. 285).
ties for Open Innovation (Mubarak & Petraite, 2020; Complex innovations are often delivered as systems
Zheng et al., 2019), service and product innovation that integrate a range of components and technologies.
(Li, 2018; Li et al., 2021; Mubarak & Petraite, 2020; While such component reuse offers powerful economies
Zheng et al., 2018), and business model innovation of scale, it provides challenges for the firm both in attain-
(Li et al., 2018; Mahmood & Mubarik, 2020). ing economic success, and also for managing the com-
However, the innovation potential of CPS identified plexity and interdependencies of the various components
in the literature chiefly relates to progress made and (Davies & Brady, 2000; Majidpour, 2016; Prencipe
brought about by the digital side, as with studies of AI, et al., 2003; West, 2006).
IoT or blockchain (Bongo et al., 2020; Li, 2018; Li et al., Within these systems, the technological progress on one
2021). In these studies, the physical aspects of the CPS or more of these components will lag the rest of the system
appear to have been largely overlooked. and thus be the limiting factor of improving overall system
This emphasis on the digital side is prominent in performance. By analogy to a military front, Hughes (1983,
research that bridges theories of digital innovation and 1987) terms such a performance-limiting component a
CPSs, which emphasize the new capabilities and relaxing reverse salient, which Lehenkari & Miettinen (2002, p. 111)
of design constraints made possible by digital control of later summarized as “when one system component—
existing systems (Hendler, 2019; Herterich et al., 2016; technical, organizational or economic—has fallen behind
Lyytinen, 2022; Sandberg et al., 2020). In fact, articles or is out of phase with the others, and impedes the further
investigating either success factors or challenges and bar- development of the system.”
riers to CPS adoption (e.g., Bongo et al., 2020; Leitão Here our interest is in the former case, specifically that
et al., 2020; Sinclair et al., 2019) appear to chiefly subset of technological bottlenecks for components that
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
534 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

limit the performance improvement of the entire system. changes to data exchange interfaces that constrained
Hughes (1983, 1987) suggests that the image of a “bottle- performance.
neck” is a less accurate metaphor for such limits than a Other reverse salients can only be resolved through
salient within a broad technological front. Some have trea- changes to sociotechnical institutions. For example,
ted the overlapping concepts of bottleneck and reverse when compatibility standards both provide interoperabil-
salient as being related issues in technological change ity but constrain innovation, improvements to the shared
(Baldwin, 2015; Windrum et al., 2017) while others have standard can become the limiting factor for improving
used the terms synonymously (e.g., Davies, 1996). performance, such as in digital telecommunications
For Hughes, the solution to such a salient comes from (Dedehayir & Hornsby, 2008; West, 2008). When a
entrepreneurs, inventors and other scientists who define reverse salient is deeply embedded in society, overcoming
“the reverse salient as a set of ‘critical problems,’” and that salient is often more complex to overcome and thus
then work to correct those problems (Hughes, 1983, requires a more comprehensive approach to resolve
p. 80). This definition is fueled by the assumption that (Mathews et al., 2018; Shen, 2019).
“such problems are believed to be solvable. The belief Overall, studies of specific reverse salients have tended
that a reverse salient might be corrected increases dra- to emphasize successfully overcoming the salient. Such
matically by its articulation in a set of critical problems” research includes salients that are “solved” (Christiansen &
(Mulder & Knot, 2001, p. 266). While studies by Hughes Buen, 2002; Windrum et al., 2017) or “resolved” (Sandberg
and many others typically emphasize absolute perfor- et al., 2020) to achieve the system's “potential” (Daim
mance, the critical problem can also be one of efficiency et al., 2014; Dedehayir & Mäkinen, 2011b; Sandberg et al.,
or cost effectiveness, particularly for industries that 2020); this may reflect a long-remarked tendency toward
require economies of scale and scope (Chandler, 1990; success bias in studies of innovation (Coopersmith, 2020;
Davies, 1996). Rogers & Shoemaker, 1971). Thus there is a gap in research
The original theory was developed from Hughes' stud- that considers salients that are not resolved or overcome.
ies of the electrical power infrastructure, but the concept
subsequently demonstrated its relevance in a wide array
of sectors and applications, mainly railroads and rail- 3 | REVERSE SALIENTS IN 3D
ways, aviation, electric vehicles, ICTs, production, tele- PRIN TING
phone, and wireless services (Dedehayir, 2009). In
particular, subsequent studies of physical systems have We are interested in how the digital and physical
examined renewable power generation (Christiansen & components of a CPS—and the interaction between these
Buen, 2002; Hoppmann et al., 2014) and airplane engines components—serve as reverse salients for that system's
(Bowles & Dawson, 1998). While many studies have technological trajectory. We do so in the context of a spe-
focused on a specific salient, a few have examined the cific CPS: 3D printing. Studying 3D printing is an
impact of multiple salients on technological progress extreme case of CPS due to the perfect integration
(Dedehayir & Hornsby, 2008; Sandberg et al., 2020; required between the digital and physical to realize the
Windrum et al., 2017). longstanding promise of a one-to-one correspondence
More recent studies have examined reverse salients of between the digital and physical worlds. Such a “bits to
digital innovation. Since the 1980s, CPU speed has been atoms” solution would fulfill the science fiction dream of
the limiting factor for most general- and special-purpose materializing any digital object as a physical one (Millard
computing systems (Daim et al., 2014; Dedehayir & et al., 2018; Nyman & Sarlin, 2014; Söderberg & Daoud,
Mäkinen, 2011a), while specialized applications have 2012). While originally the best-known subset of a broader
been limited by graphic performance (Dedehayir & category of technologies termed additive manufacturing,
Mäkineif, 2008) or portability (Windrum et al., 2017). In the term 3D printing has since been commonly used to
fact, the PC industry suggests that different types of refer to the entire category (e.g., Wohlers, 2005).
reverse salient may emerge over time, not only between Consistent with other reverse salient studies
different hardware components (Dedehayir & Mäkinen, (e.g., Lehenkari & Miettinen, 2002; Lenfle & Söderlund,
2011a), but also between hardware and software compo- 2022; Sandberg et al., 2020), we identify reverse salients
nents (Dedehayir & Mäkinen, 2011b), making identifying using a longitudinal case study that includes economic
and forecasting reverse salients particularly critical and and technological factors that shaped the evolution of
difficult (Han & Shin, 2014). In the Sandberg et al. (2020) this technology over the past 35 years. Much like these
study of digitalization of previously physical-only process earlier studies, we used a combination of (a) research on
control systems—a rare study of reverse salients in the history of the science and innovation in this field;
CPSs—tight coupling of system components limited (b) scientific literature, in this case from manufacturing
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 535

and materials science journals; (c) economic and business material (such as metal or wood) is cut down to final shape
analyses of 3D printing technology and applications; and using cutting devices such as a saw, lathe or mill. While
(d) comprehensive industry reports—particularly the the CNC process worked with materials that had been
Wohlers Report—about industry statistics and trends. Our used for centuries, as in the pre-CNC era it remained lim-
analysis was informed by previous interviews from our ited by the mechanics of the cutting devices. Initial CNC
respective earlier studies of 3D printing. adoption was also limited by the quality and availability of
The predicted rapid adoption of 3D printing digital Computer Aided-Design (CAD) software, as well as
technologies has failed to materialize due to how they dif- the employee skills necessary to use this software to create
fer from other digital technologies. Indeed, the optimistic digital designs (Åstebro, 2002; Tilbury, 2019).
predictions for 3D printing all implicitly assumed that it Thus nascent additive manufacturing technologies of
would follow the same trends, adoption and use patterns the 1980s could build upon the CAD software developed
as purely digital technologies. But as a CPS, 3D printing for CNC applications. Early 3D printers were first con-
systems require a complex integration of a broad range of trolled by dedicated CAD workstations and later using
both digital and physical technologies. Because improve- personal computers (Jacobs, 1992; Shane, 2000). They ini-
ments in physical component technologies have not kept tially faced the same limitations of CAD software and
up with the digital components, they act as reverse designer skills, but digitally enabled advances in CAD
salients for the overall performance—in terms of speed, software improved the usability of these tools (de Jong &
quality, cost and degree of automation—of how 3D print- de Bruijn, 2013).
ing can create physical artifacts. Table 2 summarizes six
reverse salients of 3D printing, of which five are tied to
the physical side of 3D printing. 3.1.2 | Consumer design tools
First, we examine the digital and physical reverse
salients limiting improvements in 3D printing perfor- The original CAD software was solely intended for profes-
mance, and we then consider the effects of those limits sionals, usually with a drafting or engineering background.
upon the societal and economic transformations Even with improvements in usability for professionals in
predicted to occur through widespread adoption of the early 21st century, further improvements are still
3D printing. needed to make 3D designing as usable for ordinary con-
sumers as 2D design tools (Rayna et al., 2015). For exam-
ple, Digital Forming, a London startup founded in 2008 to
3.1 | Digital components of 3D printing address this problem, promised that “our software har-
nesses the power of mass customization”; instead, the
3.1.1 | Design tools company was declared insolvent in 2019.

To design physical objects, 3D printing build upon one of

the earliest CPSs, Computer Numerical Control. With the 3.1.3 | Rendering layers
invention of the microprocessor, starting in 1972, CNC sys-
tems provided digital control of the mature and reliable Compared to CNC and other subtractive manufacturing,
subtractive model of manufacturing, where a raw block of 3D printing allows the design of arbitrary shapes made by

TABLE 2 Limits and reverse salients in 3D printing.

Category Limitation Current status Salient

Digital Professional design tools Mature
Consumer design tools Slowly improving X
Rendering layers Mature
Distribution of digital designs Mature
Physical Replicability: repeatability and reproducibility Ongoing challenges X
Indirect replicability: digitization Slowly improving X
Costs Slowly improving X
Time Slowly improving X
Skills and knowledge Ongoing challenges X
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
536 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

adding one layer upon top of another. However, it required 3.2.1 | Direct replicability of artifacts
invention of entirely new material science technologies for
fabricating physical objects, as well as digital algorithms to Replicability is a serious and persistent challenge for 3D
translate a 3D digital design into a series of 2D layers, printing systems. Aggregating molecules of a particular
printed one at a time. While the basic layer printing issues material layer by layer never happens the exact same way
were addressed within a decade, repeatability requires twice. These inconsistencies in printing the same digital
developing precision solutions for each new manufacturing object arise even with similar printer technologies, and
material (Gibson et al., 2014; Jacobs, 1992). sometimes with two successive prints from the same
printer and materials (Dowling et al., 2020; Kim
et al., 2018). Efforts to accurately predict the output (a 3D
3.1.4 | Digital distribution print preview) depend on physical phenomena taking
place at molecular level and today require computations
The digital designs for 3D printing are digital artifacts, by a supercomputer (Goh et al., 2020; Sahini et al., 2020).
which by their nature are stored and replicated without a The direct replicability of 3D printing assumed in
loss of information (Quah, 2003; Rayna, 2008). Since the prior forecasts require both repeatability and reproduc-
advent of digitalization in the late 20th century, the stor- ibility (Khorasani et al., 2021; Mussatto et al., 2022; Salmi
age, duplication and eventual dissemination of informa- et al., 2013). Repeatability means creating identical
tion goods has moved from the analog to the digital objects using the same operator, environment, and equip-
and from the physical to the electronic (Rayna & ment (e.g., 3D printer, software, materials) in a short
Striukova, 2021; Shapiro & Varian, 1999). Online com- interval of time. Reproducibility means obtaining the
munities were created to share or sell these 3D printable same results with different operators, materials, and
designs, of which the best known and most often studied equipment. The latter is still a dream, but even the for-
is Thingiverse (Claussen & Halbinger, 2021; Kyriakou mer remains unsolved (Dowling et al., 2020).
et al., 2017; Naik et al., 2021; Stanko & Allen, 2022; A final failure of utopian forecasts for materializing
West & Kuk, 2016). physical objects is that 3D printed objects typically
While such digital designs can be cheaply, reliably require manual post-processing, such as sanding, polish-
and quickly replicated, the same is not true for the physi- ing or even painting; sometimes these steps can only be
cal object 3D printed from such design. In principle, this done by hand (Sgarbossa et al., 2021; Thürer et al., 2021).
distinction applies to the materialization of other digital Some designs also require post-processing assembly.
goods such as a photo or sound recording, but such infor- Beyond the increase in time and cost, this nonautomated
mation goods can be viewed or played on a computer post-processing reduces the consistency of replicating 3D
without being materialized. printed artifacts.

3.2 | How the physical limits digital 3.2.2 | Digitalization and indirect
innovation replicability

Printing arbitrary 3D objects required developing new Whether by CAD design or digitization, digital files used
materials and material science. During the industry's for 3D printing typically only describe the outer shape of
first 15 years, 10 new patent families (by 10 different the object. The internal structure (including wall thick-
inventors) were introduced, each creating a different ness) needed to instantiate the object is generated prior
materials technological trajectory and seeking to over- to each print, depending on the software, the printer, the
come the limitations of other 3D printing technologies; materials and user settings. While solid objects are the
however, all had significant limitations and none cap- norm with traditional fabrication processes such as injec-
tured a dominant market position (West & Kuk, 2016). tion molding and milling, the opposite is true for 3D
The proliferation of technologies, patents and firms con- printing, where printing solid objects is slow, expensive
tinued during the initial decades of the 21st century, and failure prone (Rouf et al., 2022).
with a common core of patents shaping the entire indus- For printing digital objects, indirect replicability
try until 2008 (Wang et al., 2020). Despite this high rate would mean that copies of copies are not only identical
of innovation and new entry, these technologies have to each other, but also to the original (Quah, 2003;
failed to overcome the cost and speed advantages that Rayna, 2008). Unlike digitizing media streams or 2D
high-volume manufacturing provides for most applica- objects, commercial and consumer 3D scanning tools are
tions (Jiang et al., 2017). decades behind 2D counterparts (Conkle et al., 2018):
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 537

widely available 3D scanners can copy the outer layer of technologies. Just as physical limits mean that semicon-
the object, but not its internal structure, leaving unre- ductor manufacturing faces inextricable challenges at the
solved how the object's structure should be replicated nanoscale (Khan et al., 2018), the lack of progress in
(Javaid et al., 2021). While the photograph-scanning 3D printing speed for at least some of the technology
model has been effective for digitizing what objects look branches is due to physical constraints that will be diffi-
like, it will not tell designers how to replicate those cult to resolve, in part due to the unpredictable nature of
aspects of an object's internal structure that determine its the melting/congealing processes required by most 3D
strength, weight and other key properties.3 Even with a printing technologies (Böckin & Tillman, 2019).
3D scan that produces an accurate copy of the original
artifact, the information embedded in this scanned blue-
print would most likely be radically differ from the origi- 3.2.5 | Human skill and knowledge
nal blueprint due to inherently missing information.
Replication of digital objects has been enabled by auto-
mated processes that minimize the required human
3.2.3 | Costs knowledge and skills. Extrapolated to 3D printing, this
brought predictions that consumers would be able to
In addition to high quality, purely digital artifacts can be manufacture objects at home instead of purchasing them
replicated at a negligible cost by reusing digital storage in a store, as soon as 3D printers became widely available
for other purposes. However, 3D printing requires raw (Bogers et al., 2016; Jiang et al., 2017; Rayna &
materials that cannot be reused and are difficult to recy- Striukova, 2015). However, after decades 3D printing still
cle (Ford & Despeisse, 2016). Not only does 3D printing remains a long way from providing fully automated fabrica-
an object have a significant cost, but that cost is tion and “zero skill manufacturing” (Eyers & Potter, 2017).
irrecoverable. Because of the lack of consistency and reliability, the use
Unlike smartphones and other general-purpose com- of 3D printers requires specific craft-like knowledge for both
puting devices, 3D printing also requires specialized up effective design and manufacturing (Despeisse & Minshall,
front capital investments. Not only does 3D printing 2017; Pradel et al., 2018). Different settings are required for
require a 3D printer, but today such printer is only suit- each combination of printer and material—and often for
able for printing a particular type of object with a specific specific objects. As noted in Rouhani (2018, p. 17):
material such as metal, ceramics, or plastic. In contrast to
the multifunctionality of digital devices, the idea of a Mastering FDM 3D printing requires a lot of
single 3D printer that can automatically manufacture tinkering and patience. There are many print
arbitrary objects remains a utopian fantasy for the fore- quality issues common to FDM such as
seeable future. warping, step lines, stringing, under/over
extrusion, layer shifting, and many, many
more. While there are solutions to all of
3.2.4 | Time these problems, it can take multiple attempts
in order to get the desired finished look.
Often overlooked is that the replication of 3D printed FDM machines have a lot of moving parts
objects also requires significant time. Replicating a digital and settings which need to be adjusted
artifact is usually a matter of seconds. 3D printing time according to the geometry of the part, the
depends on the shape, size, geometry and material, but environment the printer is in, and the
typically a small object will require 1–3 h, a medium expected print quality.
sized object 12–48 h, while the largest may take up to a
week (Avinson, 2018). Some improvements have been made—for example,
While print speed for some materials has improved in self-calibrating printers—but 3D printing is unlikely to
recent years, overall speed of 3D printing technologies become “plug and play” in the foreseeable future.
has not followed a similar pattern as other digital Instead, the modest successes of 3D printing have been
enabled by tacit knowledge developed and applied by
3
In principle, the issue can be solved with a medical grade CT or MRI skilled artisans. Just as in 19th century machine tools
machine that can render (and thus replicate) a single item made of
and 20th-century CNC machines, the 21st-century opera-
homogeneous materials such as a replacement bone implant
(Randhawa et al., 2021). However, even this technology is unable to
tion of additive manufacturing systems requires skilled
automatically detect and replicate the structure (let alone materials) of a workers who know their machine and can tell what will
complex assembled product. or will not work. And while digital knowledge is reliably
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
538 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

and costlessly replicated, such tacit knowledge is inher- manufactured by any subtractive (e.g., milling) or
ently neither (Nonaka & Takeuchi, 2007) and is also diffi- transformative (e.g., injection molding) manufacturing
cult to transfer: solving one 3D printing problem does not technology.
directly solve another, although experience provides a
starting point for diagnosing and addressing problems. These factors are tied to the application of 3D printing,
whether building prototypes (Rapid Prototyping), casts,
molds and jigs (Rapid Tooling), end-user parts or objects
3.3 | Effect of 3D printing reverse (Direct Manufacturing), or for Distributed Manufacturing
salients (Rayna & Striukova, 2016).

Due to these limits, after three decades of optimistic

predictions, additive manufacturing has failed to displace 4 | A S S E S S I N G PR E D I C T I O N S
potential niche applications addressed by CNC O F 3 D P R I N T I N G - E N A B LE D
machines—let alone the broader market for high volume TR ANSFORMATIONS
formative manufacturing technologies such as injection
molding or casting. As one frustrated potential user said Using the framework a for understanding the reverse
in 2016: “At this stage, AM machines are a waste of salients of 3D printing developed in Section 3, to reassess
money due to immature technology. These machines extant knowledge and generate new insights for under-
have very high operation and maintenance costs. There- standing the phenomenon we conducted a problematiz-
fore, we are not investing any money on AM technology, ing review of research on 3D printing applications
and the bulk of our income is from CNC machining” (as detailed in Appendix A). From the corpus of 84 jour-
(Wohlers, 2017, p. 170). nal articles shown in Appendix B, we identified six major
Meanwhile, 3D printing failed to realize the promise themes for predicted societal or economic impacts of 3D
of large-scale distributed replication that is today possible printing (Table 3). For each theme, we show how the
for the digital representations but not for the physical impact of 3D printing in that area has been limited by
objects created from those representations—which would one or more of the previously mentioned reverse salients
be necessary achieve predicted transformational out- in 3D printing performance.
comes. This reveals a sharp divide between the physical
side and the digital side, because the overall performance
of the technology is limited by the reverse salients of the 4.1 | Product innovation
physical side.
Given these inherent physical limitations, 3D printing Product innovation is one of the most frequently cited
is only competitive under one or more of the following transformational impact of 3D printing. In practice, this
conditions: has meant Rapid Prototyping (RP), once the only applica-
tion of 3D printing (Hull et al., 1995), and still its most
• Short lead time: despite its comparatively slow speed, prevalent use (Sculpteo, 2020; Wohlers, 2020), because
3D printing has an advantage when time constraints limitations outldeined in Section 3 do not act as reverse
are very tight, as it does not require tools to be built salients for making prototypes.
prior to manufacturing. This applies only when a small RP has had critical benefits for product innovation
number of units are needed, after which the faster pro- (Cooper, 2021; Haleem et al., 2020; Luchs et al., 2015;
duction speed of injection molding will make up for Rayna & Striukova, 2021), especially when done collec-
the time required to build tools (Achillas et al., 2015; tively (Bogers & Horst, 2014). However, these benefits are
Blakey-Milner et al., 2021; Chaudhuri et al., 2021). being eroded by advances in virtual reality (VR), aug-
• Small batches: the lack of tooling cost makes 3D print- mented reality (AR), and digital twin technologies
ing particularly attractive when small numbers of units (Coutts et al., 2019).
are produced. However, the (global) lack of economies Product innovation can also be accelerated by Rapid
of scale for 3D printing compared to traditional Tooling (Fontana et al., 2019; Wohlers, 2020)—printing
manufacturing (Baumers & Holweg, 2019), makes it molds, casts or jigs used in conventional high-volume
less competitive (ceteris paribus) when more than a manufacturing techniques (Radstok, 1999; Rayna &
few dozen to hundred units have to be manufactured Striukova, 2016). This combines advantages of estab-
(Minguella-Canela et al., 2019) lished manufacturing processes (cost, speed, and mate-
• High complexity: 3D printers enable the manufacture rials) and 3D printing (cheaper and more complex molds
of designs of such complexity that they could not be and tools). For instance, 3D printed tire molds enabled
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 539

TABLE 3 Societal impacts of 3D printing predicted by prior research.

Societal impact Prior research Main themes

Product Weller et al. (2015), Stanko (2016), Flath et al. (2017), Disruptive innovation; innovation from data; rapid
innovation Murmura and Bravi (2017), Rindfleisch et al. and iterative prototyping done collectively; real-
(2017), Steenhuis and Pretorius (2017), Kwak time customer feedback; complex tooling to
et al. (2018), Marzi et al. (2018), Candi and enable product innovation; design freedom; 4D
Beltagui (2019), Steenhuis et al. (2020), designs; new product development; no-assembly
Rayna and Striukova (2021b), Ukobitz and products; process innovation; product remixes;
Faullant (2021) technological development; open innovation;
open source
Mass Petrick and Simpson (2013), Rayna et al. (2015), Weller Customer and consumer involvement in design and
customization/ et al. (2015), Bogers et al. (2016), Rayna and Striukova manufacturing; economies of one; knowledge
co-creation (2016), Oettmeier and Hofmann (2016), Steenhuis and reuse; large-scale product customization and
Pretorius (2016), Attaran (2017), Khorram Niaki and co-creation; platforms; pricing; value creation
Nonino (2017), Kyriakou et al. (2017), Aquino et al.
(2018), Shukla et al. (2018), Chaney et al. (2021),
Lacroix et al. (2021)
Prosumption/ Fox (2014), Appleyard (2015), Rayna et al. (2015), Bogers DIY; consumer behavior and adoption; prosumption
home et al. (2016), Kothman and Faber (2016), Rayna and (consumers involved in production process); home
fabrication Striukova (2016), Rogers et al. (2016), Steenhuis and fabrication with 3D printers; intellectual property;
Pretorius (2016), West and Kuk (2016), Wang et al. open design; online design and manufacturing
(2016), Yoo et al. (2016), Despeisse et al. (2017), Flath communities; move from e-commerce to file-
et al. (2017), Jiang et al. (2017), Halassi et al. (2019), commerce; piracy; social manufacturing; user
Kleer and Piller (2019), Naghshineh et al. (2021), Rayna innovation
and Striukova (2021a)
Distributed Garrett (2014), Gress and Kalafsky (2015), Ford and Advanced manufacturing; circular economy;
manufacturing/ Despeisse (2016), Huang et al. (2016), Laplume developing countries; digital encapsulation;
global value et al. (2016), Sasson and Johnson (2016), Despeisse production ecosystems; global factory; industry
chains et al. (2017), Durão et al. (2017), Garmulewicz et al. 4.0; internationalization; value chain
(2018), Hannibal and Knight (2018), Rehnberg and reconfiguration; resource efficiency; product
Ponte (2018), Xu et al. (2018), Kleer and Piller life cycle; geographic span and density;
(2019), Li et al. (2019), Sauerwein et al. (2019), economies of scale and scope; long tail; make-
Beltagui et al. (2020a, 2020b), Hannibal (2020), to-order: manufacturing supercenters;
Rinaldi et al. (2022) organizational structure; restructuring; spare
parts; sustainability
Logistics/supply Baumers et al. (2015), Bogers et al. (2016), Glocalized production; barriers; facility location;
chains Holmström et al. (2016) Sandström (2016), hybrid manufacturing; intellectual property rights;
Oettmeier and Hofmann (2016), Rogers et al. integration; inventories; logistics management;
(2016), Durach et al. (2017), Rylands et al. (2016), SMEs; supply chain innovation; supply chain
Li et al. (2017), Ryan et al. (2017), Schniederjans flexibility supply chain management components
(2017), Chan et al. (2018), Ghadge et al. (2018), and processes; supply chain performance; supply
Eyers et al. (2018), Martinsuo and Luomaranta chain strategy; supply chain reconfiguration;
(2018), Halassi et al. (2019), Luomaranta and responsiveness; remanufacturing; reverse logistics;
Martinsuo (2019), Tziantopoulos et al. (2019), value streams
Zanoni et al. (2019), Delic and Eyers (2020),
Ghobadian et al. (2020), Strong et al. (2020),
Choudhary et al. (2021), Engelseth et al. (2021),
Hecker (2021)
Enabling new Baumers et al. (2015), Prause (2015), Bogers et al. (2016), Business model dynamics; business ecosystems;
business Ford and Despeisse (2016), Rayna and Striukova (2016), capabilities; digital entrepreneurship; disruptive
models Rogers et al. (2016), Steenhuis and Pretorius (2016), innovation; industrial sustainability; nascent
Holzmann et al. (2017), Khorram Niaki and Nonino industries; platforms; SMEs; technology adoption;
(2017), Kwak et al. (2018), Rong et al. (2018), Xu et al. user entrepreneurship; value creation; value
(2018), Holzmann et al. (2020b), Rayna and Striukova capture; value chain reconfiguration
(2021b)

Note: All references shown in this table are provided in Appendix B.

 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
540 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Michelin to develop its CrossClimate/All Season product technologies (Rayna & Striukova, 2021), this also largely
line (Molitch-Hou, 2017). failed to materialize. Despite an early enthusiasm in the
However, this prevents reaping the full benefits of 3D 2010s, consumer 3D printing faces multiple reverse
printing for product innovation. As emphasized explicitly salients—unreliability, relative high cost, post-processing,
(Rayna & Striukova, 2021) or implicitly (Bogers skills and competencies—that have hindered professional
et al., 2016; Ford & Despeisse, 2016; Kleer & Piler, 2019; direct manufacturing applications. Many experts consider
Rindfleisch et al., 2017) in the literature, the benefits only that, aside from small maker (hobbyist) communities
happen when actual products (or a significant part of (Mavri et al., 2021), consumer 3D printing was indeed a
them) are directly manufactured (DM) with 3D printers. fad (Blueprint, 2020; Wohlers, 2020).
This provides digital freedom for physical products: pre- Consumer 3D printing uptake was a key requirement
assembled designs (Haleem et al., 2020), 4D printed prod- for the predicted emergence of transformational con-
ucts (Lee et al., 2017), or products with complex internal sumer communities that would jointly innovate, remix
structure that weight a fraction of usual products (Feng and tinker, share and distribute products and artifacts
et al., 2018; Wohlers, 2020). (Bogers et al., 2016; Jiang et al., 2017; Rayna et al., 2015;
Unfortunately, the reverse salients are also most preva- Rayna & Striukova, 2016; Rayna & Striukova, 2021), as
lent with DM: risk of low replicability, more complex (and has happened with other digital technologies. Such com-
costly) the post-processing (because of the complex struc- munities have emerged, but have had a limited impact,
ture), and requiring new skills (Rylands et al., 2016; since so few people have access to a 3D printer. The
Thomas-Seale et al., 2018). Consequently, DM has majority of 3D printing platforms identified by Rayna
remained confined to a few niche markets—medical/dental et al. (2015) have disappeared or become 3D printing ser-
prosthetics, aerospace and defense (Sculpteo, 2020; vice bureaus (e.g., 3DHub, Sculpteo, i.Materialize). Even
Wohlers, 2020). Other envisaged applications have largely on the popular Thingiverse online community, objects
been abandoned, preventing a key transformation of prod- regularly appear but the top 20 most fabricated ones are
uct innovation predicted by prior research: the real-time more than 5 years old, and contrary to expectations
integration of customer feedback to continuously improve (Stanko & Allen, 2022), have led to few remixes and are
products (Rayna & Striukova, 2016; Rayna & Striukova, seldom remixes themselves.
2021; Walsh et al., 2017; West & Bogers, 2014). Likewise, the seminal RepRap Open-Source Hardware
project that gave rise to MakerBot, Ultimaker, and dozens
of other startups (de Bruijn, 2010; Greul et al., 2018; Hu &
4.2 | Consumer involvement: Sun, 2022; Stanko, 2020; Stanko & Allen, 2022; West &
Customization, co-creation, and Kuk, 2016) has proved unable to move beyond (the rela-
prosumption tively unsophisticated) plastic filament printing technol-
ogy. Limited consumer adoption has dashed widespread
The literature also widely predicted the involvement of hopes firms might leverage 3D printing user communities
consumers and other customers in the production pro- (de Jong & de Bruijn, 2013; Rayna et al., 2015; Rayna &
cess, either in design (mass customization, co-creation) Striukova, 2016; Rindfleisch et al., 2017).
or manufacturing (prosumption, home fabrication). Also unrealized is a predicted shift from e-commerce
Because of higher costs, mass customization via DM to “file-commerce” (products purchased online to be
requires a higher willingness to pay for customized prod- printed at home) and 3D printing that would dramati-
ucts (Franke & Piller, 2004). Yet, actual willingness to cally change how products are developed, distributed,
pay appear to vary greatly (Kalantari et al., 2021), so the and acquired (Jiang et al., 2017). Consequently, the once
reverse salient for DM costs make mass customization predicted IP crisis from consumer piracy of objects
mainly viable in the medical sector (Wohlers, 2020). (Birtchnell et al., 2018; Jiang et al., 2017; Mendis
Mass customization also requires specific skills for et al., 2019; Rayna & Striukova, 2016), has also failed to
both designers and users. After failed efforts to provide materialize.
easy-to-use interfaces, mass customization has generally
been restricted to pseudo-customization (choice within a
range of pre-set attributes) where 3D printing faces 4.3 | Distributed manufacturing and
cost and time disadvantages compared to regular global value chains
manufacturing.
A more radical predicted impact was that consumers Unlike other manufacturing technologies with significant
would manufacture products with their own 3D printers. economies of scale, 3D printing is characterized by a con-
While in line with the success of other digital stant unit cost, which provides far fewer reasons to
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 541

concentrate production geographically or temporally. component are key benefits of 3D printing identified in
Thus, a key impact predicted by the literature is a shift the literature (Beltagui et al., 2020a); a famous example is
toward distributed manufacturing and, consequently, a the claim made by GE to have simplified a jet engine noz-
redefinition of global value chains. zle from 20 parts to one, and, in another case 855 parts to
Digital encapsulation (Holmström et al., 2016) just 12 3D printed parts (GE, 2020). On-demand produc-
through home 3D printing (Despeisse & Minshall, 2017; tion can reduce costs through reduced waste, as well as
Jiang et al., 2017; Rayna & Striukova, 2016) would be the lower assembly, storage and transportation costs, hereby
most extreme form of distributed manufacturing, but is saving fuel (only raw materials transported) and reducing
impractical without widespread consumer 3D printing supply chain and logistics costs (since fewer parts are
uptake. Even RepRap printers, originally designed to be involved) (Ford & Despeisse, 2016).
self-replicable—by having a 3D printer make many of the Literature has hypothesized benefits of on-demand
next printer's parts (Rayna et al., 2023; West & production for sustainability (Ford & Despeisse, 2016;
Kuk, 2016)—are generally purchased as mass-produced Gebler et al., 2014; Rouf et al., 2022), chiefly focusing on
kits or fully assembled printers. spare part production. Yet, the use of 3D printing to pro-
Distributed manufacturing is theoretically possible: duce spare parts remains rate and claims of superiority of
firms could share production facilities and use service 3D printing for spare part manufacturing (Li et al., 2017)
bureaus, while large companies could internally distrib- appear to be limited to specific niches (Frandsen
ute manufacturing across different locations, which et al., 2020) where reverse salients are not as relevant;
would improve sustainability by reducing transport even with production costs four times higher, less than
costs (Ford & Despeisse, 2016; Gebler et al., 2014). Yet, 10% of spare parts are worth 3D printing (Heinen &
we were unable to find in any significant application of Hoberg, 2019). Post-processing further worsens the 3D
this concept, beside the well-publicized (but brief) pro- printing cost disadvantage (Heinen & Hoberg, 2019),
duction of medical equipment during the first months while the environmental improvement over traditional
of the COVID-19 pandemic (Budinoff et al., 2021; manufacturing appears moderate to negligible (Böckin &
Frazer, 2022). Tillman, 2019).
Recent literature suggests that even in best-case scenar- Lack of tooling (Westerweel et al., 2018) is predicted
ios, the alleged superiority of distributed manufacturing to enable shorter lead-times, but this has little impact
may not stand scrutiny. Li et al. (2019) found that, contrary overall (Knofius et al., 2019), because of the reverse
to the existing perception, distributed manufacturing does salient (comparatively slow speed of and higher
not always guarantee a quick response, and that pooling manufacturing cost) makes that, unless unique parts are
effect makes that centralized manufacturing performs bet- involved. Otherwise, 3D printing's lead-time advantage
ter when the demand rate is relatively high. Furthermore, rapidly wears off (Muir & Haddud, 2017).
Rinaldi et al. (2022) found the benefits of distributed In regard to logistics and supply chains, 3D printing's
manufacturing rather limited even when average demand ability to manufacture shapes that embed several parts
is rather low, as centralized manufacturing provides a bet- into one (consolidation) has also been emphasized. Yet,
ter capacity utilization. this often leads to a higher total cost (Knofius et al.,
This contrast between predictions and the reality can 2019)—consolidation worsens the impacts of wear or fail-
be largely attributed to the reverse salients. When physi- ure, while greater design complexity means higher
cal properties (structural or aesthetical) of a product post-processing costs (Knofius et al., 2019; Sgarbossa
matter, distributed manufacturing requires a very tight et al., 2021).
control of the manufacturing process, which makes 3D Overall, the benefits of 3D printing on supply chain
printing a QA nightmare. In a world where transport and and logistics appear chiefly limited to niches where the
logistics are rapid, efficient, and cheap, the (mainly) cen- reverse salients do not matter as much: inherently low
tralized manufacturing model is hard to beat. The physi- volumes, no post-processing, and parts that do not wear
cal limitations of 3D printing stand in the way of the out, for example, highly specialized low-volume applica-
long-predicted global value chain reconfiguration. tions, such as industrial equipment, manufacturing
at sea, or remote supply depots for combat ships
(Kenney, 2013)—or in space, as demonstrated in 2014 by
4.4 | Logistics and supply chains NASA using an FDM 3D printer to produce spare parts
on the International Space Station (Prater et al., 2016), in
On-demand production and ability of manufacture com- which case the cost of getting cargo to outer space out-
plex designs to consolidate several parts into a single weigh the limitations of 3D printing.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
542 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

4.5 | Enabling business model 4.6 | Reevaluating societal impacts of 3D

innovation printing

Beside investigating business models for the 3D printing The literature shows a prevalent belief that 3D printing
industry (e.g., Holzmann et al., 2020), literature has pro- represents the next stage of digital innovation, blurring the
jected that 3D printing would radically reconfigure busi- boundaries between digital and physical, hereby opening
ness models in physical sectors, just like what has an entire new realm of innovation (Aryan et al., 2021;
happened in the digital realm, whether by choice or Nasiri et al., 2020; Rindfleisch et al., 2017). This brought
necessity due to consumer piracy of objects (Appleyard, predictions that the physical world would witness a free-
2015; Birtchnell et al., 2018; Depoorter, 2014; Mendis dom of creation similar to the digital one, enabling new
et al., 2019). 3D printing would not only change business modes of digital innovation that would “digitize physical
models—leading to more consumer-centric business goods much like the first digital revolution digitized infor-
models involving consumers in the manufacturing supply mation goods” (Rindfleisch et al., 2017, p. 681) and trans-
chains (Bogers et al., 2016), with, over time, a prolifera- form product development and innovation, value chain,
tion of “business models that involve increased levels inventory management, customization and business model
of customer self-service” (Bogers et al., 2016, p. 12), innovation.
but business model innovation in itself (Rayna & Aside from very specific niches, the transformative
Striukova, 2016). impacts of 3D printing identified in the literature have
Yet, beside occasional business model upgrades, it is largely failed to materialize, because of the physical limi-
hard to find cases of radical business model reconfigura- tations of 3D printing that act as reverse salients. These
tion through 3D printing. The RepRap Open-Source physical reverse salients have prevented a widespread
Hardware project has enabled shifts between open adoption of 3D printing by firms and individuals—a pre-
and closed business models (Stanko, 2020; West & requisite to most of transformative impact predictions, as
Kuk, 2016), as well as business models blending these it would enable firms to gather early customer product
two modes (Appleyard & Chesbrough, 2017; Rayna et al., feedback (West & Bogers, 2014), harvest data through 3D
2023). Even in this case, only startups have carried out printing to better understand user needs (Rindfleisch
bold business model innovation (Greul et al., 2018), with et al., 2017) and leverage users and user communities to
incumbent perpetuating essentially the same business innovate derivatively through 3D printing (de Jong & de
model. Bruijn, 2013; Rindfleisch et al., 2017). Such adoption
Expectations of transformative impact on business would lead to a potential large-scale shift to online reposi-
models were largely based on the premise of a large cus- tories for product design (Jiang et al., 2017) and bypass
tomer adoption, and involved mass customization, dis- traditional manufacturing via digital encapsulation
tributed manufacturing and reconfigured global supply (Holmström et al., 2016). In turn, this would force rethink-
chains, or on-demand production. Because of the ing of the role of brands (Rindfleisch & O'Hern, 2015;
reverse salients, all have been confined to narrow Wang et al., 2016). Widespread consumer adoption would
niches. It is not hard to understand why the predicted lead to the rise of intermediaries and platforms that would,
radical business model shifts have not materialized. The on the positive side, facilitate innovation and, on the nega-
related literature generally ignores the reverse salients tive side, bring a predicted IP doomsday when IP protec-
and instead emphasizes the benefits brought about by tion of physical goods would suddenly become as
the digital side of 3D printing, with the implicit assump- challenging as for other digital goods (Jiang et al., 2017).
tion that 3D printed artifacts behave just like purely dig- As industry experts now admit (Wohlers, 2020), wide-
ital artifacts. spread adoption is unlikely to arise until reverse salients
Even in cases when 3D printing was largely adopted, have been resolved: with decades of hindsight, predic-
business models may largely remain the same. The main- tions and expectations were simply unrealistic. Conse-
stream uptake of 3D printing in the hearing aid industry quently, discussions surrounding what is the correct
in the 1990s has still not brought radical business model business model for products marketed through 3D
reconfiguration (Sandström, 2015). Despite widespread printing—such as pricing per print, subscription or open
adoption, the very same networks of firms are involved access model (Bogers et al., 2016)—remain largely
in the value creation, with no radical shift in value net- irrelevant. Even predictions that did not require a large
works, such as consumer involvement or other changes consumer adoption have largely remained niche applica-
described in Bogers et al. (2016) and Rayna and Striukova tions, such as 3D printing mass customized medical
(2016). implants (Randhawa et al., 2020).
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 543

The same reverse salients have also impeded the pre- materialize due to persistent limitations of 3D printing
dicted expansion of distributed manufacturing, and the performance. The traditional response in prior research
advent of dispersed 3D printing micro-production centers has been that those limitations would soon be solved:
located close to customers (Luchs et al., 2015), which, in
turn, would have distributed innovation beyond a single By [comparing] direct digital manufacturing
firm. Consequently, questions of best practices for collabo- with traditional tool-based manufacturing, we
rative innovation with distributed production—particularly show that direct digital manufacturing so far
the role of digital manufacturing in B2B collaboration lags behind by several orders of magnitude
between established firms and startups and SMEs, also have compared to traditional manufacturing
been pushed into the future, except for niches such as methods. Yet we also find that direct digital
defense, aerospace and industrial equipment. manufacturing clearly is on an improvement
This consistent failure to meet expectations has led to a trajectory that eventually will see it being able
reevaluation some of the most optimistic predictions of 3D to compete with traditional manufacturing on
printing use. This can be seen in the content of the influen- a unit cost basis (Holmström et al., 2016, p. 1).
tial Wohlers Report—a yearly report by industry experts
and key source of information about the 3D printing As we have shown, the manufacturing (i.e., physical)
industry. Between the Wohlers (2012) and Wohlers (2020) aspects of 3D printing have been the reverse salient for its
annual reports, most (9 of 17) of the predicted of 3D print- use and adoption for three decades: while closing this “sev-
ing applications—avatars and figurines; consumer-created eral orders of magnitude” gap has long been a priority, on
products; fashion products; furniture, home, and office the current trajectory the eventual ability to compete with
accessories; gifts, trophies, and memorials; marketing and traditional manufacturing remains decades away.
advertising; museum displays; musical instruments; and For future research, refractory reverse salients of 3D
personal 3D printers—were dropped. Applications com- printing require a new research approach. Rather than
mon across both periods are essentially those already pre- assume an imminent (but still unfulfilled) “hockey-stick”
dicted 25 years ago: aerospace, architectural modeling, art inflection in performance or adoption, researchers should
and jewelry, automotive, machinery, medical applications, instead focus on what is known to be technologically
prototyping, tooling. feasible. By doing so, they will produce more realistic
Industry experts appear to have finally realized that (and relevant) analyses that considers the impact of these
the physical reverse salients of 3D printing were not tem- limitations, and how persistent reverse salients can be cir-
porary impediments, but were instead intrinsic to 3D cumvented with more attainable technological solutions.
printing: building objects layer by layer usually requires To ground research in reality, two different paths lie
repeated but unique reactions at molecular level, leading ahead for 3D printing researchers. The first requires revi-
to complexity, unpredictability, comparatively high siting 3D printing predictions for each of the six themes
energy usage and slow speed. In fact, it can be argued of Table 3, and, accounting for the actual limitations,
that the very limitations of 3D printing lie in its very investigate which predictions remain realistic. In particu-
strength. As a result, these reverse salients are unlikely to lar, this means moving away from the focus on direct
be resolved any time soon. manufacturing, which will remain a niche use.
Mainstream uses of 3D printing remain prototyping
and tooling, so an immediate research opportunity is
5 | FUTURE RESEARCH how product design best practices can leverage both. This
is particularly important at a time when virtual prototyp-
Given the enduring reverse salients of 3D printing, how ing technologies—such as augmented or virtual reality—
does that change future research? Here we consider the are rapidly increasing in performance and use. These
implications of this for the future of digital manufactur- newer technologies may undercut prior 3D printing suc-
ing research, and more broadly for our understanding of cesses, as digital substitutes evolve more quickly than
digital innovation, CPSs, and reverse salients, concluding physical-limited processes. Likewise, fast moving high-
with managerial implications of this research. flexibility robotic automation may reduce the appeal of
rapid tooling.
A second research path would study how other digital
5.1 | Implications for 3D printing and technologies might help overcome the physical limita-
digital manufacturing tions of 3D printing. For instance, combining 3D printing
with other digital manufacturing technologies (e.g., CNC,
As analyzed in Section 4, research on 3D printing has laser cutting) could alleviate the speed and post-
brought high expectations that have repeatedly failed to processing issues that plague 3D printing, while using AI
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
544 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

in design may diminish the complexity and time needed • Business model innovation: Besides direct manufactur-
to customize 3D printed items. Augmented reality could ing and home fabrication, could rapid prototyping and
finally provide the intuitive and easy-to-use user interface tooling enable reconfigured or entirely new business
(equivalent the mouse and windows for PCs) that is models? Are there disruptive (Christensen, 1997) busi-
needed to fully leverage 3D printing technologies. ness models for direct or local manufacturing
Overall, both research directions are examples of how analogous to how lower quality MP3 music disrupted
to reduce the impact of persistent reverse salients on music consumption in the 1990s?
improving the performance and economic impact of 3D
printing. Future research and commercial applications
cannot ignore these reverse salients, but instead must 5.2 | Implications for CPSs and digital
consider those trajectories that either minimize or over- innovation
come their impact.
Considering the main themes identified in the litera- Considering the differences of the digital and physical
ture (Table 3), opportunities for further research include: models suggests that we need a more nuanced model of
innovation and technological change in today's digital
• Product innovation: Assuming direct manufacturing world, particularly with regard to reverse salients. In
will remain a niche application, how can 3D printed Table 4, we compare the rate of innovation and reverse
tooling enable complex molds and gigs that can be salients for all possible combinations of digital and physi-
used as a source of product innovation? How can com- cal innovations: the two extremes of purely digital or
bining 3D printing with virtual prototyping technolo- physical systems, and four possible configurations of
gies be leveraged to facilitate user feedback in near CPSs. We contrast the emerging technologies (where dra-
real-time? matic performance improvements are still possible) with
• Consumer involvement: If consumer adoption of 3D mature technologies that have approached the limits of
printing remains low, how can combining these tech- their technological trajectory.
nologies enable large-scale co-creation with con- The Purely Digital configuration corresponds to the
sumers? Could digital technologies supplant 3D digital innovation research of the past decade. In most
printed physical prototypes, thereby enabling online cases, the malleability and rapidly improving perfor-
communities to utilize digital twins that provide a true mance of digital systems have enabled dramatic progress
correspondence between the digital artifacts and their in such systems. A notable exception is artificial intelli-
physical materialization? Could mass-customization gence, where the rate of progress has been “patchy and
serve intermediate-sized global markets with similar unpredictable” (Stone et al., 2016, p. 4).
tastes, desires, or needs with 3D printing be used to Among CPSs, we identify four configurations.
decrease tooling cost to make economical micro-batch (1) Digital-Limited have mature physical technologies,
production? but still require dramatic improvements in the digital
• Distributed manufacturing and global value chains: technologies. (2) Emerging Digital and Physical, where
Assuming the limitations of direct manufacturing can- both modes are significant limiting factors, but neither
not be overcome, to which extent other digital technology has reached maturity. (3) Mature Digital and
manufacturing technologies (such as laser cutting or Physical technologies that are combined for an immedi-
CNC) could be combined with 3D printing in inte- ate wave of combinatorial (often digital) innovation
grated and universal manufacturing units in order to (Sandberg et al., 2020). However, once the combination
deliver the promise of distributed manufacturing, and, reaches maturity, subsequent improvements are slow and
consequently, enable more sustainable global value incremental. (4) Physical-Limited, where the rate of digi-
chain reconfigurations? tal innovation is much faster than the physical and thus
• Logistics and supply chains: Could integrated digital overall performance is limited by the physical.
manufacturing units be leveraged to deliver pseudo Finally, the Purely Physical systems are the familiar
on-demand manufacturing, possibly with the help of domain of innovation studies of the past 40 years—such as
rapid tooling enabling to produce just-in-time small railroads, power systems airplanes, automobiles and
batches production using traditional manufacturing? manufacturing processes (Anderson & Tushman, 1990;
Likewise, could rapid tooling meet very tight lead Hughes, 1983; Utterback, 1994). Today, most mechanical
times by utilizing local integrated digital manufactur- innovations have some digital component, but purely phys-
ing units? Also could hybrid production strategies ical innovation can be found in materials science—such as
manufacture parts less prone to wearing out through photovoltaic cells and modern battery materials—where
3D printing, while consumable parts remain mass the rate of progress remains slow despite billions of dollars
produced? in annual R&D spending.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 545

TABLE 4 Impact of reverse salients in digital and physical innovation.

Reverse Rate of
Domain Digital Physical salient innovation Example
Purely digital Emerging N/a Digital Typically high Artificial intelligence (Haenlein &
Kaplan, 2019)
CPS (digital-limited) Emerging Mature Digital Typically high Facial recognition, edge computing, IoT
CPS (digital and physical Emergent Emergent Both Slow Virtual reality (Zhan et al., 2020),
limited) autonomous vehicles (Bagloee
et al., 2016)
CPS (mature digital and Mature Mature Both Initially high, CNC (Ernst, 1997), Internet-enabled
physical) then slow appliances (Boutellier & Heinzen, 2014)
CPS (physical-limited) Mature Emerging Physical Slow 3D printers, wind turbines (Beiter
et al., 2021; Veers et al., 2019)
Purely physical N/a Emerging Physical Slow Photovoltaic cells, battery materials (Chu
et al., 2017)

Research on CPS innovation so far has emphasized the digital constraints on system performance, what is
digital-limited case, but future research should examine needed is a more nuanced discussion of the interaction of
CPSs that are physical-limited. Since CPS performance can this digital innovation with the inevitable physical
be limited by either digital or physical components, if constraints.
researchers overestimate the rate of improvement for the Additional research should examine other hybrid
physical components, this leads to incorrect predictions of CPSs (beyond digital manufacturing) where the perfor-
the innovation trajectory of a CPS. Future research is also mance drivers are more physical than digital. Such stud-
needed on the interdependence of component innovation ies should consider two possible technological
for CPS and their respective innovation rates and techno- trajectories. The first is when physical components limit
logical bottlenecks. the rate of innovation for digital components or the over-
At the same time, theories of digital innovation (and all system. The second is when improvements in digital
innovation more generally) have rarely considered what components work around the limitations of slowly
happens when the digital meets the physical. Thus improving physical ones, as appears to be happening
research on digital innovation should consider the rarely with self-driving cars and virtual reality displays. Theo-
explored middle ground of Table 4, between traditional ries and measures for reverse salients could help explain
(physical) and digital innovation theories. The case of 3D both technological trajectory. Finally, future research
printing reminds us that in the end, digital innovation should examine reverse salients where progress depends
must serve material beings living in a material world. on a complex interplay of digital and physical innovation;
Cyber-physical and other systems that span both domains such studies could link to the rich literature on
inherently face material constraints not applicable to co-evolutionary innovation processes (e.g., Garud &
purely digital innovations, but even the latter will eventu- Rappa, 1994; Geels, 2005; van den Ende et al., 2012).
ally face material constraints when they interface with
the material world.
Future research on digital innovation is thus at a 5.3 | Implications for reverse salients
crossroads as to how to address this tangible reality. The
easiest way would be to define the scope of the theory While Hughes' concept of a reverse salient continues to
and its predictions as bounded by purely digital influence innovation research (Kapoor & McGrath, 2014;
innovations—emphasizing the boundaries of previous Lenfle & Söderlund, 2022; Mäkitie et al., 2022; Mathews
theoretical findings and perhaps the crossover points to et al., 2018; Sandberg et al., 2020; Schubert et al., 2013),
adjacent theories. The path less traveled would be to that research continues to have a singular blind spot, by
adapt and extend theories of digital innovation to explain assuming reverse salients can and will be overcome. While
digitally-enabled physical innovation, as we have done innovation studies examine successful technologies—or
with 3D printing and others have done with automobiles occasionally failed projects within a successful technology
(Lee & Berente, 2012). While previous studies of digital- (Law & Callon, 1992)—few reverse salients studies have
physical (and CPS) systems studied have emphasized examined when the salient is not overcome.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
546 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

This could be just a methodological issue: Hughes predictions. These suggest multiple opportunities to study
(1983) did not publish his study of electric power in 1890, persistent reverse salients that shape technological trajec-
just as West (2008) did not publish his study of deep tories, problem recognition and commercialization over
space communications in 1965 and Lenfle and Söderlund decades.
(2022) did not publish their study of numerical weather
prediction in 1990. In 2000, researchers would have con-
cluded photovoltaic solar energy equipment was a tech- 5.4 | Managerial implications
nology that failed to meet its goals for decades
(West, 2014), but today such equipment generates more As Hughes emphasized, reverse salients can provide great
than $100 billion in global revenue and a double-digit market opportunities. However, 3D printing points to
compound growth rate. what happens when the time horizon for resolving them
Future research should thus consider this fundamental is beyond the time horizon of private managers. While
question: What happens to a technological trajectory when utopian predictions sell analyst reports, consulting ser-
salients are not overcome? As noted earlier, 3D printing vices and stock offerings, the longstanding salients in 3D
suggests that two possible options are to target market printing performance provide exemplars of how man-
niches or to compensate by innovating in other compo- agers can identify inherent (and difficult to resolve) tech-
nents of the system. But in both cases, the salient remains. nological limits. Although rarely identified in prior
How might this be studied? While Coopersmith research, multiple simultaneous reverse salients make it
(2020) suggests that one explanation is the lack of data even less likely that the technology will make immediate
for failed technologies, a well-documented example of and rapid progress.
utopian predictions that are unrealized due to persistent At the same time, not all hybrid digital-physical sys-
reverse salients would include artificial intelligence. tems are the same. We offer a framework for identifying
Beginning with the field's origins at a 1956 Dartmouth which systems are most capable of rapid progress and
Summer conference that included Claude Shannon, which ones are not. At the same time, digital innovation
Herb Simon and Marvin Minksy, AI missed key predic- promises rapid and malleable innovation that can deliver
tions for decades. For example, in 1973 AI experts pre- rapid initial improvements in CPSs until the physical lim-
dicted (Nilson, 2009, p. 515) the commercial availability itations are reached. Together, these suggest caution for
of automatic medical diagnosis (in 1980), automated managers investing in CPS innovation where physical
voice-to-text (in 1992) and self-driving vehicles for city innovation is required to realize the hypothesized benefit
streets (in 2000) but the actual wait was at least 50% of that system.
longer. For companies involved in 3D printing, whether as
Empirical research on reverse salients could thus providers or customers, a key implication is to take the
develop a more complete framework for the co-evolution physical limitations of 3D printing as they are, without
of salients, critical problems and technological trajecto- expecting that they will eventually disappear. Instead of
ries. Prior research has noted the difference between considering uses where 3D printing would have a major
salients that are quickly recognized and those that are impact if limitations were eliminated, they should instead
not. Future research could consider both salients that focus on uses where 3D printing already can make a dif-
take decades to resolve and those not yet resolved as of ference. Such analysis should leverage convergence with
the current writing. Even among successful innovations, other digital technologies, whether on the digital
such decades-long trajectories are more common than (AI, AR) or the physical side (robotics, CNC, IoT) to fully
often assumed: The 1987 Knowledge Navigator video that unleash the potential of 3D printing as it now is, not as
made Apple CEO John Sculley the butt of endless jokes we dream it could be.
was largely realized 24 years later when Apple added the
Siri voice assistant to the iPhone. ACKNOWLEDGMENTS
Another complex systemic innovation that has been We are grateful to the reviewers and guest editor Michael
the subject of several salient studies is the transition to Stanko for their guidance and support throughout the
renewable energy, as when Mäkitie et al. (2022) exam- development of this article; we particularly appreciate
ined how reverse salients in three RE transportation tech- the reviewer who suggested a possible connection to the
nologies influenced investment decisions. Beyond AI and influential work of Thomas Hughes.
RE, data are readily available on reverse salients for other
key technologies such as fusion power production, F U N D I N G IN F O R M A T I O N
humanoid robots or interplanetary exploration; more This research was in part funded through the Chair Tech-
recently, small-scale nuclear power plants and flying tax- nology for Change of the Institut Polytechnique de Paris,
ies are growing closer to realizing decades-long utopian as supported by the patronage of Accenture.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 547

CONFLICT OF INTEREST STATEMENT Beiter, P., A. Cooperman, E. Lantz, T. Stehly, M. Shields, R. Wiser,
The authors declare no conflicts of interest. T. Telsnig, L. Kitzing, V. Berkhout, and Y. Kikuchi. 2021.
“Wind Power Costs Driven by Innovation and Experience With
Further Reductions on the Horizon.” WIREs Energy and Envi-
E TH IC S ST A T EME N T
ronment 10(5): e398. Portico. https://doi.org/10.1002/wene.398.
The authors have read and agreed to the Committee on Ben‐Ner, A., and E. Siemsen. 2017. “Decentralization and Localiza-
Publication Ethics (COPE) international standards for tion of Production.” California Management Review 59(2): 5–23.
authors. https://doi.org/10.1177/0008125617695284.
Bijker, W. E., and J. Law, eds. 1994. Shaping Technology/Building
ORCID Society: Studies in Sociotechnical Change. Cambridge, MA: MIT
Thierry Rayna https://orcid.org/0000-0003-0144-6031 Press.
Birtchnell, T., A. Daly, T. Rayna, and L. Striukova. 2018. 3D Print-
Joel West https://orcid.org/0000-0002-7958-5229
ing and Intellectual Property Futures (Independent Report No.
2018/03). Newport, UK: UK Intellectual Property Office.
R EF E RE N C E S Blakey-Milner, B., P. Gradl, G. Snedden, M. Brooks, J. Pitot, E.
Achillas, C., D. Aidonis, E. Iakovou, M. Thymianidis, and D. Lopez, M. Leary, F. Berto, and A. du Plessis. 2021. “Metal
Tzetzis. 2015. “A Methodological Framework for the Inclusion Additive Manufacturing in Aerospace: A Review.” Materials &
of Modern Additive Manufacturing into the Production Portfo- Design 209: 110008. https://doi.org/10.1016/j.matdes.2021.
lio of a Focused Factory.” Journal of Manufacturing Systems 37: 110008.
328–39. https://doi.org/10.1016/j.jmsy.2014.07.014. Blueprint. 2020. “CES 2020 Shows Us What the End of the Consumer
Appleyard, M. M., and H. W. Chesbrough. 2017. “The Dynamics of 3D Printing Hype Cycle Looks Like.” 3DPrint.ComjThe Voice of
Open Strategy: From Adoption to Reversion.” Long Range Plan- 3D Printing/Additive Manufacturing. January 24, 2020. https://
ning 50(3): 310–21. https://doi.org/10.1016/j.lrp.2016.07.004. 3dprint.com/262622/ces-2020-shows-us-what-the-end-of-the-
Aryan, V., J. Bertling, and C. Liedtke. 2021. “Topology, Typology, consumer-3d-printing-hype-cycle-looks-like/.
and Dynamics of Commons-Based Peer Production: On Plat- Bogers, M., R. Hadar, and A. Bilberg. 2016. “Additive Manufactur-
forms, Actors, and Innovation in the Maker Movement.” Crea- ing for Consumer-Centric Business Models: Implications for
tivity and Innovation Management 30(1): 63–79. https://doi.org/ Supply Chains in Consumer Goods Manufacturing.” Technolog-
10.1111/caim.12392. ical Forecasting and Social Change 102: 225–39. https://doi.org/
Åstebro, T. 2002. “Noncapital Investment Costs and the Adoption 10.1016/j.techfore.2015.07.024.
of CAD and CNC in U.S. Metalworking Industries.” RAND Bogers, M., and W. Horst. 2014. “Collaborative Prototyping:
Journal of Economics 33(4): 672–88. https://doi.org/10.2307/ Cross-Fertilization of Knowledge in Prototype-Driven Problem
3087480. Solving.” Journal of Product Innovation Management 31(4):
Autio, E., R. Mudambi, and Y. Yoo. 2021. “Digitalization and Glob- 744–64. https://doi.org/10.1111/jpim.12121.
alization in a Turbulent World: Centrifugal and Centripetal Boland, R. J., K. Lyytinen, and Y. Yoo. 2007. “Wakes of Innovation
Forces.” Global Strategy Journal 11(1): 3–16. https://doi.org/10. in Project Networks: The Case of Digital 3‐D Representations
1002/gsj.1396. in Architecture, Engineering, and Construction.” Organization
Avinson. 2018. “How Long Does 3D Printing Take?” September Science 18(4): 631–47. https://doi.org/10.1287/orsc.1070.0304.
20, 2018. https://3space.com/blog/how-long-does-3d-printing- Borgia, E. 2014. “The Internet of Things Vision: Key Features,
take/. Applications and Open Issues.” Computer Communications 54:
Böckin, D., and A.-M. Tillman. 2019. “Environmental Assessment 1–31. https://doi.org/10.1016/j.comcom.2014.09.008.
of Additive Manufacturing in the Automotive Industry.” Jour- Boutellier, R., and M. Heinzen. 2014. Growth Through Innovation:
nal of Cleaner Production 226: 977–87. https://doi.org/10.1016/ Managing the Technology-Driven Enterprise. Cham, Switzer-
j.jclepro.2019.04.086. land: Springer Science & Business Media.
Baars, H., C. Felden, P. Gluchowski, A. Hilbert, H.-G. Kemper, and Bowles, M. D., and V. P. Dawson. 1998. “The Advanced Turboprop
S. Olbrich. 2014. “Shaping the Next Incarnation of Business Project: Radical Innovation in a Conservative Environment.” In
Intelligence.” Business & Information Systems Engineering 6(1): From Engineering Science to Big Science—The NACA and NASA
11–6. https://doi.org/10.1007/s12599-013-0307-z. Collier Trophy Research Project Winners, edited by Pamela Etter
Bagloee, S. A., M. Tavana, M. Asadi, and T. Oliver. 2016. “Autono- Mack, 321–43. Honolulu: University Press of the Pacific.
mous Vehicles: Challenges, Opportunities, and Future Implica- Budinoff, H. D., J. Bushra, and M. Shafae. 2021. “Community-
tions for Transportation Policies.” Journal of Modern Driven PPE Production Using Additive Manufacturing During
Transportation 24(4): 284–303. https://doi.org/10.1007/s40534- the COVID-19 Pandemic: Survey and Lessons Learned.” Jour-
016-0117-3. nal of Manufacturing Systems 60: 799–810. https://doi.org/10.
Baldwin, C. Y. 2015. “Bottlenecks, Modules, and Dynamic Architec- 1016/j.jmsy.2021.07.010.
tural Capabilities.” In The Oxford Handbook of Dynamic Capa- Campbell, T., C. Williams, O. Ivanova, and B. Garrett. 2011. Could
bilities, Online ed., edited by David J. Teece and Sohvi Heaton. 3D Printing Change the World? Washington, DC: Atlantic
Oxford: Oxford University Press. https://doi.org/10.1093/ Council.
oxfordhb/9780199678914.013.011. Chandler, A. D. 1990. Scale and Scope: The Dynamics of Industrial
Baumers, M., and M. Holweg. 2019. “On the Economics of Additive Capitalism. Cambridge, MA: Belknap Press.
Manufacturing: Experimental Findings.” Journal of Operations Chaudhuri, A., H. A. Gerlich, J. Jayaram, A. Ghadge, J. Shack,
Management 65(8): 794–809. https://doi.org/10.1002/joom. B. H. Brix, L. H. Hoffbeck, and N. Ulriksen. 2021. “Selecting
1053. Spare Parts Suitable for Additive Manufacturing: A Design
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
548 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Science Approach.” Production Planning & Control 32(8): 670– Dedehayir, O., and S. J. Mäkineif. 2008. “Dynamics of Reverse
87. https://doi.org/10.1080/09537287.2020.1751890. Salience as Technological Performance Gap: An Empirical
Christiansen, A. C., and J. Buen. 2002. “Managing Environmental Study of the Personal Computer Technology System.” Journal
Innovation in the Energy Sector: The Case of Photovoltaic and of Technology Management & Innovation 3(3): 55–66. https://
Wave Power Development in Norway.” International Journal of doi.org/10.4067/S0718-27242008000100006.
Innovation Management 6(3): 233–56. https://doi.org/10.1142/ Dedehayir, O., and S. J. Mäkinen. 2011b. “Determining Reverse
s1363919602000586. Salient Types and Evolutionary Dynamics of Technology Sys-
Chu, S., Y. Cui, and N. Liu. 2017. “The Path Towards Sustainable tems With Performance Disparities.” Technology Analysis &
Energy.” Nature Materials 16(1): 16–22. https://doi.org/10. Strategic Management 23(10): 1095–114. https://doi.org/10.
1038/nmat4834. 1080/09537325.2011.621308.
Claussen, J., and M. A. Halbinger. 2021. “The Role of Pre- Depoorter, B. 2014. “Intellectual Property Infringements & 3D Printing:
Innovation Platform Activity for Diffusion Success: Evidence Decentralized Piracy.” Hastings Law Journal 65(6): 1483–504.
From Consumer Innovations on a 3D Printing Platform.” https://repository.uclawsf.edu/hastings_law_journal/vol65/iss6/2.
Research Policy 50(8): 103943. https://doi.org/10.1016/j.respol. Despeisse, M., and T. Minshall. 2017. “Skills and Education for
2020.103943. Additive Manufacturing: A Review of Emerging Issues.” In
Conkle, J., P. S. Suchdev, E. Alexander, R. Flores-Ayala, Advances in Production Management Systems. The Path to Intel-
U. Ramakrishnan, and R. Martorell. 2018. “Accuracy and Reli- ligent, Collaborative and Sustainable Manufacturing, edited by
ability of a Low-Cost, Handheld 3D Imaging System for Child H. Lödding, R. Riedel, K.-D. Thoben, G. von Cieminski, and D.
Anthropometry.” PLoS One 13(10): e0205320. https://doi.org/ Kiritsis, 289–97. Cham, Switzerland: Springer International
10.1371/journal.pone.0205320. Publishing. https://doi.org/10.1007/978-3-319-66923-6_34.
Cooper, R. G. 2021. “Accelerating Innovation: Some Lessons From Dosi, G. 1982. “Technological Paradigms and Technological Trajec-
the Pandemic.” Journal of Product Innovation Management tories.” Research Policy 11(3): 147–62. https://doi.org/10.1016/
38(2): 221–32. https://doi.org/10.1111/jpim.12565. 0048-7333(82)90016-6.
Coopersmith, J. 2020. “Spectacular Flops: Game‐Changing Technol- Dowling, L., J. Kennedy, S. O'Shaughnessy, and D. Trimble. 2020.
ogies That Failed by Michael Brian Schiffer.” Technology and “A Review of Critical Repeatability and Reproducibility Issues
Culture 61(4): 1235–6. https://doi.org/10.1353/tech.2020.0134. in Powder Bed Fusion.” Materials & Design 186: 108346.
Coutts, E. R., A. Wodehouse, and J. Robertson. 2019. “A Compari- https://doi.org/10.1016/j.matdes.2019.108346.
son of Contemporary Prototyping Methods.” Proceedings of the Ernst, H. 1997. Small Business Economics 9(4): 361–81. https://doi.
Design Society: International Conference on Engineering Design org/10.1023/a:1007921808138.
1(1): 1313–22. https://doi.org/10.1017/dsi.2019.137. Eyers, D. R., and A. T. Potter. 2017. “Industrial Additive
Daim, T., J. Justice, L. Hogaboam, S. J. Mäkinen, and O. Dedehayir. Manufacturing: A Manufacturing Systems Perspective.” Com-
2014. “Identifying and Forecasting the Reverse Salient in Video puters in Industry 92–93: 208–18. https://doi.org/10.1016/j.
Game Consoles: A Performance Gap Ratio Comparative Analy- compind.2017.08.002.
sis.” Technological Forecasting and Social Change 82: 177–89. Feng, J., J. Fu, Z. Lin, C. Shang, and B. Li. 2018. “A Review of the
https://doi.org/10.1016/j.techfore.2013.06.007. Design Methods of Complex Topology Structures for 3D Print-
Davies, A. 1996. “Innovation in Large Technical Systems: The Case ing.” Visual Computing for Industry, Biomedicine, and Art 1(1):
of Telecommunications.” Industrial and Corporate Change 5(4): 5. https://doi.org/10.1186/s42492-018-0004-3.
1143–80. https://doi.org/10.1093/icc/5.4.1143. Fontana, F., C. Klahn, and M. Meboldt. 2019. “Value-Driven Clus-
Davies, A., and T. Brady. 2000. “Organisational Capabilities and tering of Industrial Additive Manufacturing Applications.”
Learning in Complex Product Systems: Towards Repeatable Journal of Manufacturing Technology Management 30(2): 366–
Solutions.” Research Policy 29(7–8): 931–53. https://doi.org/10. 90. https://doi.org/10.1108/JMTM-06-2018-0167.
1016/s0048-7333(00)00113-x. Ford, S., and M. Despeisse. 2016. “Additive Manufacturing and Sus-
de Bruijn, E. 2010. On the Viability of the Open Source Development tainability: An Exploratory Study of the Advantages and Chal-
Model for the Design of Physical Objects: Lessons Learned From lenges.” Journal of Cleaner Production 137: 1573–87. https://
the RepRap Project. Unpublished MSc Thesis. Tilburg: Tilburg doi.org/10.1016/j.jclepro.2016.04.150.
University. Frandsen, C. S., M. M. Nielsen, A. Chaudhuri, J. Jayaram, and K.
de Jong, J. P., and E. de Bruijn. 2013. “Innovation Lessons From Govindan. 2020. “In Search for Classification and
3-D Printing.” MIT Sloan Management Review 54(2): 43–52. Selection of Spare Parts Suitable for Additive Manufacturing:
Dedehayir, O., and S. J. Mäkinen. 2011a. “Measuring Industry Clock- A Literature Review.” International Journal of Production
speed in the Systemic Industry Context.” Technovation 31(12): Research 58(4): 970–96. https://doi.org/10.1080/00207543.
627–37. https://doi.org/10.1016/j.technovation.2011.07.008. 2019.1605226.
Dedehayir, O. 2009. “Bibliometric Study of the Reverse Salient Con- Franke, N., and F. Piller. 2004. “Value Creation by Toolkits for User
cept.” Journal of Industrial Engineering and Management 2(3): Innovation and Design: The Case of the Watch Market.” Jour-
3. https://doi.org/10.3926/jiem.v2n3.p569-591. nal of Product Innovation Management 21(6): 401–15. https://
Dedehayir, O., and A. Hornsby. 2008. “Staying at the Technology doi.org/10.1111/j.0737-6782.2004.00094.x.
Forefront by Spotting the Reverse Salient: The Case of Digital Frazer, J. S. 2022. “The Role of Distributed Manufacturing and 3D
Video Broadcasting.” In 2008 4th IEEE International Conference Printing in Development of Personal Protective Equipment
on Management of Innovation and Technology 189–94. https:// Against COVID-19.” In Emerging Applications of 3D Printing
doi.org/10.1109/icmit.2008.4654360. During CoVID 19 Pandemic, edited by K. Sandhu, S. Singh, C.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 549

Prakash, N. R. Sharma, and K. Subburaj, 15–34. Singapore: Operations Management 65(8): 810–26. https://doi.org/10.1002/
Springer. https://doi.org/10.1007/978-981-33-6703-6_2. joom.1054.
Fujita, M. 2001. “AIBO: Toward the Era of Digital Creatures.” The Hendler, S. 2019. “Digital‐Physical Product Development: A
International Journal of Robotics Research 20(10): 781–94. Qualitative Analysis.” European Journal of Innovation Man-
https://doi.org/10.1177/02783640122068092. agement 22(2): 315–34. https://doi.org/10.1108/ejim-01-2018-
Garcia, R., and R. Calantone. 2002. “A Critical Look at Technologi- 0026.
cal Innovation Typology and Innovativeness Terminology: A Herterich, M. M., T. Buehnen, F. Uebernickel, and W. Brenner.
Literature Review.” Journal of Product Innovation Management 2016. “A Taxonomy of Industrial Service Systems Enabled by
19(2): 110–32. https://doi.org/10.1111/1540-5885.1920110. Digital Product Innovation.” In Proceedings of the 49th Hawaii
Garud, R., and M. A. Rappa. 1994. “A Socio‐Cognitive Model of International Conference on System Sciences (HICSS-49) 1236–
Technology Evolution: The Case of Cochlear Implants.” Orga- 45. https://doi.org/10.1109/HICSS.2016.157.
nization Science 5(3): 344–62. https://doi.org/10.1287/orsc.5. Hill, C. W. L., and F. T. Rothaermel. 2003. “The Performance of
3.344. Incumbent Firms in the Face of Radical Technological Innova-
GE. 2020. Better Performance From Fewer Parts. GE.com. https:// tion.” The Academy of Management Review 28(2): 257. https://
web.archive.org/web/20200522182434/https://www.ge.com/ doi.org/10.2307/30040712.
additive/additive-manufacturing/industries/aviation-aerospace. Holmström, J., M. Holweg, S. H. Khajavi, and J. Partanen. 2016.
Gebler, M., A. J. S. Uiterkamp, and C. Visser. 2014. “A Global Sus- “The Direct Digital Manufacturing (R)evolution: Definition of a
tainability Perspective on 3D Printing Technologies.” Energy Research Agenda.” Operations Management Research 9(1–2): 1–
Policy 74: 158–67. https://doi.org/10.1016/j.enpol.2014.08.033. 10. https://doi.org/10.1007/s12063-016-0106-z.
Geels, F. W. 2005. Technological Transitions and System Innova- Holzmann, P., R. J. Breitenecker, and E. J. Schwarz. 2020. “Business
tions: A Co-Evolutionary and Socio-Technical Analysis. Chelten- Model Patterns for 3D Printer Manufacturers.” Journal of
ham, UK: Elgar. Manufacturing Technology Management 31(6): 1281–300.
Ghaffar, S., and P. Mullett. 2021. “Discussion: Commentary: 3D https://doi.org/10.1108/JMTM-09-2018-0313.
Printing Set to Transform the Construction Industry.” Proceed- Hoppmann, J., J. Huenteler, and B. Girod. 2014. “Compulsive Pol-
ings of the Institution of Civil Engineers—Structures and Build- icy‐Making—The Evolution of the German Feed‐In Tariff Sys-
ings 174(4): 336–7. https://doi.org/10.1680/jstbu.21.00024. tem for Solar Photovoltaic Power.” Research Policy 43(8): 1422–
Gibson, I., D. Rosen, B. Stucker, and M. Khorasani. 2014. Additive 41. https://doi.org/10.1016/j.respol.2014.01.014.
Manufacturing Technologies. New York: Springer. Hu, B., and Z. Sun. 2022. “Managing Self-Replicating Innovative
Goh, G. D., S. L. Sing, and W. Y. Yeong. 2020. “A Review on Goods.” Management Science 68(1): 399–419. https://doi.org/10.
Machine Learning in 3D Printing: Applications, Potential, and 1287/mnsc.2020.3936.
Challenges.” Artificial Intelligence Review 54(1): 63–94. https:// Hughes, T. P. 1983. Networks of Power: Electrification in Western
doi.org/10.1007/s10462-020-09876-9. Society, 1880–1930. Baltimore: Johns Hopkins University Press.
Greer, C., M. Burns, D. Wollman, and E. Griffor. 2019. Cyber- Hughes, T. P. 1987. “The Evolution of Large Technological Sys-
Physical Systems and Internet of Things. Gaithersburg, MD: tems.” In The Social Construction of Technological Systems, edi-
National Institute of Standards and Technology. ted by W. E. Bijker, T. P. Hughes, and T. J. Pinch, 51–82.
Greul, A., J. West, and S. Bock. 2018. “Open at Birth? Why New Cambridge, MA: MIT Press.
Firms Do (or Don’t) Use Open Innovation.” Strategic Entrepre- Hull, C., M. Feygin, Y. Baron, R. Sanders, E. Sachs, A. Lightman,
neurship Journal 12(3): 392–420. Portico. https://doi.org/10. and T. Wohlers. 1995. “Rapid Prototyping: Current Technology
1002/sej.1282. and Future Potential.” Rapid Prototyping Journal 1(1): 11–9.
Haenlein, M., and A. Kaplan. 2019. “A Brief History of Artificial https://doi.org/10.1108/13552549510732026.
Intelligence: On the Past, Present, and Future of Artificial Intel- Iansiti, M., and K. R. Lakhani. 2020. Competing in the Age of AI:
ligence.” California Management Review 61(4): 5–14. https:// Strategy and Leadership When Algorithms and Networks Run
doi.org/10.1177/0008125619864925. the World. Boston: Harvard Business Review Press.
Haleem, A., M. Javaid, and S. Rab. 2020. “Impact of Additive Jacobs, P. F. 1992. Rapid Prototyping & Manufacturing: Fundamen-
Manufacturing in Different Areas of Industry 4.0.” Interna- tals of Stereolithography. Dearborn, MI: Society of Manufactur-
tional Journal of Logistics Systems and Management 37(2): 239. ing Engineers.
https://doi.org/10.1504/IJLSM.2020.110578. Javaid, M., A. Haleem, R. Pratap Singh, and R. Suman. 2021.
Han, K., and J. Shin. 2014. “A Systematic Way of Identifying and “Industrial Perspectives of 3D Scanning: Features, Roles and
Forecasting Technological Reverse Salients Using QFD, Biblio- It's Analytical Applications.” Sensors International 2: 100114.
metrics, and Trend Impact Analysis: A Carbon Nanotube Bio- https://doi.org/10.1016/j.sintl.2021.100114.
sensor Case.” Technovation 34(9): 559–70. https://doi.org/10. Jiang, R., R. Kleer, and F. T. Piller. 2017. “Predicting the Future of
1016/j.technovation.2014.05.009. Additive Manufacturing: A Delphi Study on Economic and
Hanelt, A., E. Piccinini, R. W. Gregory, B. Hildebrandt, and L. M. Societal Implications of 3D Printing for 2030.” Technological
Kolbe. 2015, March. “Digital Transformation of Primarily Physical Forecasting and Social Change 117: 84–97. https://doi.org/10.
Industries-Exploring the Impact of Digital Trends on Business 1016/j.techfore.2017.01.006.
Models of Automobile Manufacturers.” Proceedings der 12. Inter- Kalantari, H. D., L. W. Johnson, and C. R. Perera. 2021. “The Effect
nationalen Tagung Wirtschaftsinformatik: 1313–27. of Gender on Willingness to Pay for Mass Customised Running
Heinen, J. J., and K. Hoberg. 2019. “Assessing the Potential of Addi- Shoes.” Journal of Global Fashion Marketing 12(2): 161–75.
tive Manufacturing for the Provision of Spare Parts.” Journal of https://doi.org/10.1080/20932685.2020.1867878.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
550 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Kapoor, R., and P. J. McGrath. 2014. “Unmasking the Interplay Open Journal of the Industrial Electronics Society 1: 298–310.
Between Technology Evolution and R&D Collaboration: Evi- https://doi.org/10.1109/OJIES.2020.3031660.
dence From the Global Semiconductor Manufacturing Indus- Lenfle, S., and J. Söderlund. 2022. “Project‐Oriented Agency and
try, 1990–2010.” Research Policy 43(3): 555–69. https://doi.org/ Regeneration in Socio‐Technical Transition: Insights From the
10.1016/j.respol.2013.08.002. Case of Numerical Weather Prediction (1978–2015).” Research
Kenney, M. E. 2013. Cost Reduction Through the Use of Additive Policy 51(3): 104455. https://doi.org/10.1016/j.respol.2021.
Manufacturing (3D Printing) and Collaborative Product Life Cycle 104455.
Management Technologies to Enhance the Navy's Maintenance Li, L. 2018. “China's Manufacturing Locus in 2025: With a Compar-
Programs. Doctoral dissertation. Monterey, CA: Naval Postgradu- ison of “Made-in-China 2025” and “Industry 4.0”.” Technologi-
ate School. https://apps.dtic.mil/sti/citations/ADA591795. cal Forecasting and Social Change 135: 66–74. https://doi.org/
Khan, H. N., D. A. Hounshell, and E. R. H. Fuchs. 2018. “Science 10.1016/j.techfore.2017.05.028.
and Research Policy at the End of Moore's Law.” Nature Elec- Li, S. 2013. A Desktop Factory; 3-D Printers, Which Can Make Plas-
tronics 1(1): 14–21. https://doi.org/10.1038/s41928-017-0005-9. tic Objects, Are Creeping Into the Home. Los Angeles Times,
Khorasani, M., A. Ghasemi, B. Rolfe, and I. Gibson. 2021. “Additive July 28, p. B1.
Manufacturing a Powerful Tool for the Aerospace Industry.” Lien, Shao-Yu, Shin-Ming Cheng, Sung-Yin Shih, and Kwang-
Rapid Prototyping Journal 28(1): 87–100. https://doi.org/10. Cheng Chen. 2012. “Radio Resource Management for QoS
1108/RPJ-01-2021-0009. Guarantees in Cyber-Physical Systems.” IEEE Transactions on
Kim, D.‐J., and B. Kogut. 1996. “Technological Platforms and Diver- Parallel and Distributed Systems 23(9): 1752–61. https://doi.org/
sification.” Organization Science 7(3): 283–301. https://doi.org/ 10.1109/TPDS.2012.151.
10.1287/orsc.7.3.283. Luchs, M. G., K. S. Swan, and M. E. H. Creusen. 2015. “Perspective:
Kim, H., Y. Lin, and T.-L. B. Tseng. 2018. “A Review on Quality A Review of Marketing Research on Product Design With
Control in Additive Manufacturing.” Rapid Prototyping Journal Directions for Future Research.” Journal of Product Innovation
24(3): 645–69. https://doi.org/10.1108/RPJ-03-2017-0048. Management 33(3): 320–41. Portico. https://doi.org/10.1111/
Kling, R. 1996. “Hopes and Horrors: Technological Utopianism and jpim.12276.
Anti-Utopianism in Narratives of Computerization.” In Com- Lyytinen, K. 2022. “Innovation Logics in the Digital Era: A Sys-
puterization and Controversy: Value Conflicts and Social temic Review of the Emerging Digital Innovation Regime.”
Choices, edited by R. Kling, 40–58. San Diego: Academic Press. Innovation: Organization and Management 24(1): 13–34.
Knofius, N., M. C. van der Heijden, and W. H. M. Zijm. 2019. “Con- https://doi.org/10.1080/14479338.2021.1938579.
solidating Spare Parts for Asset Maintenance With Additive Mäkitie, T., J. Hanson, M. Steen, T. Hansen, and A. D. Andersen.
Manufacturing.” International Journal of Production Economics 2022. “Complementarity Formation Mechanisms in Technol-
208: 269–80. https://doi.org/10.1016/j.ijpe.2018.11.007. ogy Value Chains.” Research Policy 51(7): 104559. https://doi.
Law, J., and M. Callon. 1992. “The Life and Death of an Aircraft: A org/10.1016/j.respol.2022.104559.
Network Analysis of Technical Change.” In Shaping Mahmood, T., and M. S. Mubarik. 2020. “Balancing Innovation and
Technology/Building Society: Studies in Sociotechnical Change, Exploitation in the Fourth Industrial Revolution: Role of Intel-
edited by W. E. Bijker and J. Law, 21–52. Cambridge, MA: MIT lectual Capital and Technology Absorptive Capacity.” Techno-
Press. logical Forecasting and Social Change 160: 120248. https://doi.
Lee, E. A. 2008. “Cyber Physical Systems: Design Challenges.” In org/10.1016/j.techfore.2020.120248.
2008 11th IEEE International Symposium on Object and Compo- Majidpour, M. 2016. “Technological Catch‐Up in Complex Product
nent‐Oriented Real‐Time Distributed Computing (ISORC). Systems.” Journal of Engineering and Technology Management
https://doi.org/10.1109/isorc.2008.25. 41: 92–105. https://doi.org/10.1016/j.jengtecman.2016.07.003.
Lee, J., B. Bagheri, and H.‐A. Kao. 2015. “A Cyber‐Physical Systems Markus, M. L. 1994. “Electronic Mail as the Medium of Managerial
Architecture for Industry 4.0‐Based Manufacturing Systems.” Choice.” Organization Science 5(4): 502–27. https://doi.org/10.
Manufacturing Letters 3: 18–23. https://doi.org/10.1016/j. 1287/orsc.5.4.502.
mfglet.2014.12.001. Mathews, J. A., H. Tan, and M. C. Hu. 2018. “Moving to a Circular
Lee, J., and N. Berente. 2012. “Digital Innovation and the Division Economy in China: Transforming Industrial Parks into Eco-
of Innovative Labor: Digital Controls in the Automotive Indus- Industrial Parks.” California Management Review 60(3): 157–81.
try.” Organization Science 23(5): 1428–47. https://doi.org/10. https://doi.org/10.1177/0008125617752692.
1287/orsc.1110.0707. Mavri, M., E. Fronimaki, and A. Kadrefi. 2021. “Survey Analysis for
Lee, J., H.-C. Kim, J.-W. Choi, and I. H. Lee. 2017. “A Review on the Adoption of 3D Printing Technology: Consumers' Perspec-
3D Printed Smart Devices for 4D Printing.” International Jour- tive.” Journal of Science and Technology Policy Management 14
nal of Precision Engineering and Manufacturing-Green Technol- (2): 353–85. https://doi.org/10.1108/JSTPM-02-2020-0023.
ogy 4(3): 373–83. https://doi.org/10.1007/s40684-017-0042-x. Mendis, D. K., M. A. Lemley, and M. Rimmer, eds. 2019. 3D Print-
Lehenkari, J., and R. Miettinen. 2002. “Standardisation in the ing and Beyond: Intellectual Property and Regulation. Chelten-
Construction of a Large Technological System—The Case of the ham, UK: Elgar.
Nordic Mobile Telephone System.” Telecommunications Policy Millard, J., M. N. Sorivelle, S. Deljanin, E. Unterfrauner, and
26(3–4): 109–27. https://doi.org/10.1016/s0308-5961(02)00004-6. C. Voigt. 2018. “Is the Maker Movement Contributing to
Leitão, P., F. Pires, S. Karnouskos, and A. W. Colombo. 2020. “Quo Sustainability?” Sustainability 10(7): 7. https://doi.org/10.3390/
Vadis Industry 4.0? Position, Trends, and Challenges.” IEEE su10072212.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 551

Minguella-Canela, J., S. M. Planas, J. R. G. Ayats, and M. A. S. Peltz, J. F. . 1987. Sylmar Firm Puts Hot Idea for Model-Making
L
opez. 2019. “Study and Comparison of the Different Costs' Machine on the Market Via Canada. Los Angeles Times,
Schema Associated to Geometry, Material and Processing October 11, 1987. p. OC-D13.
Between 3D Printing, Injection Molding and Machining Petrick, I. J., and T. W. Simpson. 2013. “3D Printing Disrupts
Manufacturing Technologies.” Procedia Manufacturing 41: Manufacturing: How Economies of One Create New Rules of
280–7. https://doi.org/10.1016/j.promfg.2019.09.010. Competition.” Research‐Technology Management 56(6): 12–6.
Molitch-Hou, M. 2017. “Michelin and Fives' AddUp: From Making https://doi.org/10.5437/08956308x5606193.
Tires to Making Metal 3D Printers.” January 30, 2017. https:// Pradel, P., Z. Zhu, R. Bibb, and J. Moultrie. 2018. “Investigation of
www.engineering.com/story/michelin-and-fives-addup-from- Design for Additive Manufacturing in Professional Design Prac-
making-tires-to-making-metal-3d-printers. tice.” Journal of Engineering Design 29(4–5): 165–200. https://
Moore, G. E. 2006. “Moore's Law at 40.” In Understanding Moore's doi.org/10.1080/09544828.2018.1454589.
Law: Four Decades of Innovation, edited by D. C. Brock, 67–84. Prater, T. J., Q. A. Bean, R. D. Beshears, T. D. Rolin, N. J.
Philadelphia: CHC Press. Werkheiser, E. A. Ordonez, R. M. Ryan, and F. E. Ledbetter,
Mubarak, M. F., and M. Petraite. 2020. “Industry 4.0 Technologies, III. 2016. “Summary Report on Phase I Results From the 3D
Digital Trust and Technological Orientation: What Matters in Printing in Zero G Technology Demonstration Mission.” https://
Open Innovation?” Technological Forecasting and Social Change ntrs.nasa.gov/citations/20160008972.
161: 120332. https://doi.org/10.1016/j.techfore.2020.120332. Prencipe, A., A. Davies, and M. Hobday, eds. 2003. The Business of
Muir, M., and A. Haddud. 2017. “Additive Manufacturing in the Systems Integration. Oxford: Oxford.
Mechanical Engineering and Medical Industries Spare Parts Sup- Quah, D. 2003. “Digital Goods and the New Economy.” In New
ply Chain.” Journal of Manufacturing Technology Management Economy Handbook, edited by D. Jones, 289–321. San Diego:
29(2): 372–97. https://doi.org/10.1108/JMTM-01-2017-0004. Academic Press.
Mulder, K., and M. Knot. 2001. “PVC Plastic: A History of Systems Raasch, C., C. Herstatt, and K. Balka. 2009. “On the Open Design of
Development and Entrenchment.” Technology in Society 23(2): Tangible Goods.” R&D Management 39(4): 382–93. https://doi.
265–86. https://doi.org/10.1016/s0160-791x(01)00013-6. org/10.1111/j.1467-9310.2009.00567.x.
Mussatto, A., R. Groarke, R. K. Vijayaraghavan, C. Hughes, M. A. Radstok, E. 1999. “Rapid Tooling.” Rapid Prototyping Journal 5(4):
Obeidi, M. N. Dogu, M. A. Yalçin, P. J. McNally, Y. Delaure, 164–9. https://doi.org/10.1108/13552549910295488.
and D. Brabazon. 2022. “Assessing Dependency of Part Proper- Rajkumar, R. ( Raj), I. Lee, L. Sha, and J. Stankovic. 2010. “Cyber‐
ties on the Printing Location in Laser-Powder Bed Fusion Metal Physical Systems.” In Proceedings of the 47th Design Automation
Additive Manufacturing.” Materials Today Communications 30: Conference. https://doi.org/10.1145/1837274.1837461.
103209. https://doi.org/10.1016/j.mtcomm.2022.103209. Randhawa, K., J. West, K. Skellern, and E. Josserand. 2020. “Evolv-
Naik, H. S., A. Fritzsche, and K. M. Moeslein. 2021. “Modularity in ing a Value Chain to an Open Innovation Ecosystem: Cognitive
Making: Simplifying Solution Space for User Innovation.” R&D Engagement of Stakeholders in Customizing Medical
Management 51(1): 57–72. https://doi.org/10.1111/radm.12427. Implants.” California Management Review 63(2): 101–34.
Nasiri, M., M. Saunila, J. Ukko, T. Rantala, and H. Rantanen. 2020. https://doi.org/10.1177/0008125620974435.
“Shaping Digital Innovation Via Digital-Related Capabilities.” Rayna, T., and L. Striukova. 2015. “Open Innovation 2.0: Is Co‐crea-
Information Systems Frontiers. https://doi.org/10.1007/s10796- tion the Ultimate Challenge?” International Journal of Technol-
020-10089-2. ogy Management 69(1): 38. https://doi.org/10.1504/ijtm.2015.
Negroponte, N. 1996. Being Digital. New York: Vintage. 071030.
Nelson, R. R., and S. G. Winter. 1982. An Evolutionary Theory of Rayna, T., and L. Striukova. 2016. “From Rapid Prototyping to Home
Economic Change. Cambridge, MA: Belknap Press. Fabrication: How 3D Printing is Changing Business Model Inno-
Nerlich, B. 2005. “From Nautilus to Nanobo(a)ts: the visual con- vation.” Technological Forecasting and Social Change 102: 214–
struction of nanoscience.” AZojono: Journal of Nanotechnology. 24. https://doi.org/10.1016/j.techfore.2015.07.023.
https://doi.org/10.2240/azojono0109. Rayna, T., and L. Striukova. 2021. “Involving Consumers: The Role
NIST. 2020. “Cyber-Physical Systems, National Institute of Standards of Digital Technologies in Promoting ‘Prosumption’ and User
and Technology.” U.S. Department of Commerce, March Innovation.” Journal of the Knowledge Economy 12(1): 218–37.
16, 2020. https://web.archive.org/web/20200316224708/https:// https://doi.org/10.1007/s13132-016-0390-8.
www.nist.gov/el/cyber-physical-systems. Rayna, T. 2008. “Understanding the Challenges of the Digital Econ-
Nonaka, I., and H. Takeuchi. 2007. “The Knowledge-Creating Com- omy: The Nature of Digital Goods.” Communications and Strat-
pany.” Harvard Business Review 85(7/8): 162. egies 71: 13–6.
Nylén, D., and J. Holmström. 2015. “Digital Innovation Strategy: A Rayna, T., L. Striukova, and J. Darlington. 2015. “Co-Creation and
Framework for Diagnosing and Improving Digital Product and User Innovation: The Role of Online 3D Printing Platforms.”
Service Innovation.” Business Horizons 58(1): 57–67. https:// Journal of Engineering and Technology Management 37: 90–102.
doi.org/10.1016/j.bushor.2014.09.001. https://doi.org/10.1016/j.jengtecman.2015.07.002.
Nyman, H. J., and P. Sarlin. 2014. “From Bits to Atoms: 3D Printing Rindfleisch, A., M. O'Hern, and V. Sachdev. 2017. “The Digital
in the Context of Supply Chain Strategies.” In Proceedings of Revolution, 3D Printing, and Innovation as Data.” Journal of
the 47th Hawaii International Conference on System Sciences Product Innovation Management 34(5): 681–90. Portico. https://
4190–9. https://doi.org/10.1109/HICSS.2014.518. doi.org/10.1111/jpim.12402.
Osterwalder, A., and Y. Pigneur. 2004. “An Ontology for E-Business Rindfleisch, A., and M. O’Hern. 2015. “Brand Remixing: 3D Print-
Models.” In Value Creation From E-Business Models, edited by ing the Nokia Case.” Review of Marketing Research: 53–81.
W. Currie, 65–97. Oxford: Butterworth-Heinemann. https://doi.org/10.1108/s1548-643520150000012003.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
552 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Rogers, E. M., and F. F. Shoemaker. 1971. Communication of Inno- Stanko, M. A. 2020. “Building an Understanding of How Winning
vations: A Cross-Cultural Approach. New York: FreePress. Products Emerge When Open and Proprietary Products Coex-
Rouf, S., A. Raina, M. Irfan Ul Haq, N. Naveed, S. Jeganmohan, ist: Evidence From the RepRap Community.” Creativity and
and A. Farzana Kichloo. 2022. “3D Printed Parts and Mechani- Innovation Management 29(3): 398–412. https://doi.org/10.
cal Properties: Influencing Parameters, Sustainability Aspects, 1111/caim.12376.
Global Market Scenario, Challenges and Applications.” Stanko, M. A., and B. J. Allen. 2022. “Disentangling the Collective
Advanced Industrial and Engineering Polymer Research. 5: Motivations for User Innovation in a 3D Printing Community.”
143–58. https://doi.org/10.1016/j.aiepr.2022.02.001. Technovation 111: 102387. https://doi.org/10.1016/j.
Rouhani, C. . 2018. “Three BIG Challenges for Using FDM 3D Print- technovation.2021.102387.
ing at Scale.” Sculpteo.com. August 2, 2018. https://www. Stone, P., R. Brooks, E. Brynjolfsson, R. Calo, O. Etzioni, G. Hager,
sculpteo.com/blog/2018/08/02/three-big-challenges-for-using- J. Hirschberg, et al. 2016. Artificial Intelligence and Life in 2030:
fdm-3d-printing-at-scale/. The One Hundred Year Study on Artificial Intelligence. Stanford,
Söderberg, J., and A. Daoud. 2012. “Atoms Want to Be Free Too! CA: Stanford University.
Expanding the Critique of Intellectual Property to Physical Sturgeon, T. J. 2019. “Upgrading Strategies for the Digital Econ-
Goods.” TripleC: Communication, Capitalism & Critique. omy.” Global Strategy Journal 11(1): 34–57. Portico. https://doi.
Open Access Journal for a Global Sustainable Information org/10.1002/gsj.1364.
Society 10(1): 66–76. https://doi.org/10.31269/triplec. Thürer, M., Y. Huang, and M. Stevenson. 2021. “Workload Control
v10i1.288. in Additive Manufacturing Shops Where Post-Processing is a
Sahini, D. K., J. Ghose, S. K. Jha, A. Behera, and A. Mandal. 2020. Constraint: An Assessment by Simulation.” International Jour-
“Optimization and Simulation of Additive Manufacturing Pro- nal of Production Research 59(14): 4268–86. https://doi.org/10.
cesses: Challenges and Opportunities—A Review.” In Additive 1080/00207543.2020.1761038.
Manufacturing Applications for Metals and Composites, edited Thomas-Seale, L. E. J., J. C. Kirkman-Brown, M. M. Attallah, D. M.
by K. R. Balasubramanian and V. Senthilkumar, 187–209. Her- Espino, and D. E. T. Shepherd. 2018. “The Barriers to the Pro-
shey, Penn: IGI Global. gression of Additive Manufacture: Perspectives From UK
Salmi, M., K.-S. Paloheimo, J. Tuomi, J. Wolff, and A. Mäkitie. Industry.” International Journal of Production Economics 198:
2013. “Accuracy of Medical Models Made by Additive 104–18. https://doi.org/10.1016/j.ijpe.2018.02.003.
Manufacturing (Rapid Manufacturing).” Journal of Cranio- Tilbury, D. M. 2019. “Cyber‐Physical Manufacturing Systems.”
Maxillofacial Surgery 41(7): 603–9. https://doi.org/10.1016/j. Annual Review of Control, Robotics, and Autonomous Systems
jcms.2012.11.041. 2(1): 427–43. https://doi.org/10.1146/annurev-control-053018-
Sandberg, J., J. Holmström, and K. Lyytinen. 2020. “Digitization 023652.
and Phase Transitions in Platform Organizing Logics: Evidence Turner, F. 2010. From Counterculture to Cyberculture: Stewart
from the Process Automation Industry.” MIS Quarterly 44(1): Brand, the Whole Earth Network, and the Rise of Digital Utopi-
129–53. https://doi.org/10.25300/MISQ/2020/14520. anism. Chicago: University of Chicago Press.
Schaller, R. R. 1997. “Moore's Law: Past, Present and Future.” IEEE van den Ende, J., G. van de Kaa, S. den Uijl, and H. J. de Vries.
Spectrum 34(6): 52–9. https://doi.org/10.1109/6.591665. 2012. “The Paradox of Standard Flexibility: The Effects of Co‐
Schubert, C., J. Sydow, and A. Windeler. 2013. “The Means of Man- evolution Between Standard and Interorganizational Network.”
aging Momentum: Bridging Technological Paths and Organisa- Organization Studies 33(5–6): 705–36. https://doi.org/10.1177/
tional Fields.” Research Policy 42(8): 1389–405. https://doi.org/ 0170840612443625.
10.1016/j.respol.2013.04.004. Veers, P., K. Dykes, E. Lantz, S. Barth, C. L. Bottasso, O. Carlson,
Sculpteo. 2020. The State of 3D Printing 2020. San Francisco: A. Clifton, et al. 2019. “Grand Challenges in the Science of
Sculpteo. Wind Energy.” Science 366(6464): eaau2027.
Sgarbossa, F., M. Peron, F. Lolli, and E. Balugani. 2021. “Conven- Walsh, G. S., J. Przychodzen, and W. Przychodzen. 2017. “Support-
tional or Additive Manufacturing for Spare Parts Management: ing the SME Commercialization Process: The Case of
An Extensive Comparison for Poisson Demand.” International 3D Printing Platforms.” Small Enterprise Research 24(3):
Journal of Production Economics 233: 107993. https://doi.org/ 257–73. https://doi.org/10.1080/13215906.2017.1396490.
10.1016/j.ijpe.2020.107993. Wang, L., S. Jiang, and S. Zhang. 2020. “Mapping Technological
Shane, S. 2000. “Prior Knowledge and the Discovery of Entrepre- Trajectories and Exploring Knowledge Sources: A Case Study
neurial Opportunities.” Organization Science 11(4): 448–69. of 3D Printing Technologies.” Technological Forecasting and
https://doi.org/10.1287/orsc.11.4.448.14602. Social Change 161: 120251. https://doi.org/10.1016/j.techfore.
Shapiro, C., and H. R. Varian. 1999. Information Rules: A Strategic 2020.120251.
Guide to the Network Economy 2163. Boston, MA: Harvard Busi- Wang, Q., X. Sun, S. Cobb, G. Lawson, and S. Sharples. 2016. “3D
ness School Press. Printing System: An Innovation for Small-Scale Manufacturing
Shen, X. 2019. “Building ‘Mass’ and ‘Momentum’—A Latecomer in Home Settings?—Early Adopters of 3D Printing Systems in
Country's Passage to Large Technological Systems—The Case China.” International Journal of Production Research 54(20):
of China.” Technology Analysis & Strategic Management 31(5): 6017–32. https://doi.org/10.1080/00207543.2016.1154211.
503–16. https://doi.org/10.1080/09537325.2018.1519184. West, J. 2008. “Commercializing Open Science: Deep Space Com-
Stanko, M. A. 2016. “Toward a Theory of Remixing in Online Inno- munications as the Lead Market for Shannon Theory, 1960–
vation Communities.” Information Systems Research 27(4): 73.” Journal of Management Studies 45(8): 1506–32. https://doi.
773–91. https://doi.org/10.1287/isre.2016.0650. org/10.1111/j.1467-6486.2008.00807.x.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 553

West, J. 2014. “Too Little, Too Early: California's Transient Advan- Paradigm With Lifecycle Personalization Concerns.” Interna-
tage in the Photovoltaic Solar Industry.” The Journal of Tech- tional Journal of Production Research 57(8): 2571–84. https://
nology Transfer 39(3): 487–501. https://doi.org/10.1007/s10961- doi.org/10.1080/00207543.2018.1530475.
012-9291-6.
West, J., and G. Kuk. 2016. “The Complementarity of Openness:
How MakerBot Leveraged Thingiverse in 3D Printing.” Techno- AUTHOR BIOGRAPHIES
logical Forecasting and Social Change 102: 169–81. https://doi.
org/10.1016/j.techfore.2015.07.025. Thierry Rayna is Chaired Professor of Technology
West, J., and M. Bogers. 2014. “Leveraging External Sources of
for Change at the École Polytechnique and a Fellow
Innovation: A Review of Research on Open Innovation.” Jour-
nal of Product Innovation Management 31(4): 814–31. Portico.
of the CNRS Innovation Interdisciplinary Institute
https://doi.org/10.1111/jpim.12125. (i³), France. His research investigates the changes
West, J. 2006. “Does Appropriability Enable or Retard Open Inno- brought about by digital technologies (such as 3D
vation?” In Open Innovation: Researching a New Paradigm, edi- printing, artificial intelligence, blockchain, cloud, IoT,
ted by H. Chesbrough, W. Vanhaverbeke, and J. West, 109–33. platforms) on business strategy, organization, and
Oxford: Oxford University Press. public policy. In particular, his research focuses on the
Westerweel, B., R. J. I. Basten, and G.-J. van Houtum. 2018. “Tradi-
issues of business model innovation, ecosystem gover-
tional or Additive Manufacturing? Assessing Component
nance, and impact (environmental, social, economic),
Design Options through Lifecycle Cost Analysis.” European
Journal of Operational Research 270(2): 570–85. https://doi.org/ in relation to digital technologies. Prof. Rayna has a
10.1016/j.ejor.2018.04.015. long track record of collaboration with companies
Windrum, P., K. Frenken, and L. Green. 2017. “The Importance of (small and large) in various industries “disrupted” by
Ergonomic Design in Product Innovation. Lessons From the digital technologies, and policymakers.
Development of the Portable Computer.” Industrial and Corpo-
rate Change 26(6): 953–71. https://doi.org/10.1093/icc/dtx006. Joel West is Professor of Entrepreneurship at Hilde-
Wohlers, T. 2005. “New Trends and Developments in Additive Fab- gard College, and Professor Emeritus at the Keck
rication.” In Virtual Modeling and Rapid Manufacturing: Graduate Institute. His research has examined the
Advanced Research in Virtual and Rapid Prototyping, edited by impact of technological evolution upon industry struc-
P. B
artolo. London: Taylor & Francis. ture in personal computers, mobile phones, digital
Wohlers. 2012. Wohlers Report 2012. Ft Collins, CO: Wohlers communications, solar energy, and 3D printing. His
Associates.
has also pioneered the understanding of how firms
Wohlers. 2017. Wohlers Report 2017. Ft Collins, CO: Wohlers
Associates.
utilize strategic openness to build competitive advan-
Wohlers. 2020. Wohlers Report 2020. Ft Collins, CO: Wohlers tage through open platforms, open standards, open
Associates. communities, and open consortia. He is the co‐editor
Yoo, Y., O. Henfridsson, and K. Lyytinen. 2010. “Research Com- of Open Innovation: Researching a New Paradigm
mentary—The New Organizing Logic of Digital Innovation: An (2006), New Frontiers of Open Innovation (2014) and
Agenda for Information Systems Research.” Information Sys- the Oxford Handbook of Open Innovation (2023) and
tems Research 21(4): 724–35. https://doi.org/10.1287/isre.1100.
cofounder of the World Open Innovation Conference.
0322.
Yoo, Y., R. J. Boland, K. Lyytinen, and A. Majchrzak. 2012. “Orga-
nizing for Innovation in the Digitized World.” Organization Sci-
ence 23(5): 1398–408. https://doi.org/10.1287/orsc.1120.0771.
Zhan, T., K. Yin, J. Xiong, Z. He, and S.‐T. Wu. 2020. “Augmented SU PP O R TI N G I N F O RMA TI O N
Reality and Virtual Reality Displays: Perspectives and Chal- Additional supporting information can be found online
lenges.” IScience 23(8): 101397. https://doi.org/10.1016/j.isci. in the Supporting Information section at the end of this
2020.101397. article.
Zhang, J., F.‐Y. Wang, K. Wang, W.‐H. Lin, X. Xu, and C. Chen.
2011. “Data‐Driven Intelligent Transportation Systems: A Sur-
vey.” IEEE Transactions on Intelligent Transportation Systems How to cite this article: Rayna, Thierry, and
12(4): 1624–39. https://doi.org/10.1109/tits.2011.2158001. Joel West. 2023. “Where Digital Meets Physical
Zheng, P., T.-J. Lin, C.-H. Chen, and X. Xu. 2018. “A Systematic
Innovation: Reverse Salients and the Unrealized
Design Approach for Service Innovation of Smart Product-
Dreams of 3D Printing.” Journal of Product
Service Systems.” Journal of Cleaner Production 201: 657–67.
https://doi.org/10.1016/j.jclepro.2018.08.101. Innovation Management 40(4): 530–553. https://
Zheng, P., Y. Lin, C.-H. Chen, and X. Xu. 2019. “Smart, Connected doi.org/10.1111/jpim.12681
Open Architecture Product: An IT-Driven Co-Creation
Received: 31 May 2021 Revised: 29 October 2022 Accepted: 11 March 2023
DOI: 10.1111/jpim.12681

ORIGINAL ARTICLE

Where digital meets physical innovation: Reverse salients

and the unrealized dreams of 3D printing

Thierry Rayna 1 | Joel West 2,3

1

i³-CRG, Ecole Polytechnique, CNRS,
Institut polytechnique de Paris, Palaiseau, Abstract
France For more than three decades, enthusiasts have predicted that direct
2
Hildegard College, Costa Mesa, manufacturing enabled by 3D printing would inevitably supplant traditional
California, USA
3
manufacturing methods. Alas, for nearly as long, these utopian predictions
Riggs School of Applied Life Sciences,
Keck Graduate Institute, Claremont,
have failed to materialize. One reason is a flawed assumption that hybrid
United States digital-physical systems such as 3D printing would advance as rapidly as
purely digital innovations enabled by Moore's law. Instead, like other exam-
Correspondence


Thierry Rayna, i³-CRG, Ecole ples of cyber-physical systems (CPSs), technological progress in 3D printing
Polytechnique, CNRS, Institut faces inherent limitations that are emblematic of the differences between
polytechnique de Paris, Palaiseau, France.
CPSs and purely digital innovations. As with any complex CPS, improved
Email: [email protected]
performance of a 3D printing system has been limited by that of its key
Special Issue Guest Editors: Michael components—the sort of limiting problem previously defined as a reverse
Stanko and Aric Rindfleisch
salient. Unlike previously studied technologies, several reverse salients for
3D printing performance have neither resolved nor signs of resolving soon.
Here we analyze these key reverse salients, and show how they have ham-
pered the suitability of 3D printing for direct manufacturing and other pre-
dicted applications. We contrast predicted versus actual capabilities for 3D
printing-enabled transformation in six key areas: product innovation, mass
customization, home fabrication, distributed manufacturing, supply chain
optimization and business model innovation. From this, we suggest opportu-
nities for greater realism in future 3D printing research, as well as broader
implications for our understanding of CPSs and reverse salients.

KEYWORDS
3D printing, cyber-physical systems, digital innovation, innovation, reverse salients

1 | INTRODUCTION synthesizer,” a 23rd-century technology that allowed

food to be manufactured on demand for crew members
In September 1966, viewers of the American science light years from home. In 1987, the next Star Trek
fiction series Star Trek were introduced to the “food series featured a voice-controlled “replicator,” a 24th-
century technology that could make almost anything.
Thierry Rayna and Joel West contributed equally to this study. These thus popularized the dream of on-demand

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2023 The Authors. Journal of Product Innovation Management published by Wiley Periodicals LLC on behalf of Product Development & Management Association.

530 wileyonlinelibrary.com/journal/jpim J Prod Innov Manag. 2023;40:530–553.

 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 531

computer-controlled production of arbitrary physical

objects.1 Practitioner points
Also in 1987, in a Los Angeles suburb 20 miles from
Star Trek's Hollywood studio, a new firm named • 3D printing points out the risk of planning
3D Systems, Inc. produced and sold the world's first addi- around utopian predictions of technology
tive manufacturing system. Based on a technology called performance.
stereolithography—patented by founder Chuck Hull— • We show how persistent reverse salients can
these systems could create complex 3D shapes by use of a slow the performance and adoption of a hybrid
computer-controlled laser that solidified liquid resin. digital + physical technology.
DTM Corporation was formed that same year to commer- • We offer a framework for identifying which
cialize selective laser sintering, a UT Austin invention by systems and systems components are most
Carl Deckard that created 3D objects by fusing metal likely to face these persistent performance
powder.2 These two firms thus pioneered what eventually limitations.
became known as 3D printing.
As in their sci-fi archetypes, such products offered
the promise to render arbitrary digital designs into a transformation made possible by the adoption of digital
tangible form. For more than 30 years, these complex computing and communications enabled by large-scale
and expensive systems have been predicted to inevitably circuit integration, rapid improvements in computing
transform manufacturing industries (e.g., Ben-Ner & performance and the malleability of software (Schaller,
Siemsen, 2017; Ghaffar & Mullett, 2021; Li, 2013; 1997; West, 2008).
Peltz, 1987; Petrick & Simpson, 2013). For example, in a We thus consider the impact of the limits of digital
2011 story entitled “Could 3D printing change the and physical technologies that are inherent for digital
world?” a leading U.S. foreign policy think tank breath- systems that serve physical beings who are living in a
lessly predicted: material world. We do so for a broad class of digital-
physical hybrid innovation known as cyber-physical sys-
3D Printing/Additive Manufacturing (AM) is tems (CPSs) that integrate digital and physical compo-
a revolutionary emerging technology that nents in a single system (Lee, 2008; Rajkumar
could up-end the last two centuries of et al., 2010). The study of CPSs often emphasizes adding
approaches to design and manufacturing digital intelligence and connectivity to mature physical
with profound geopolitical, economic, social, technologies, but here we study the case of 3D printing,
demographic, environmental, and security in which digital technologies are combined with imma-
implications (Campbell et al., 2011, p. 1). ture physical technologies.
We begin the article by reviewing how theories of
Two years later, President Barack Obama predicted in his both digital innovation and CPSs explain innovation tra-
2013 State of the Union address that 3D printing “has the jectories for systems that combine digital and physical
potential to revolutionize the way we make almost components. We then consider how these trajectories are
everything.” often limited by a critical technological bottleneck that
Instead, the reality of 3D printing has fallen short of Thomas Hughes (1983, 1987) defines as a reverse salient.
these early visions, as technological performance and This article examines 3D printing as a specific case study
thus adoption have consistently failed to meet predictions for understanding the reverse salients of CPSs. Using more
during these years. Such hype is not unique to 3D print- than 80 articles from scientific journals on digital manufactur-
ing, as many forms of digital innovation have attracted ing and other applications of 3D printing, we trace the pro-
technologically utopian visions to mobilize public sup- gress and limitations of both the digital and physical
port and adoption (Kling, 1996; Turner, 2010). However, components of this technology over the past 35 years, and
purely digital innovations have fulfilled such visions identify the key reverse salients for 3D printing. From this,
during this same period, as witnessed by the global we offer broader implications and research opportunities for
3D printing, CPSs and the study of reverse salients.
1
The first known mention of additive manufacturing in science fiction
came in a 1947 short story in Astounding magazine by Eric Frank
Russell, which referred to “atom fed to atom like brick after brick to
2 | L I T E R A T UR E R E V I E W
build a house” (Nerlich, 2005).
2
3D Systems acquired DTM in 2001, the first of many such acquisitions.
Deckard died in 2019, while Hull remains executive vice president of 3D In recent decades, our world has been transformed by
Systems. ubiquitous digital innovation. One widely cited definition
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
532 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

of such innovation is “the carrying out of new combina- physical, one option is that the physical enables the digi-
tions of digital and physical components to produce novel tal. The most dramatic example of ongoing improve-
products” (Yoo et al., 2010, p. 725). However, subsequent ments in semiconductor manufacturing under Moore's
researchers concluded “What is missing to date is an law have made predictable and rapid improvements in
understanding of how digital transformation manifests computing performance (Moore, 2006; Schaller, 1997)—
itself in industries in which the core products are primar- in turn enabling advances in digital innovation (Autio
ily physical” (Hanelt et al., 2015, p. 1313). We therefore et al., 2021), However, in this article, we focus on the
are interested in this question: How do technological lim- converse case: how the digital enables the physical. An
itations (specifically reverse salients) constrain the tech- examination of digitally enabled technologies of the late
nological trajectory of a system that requires both digital 20th and early 21st century suggests three different pat-
and physical innovations? We also consider what hap- terns for how the digital enables the physical (Table 1):
pens when such limits are either not recognized, or are
assumed to be temporary. 1. Digital represents the physical, where the digital good
We thus review three prior bodies of research that is an intermediate representation of a good that will
relate to digital + physical innovation. The first is how be consumed in physical form. This includes objects
digital innovation considers systems combining digital printed on demand, whether a book or a spare part
and physical components, while the second is research (Jiang et al., 2017; Osterwalder & Pigneur, 2004).
on CPSs. We extend this with prior research on reverse Information goods such as text, pictures and
salients in physical and digital systems to suggest how audiovisual material are a special case, because the
reverse salients could impact the performance trajectory value is the information contained in the object
of a CPS such as 3D printing. rather than the object itself (Negroponte, 1996;
Rayna, 2008).
2. Digital coordinates the physical: digital technologies
2.1 | The promise of integrating digital are used to coordinate activities in the physical world,
and physical innovation whether e-mail, a project plan or an ecosystem map
(Boland et al., 2007; Markus, 1994; Randhawa
Digital technology removes the existing constraints of et al., 2020). For example, to design and make physi-
physical innovation and makes possible a great degree of cal objects, the open-source hardware movement has
redefining and reconfiguring the nature of the innovation used many of the same collaboration processes, social
(Rindfleisch et al., 2017; Yoo et al., 2012). The malleabil- norms and IP rules as those used for producing open
ity of digital technologies thus enables experimentation source software, a purely digital good (Raasch
and improvisation in a way not possible with physical et al., 2009; West & Kuk, 2016).
innovation (Nylén & Holmström, 2015), which has pow-
erful implications for the nature of innovation itself.
First, these reduced constraints—plus the use of a com-
mon digital interface for a wide range of problem T A B L E 1 Taxonomy of digital innovations that depend on the
domains—mean that the pace of innovation has acceler- physical world.
ated (Yoo et al., 2010).
Category Example
Second, digital innovations are often not governed by
Digital represents Print-on-demand 2D objects
the well-known S-curves limiting performance trajecto-
the physical (Osterwalder & Pigneur, 2004) and
ries of physical technologies (Priestly et al., 2020). Even if
3D objects (Jiang et al., 2017);
such S-curves exist for specific domains for key enabling Entertainment and other information
technologies—such as quantum computing (Hill & goods (Negroponte, 1996;
Rothaermel, 2003) or artificial intelligence (AI) (Iansiti & Rayna, 2008); innovation as data
Lakhani, 2020)—advances in general purpose enabling (Rindfleisch et al., 2017)
technologies will accelerate societal innovation more Digital coordinates E-mail (Markus, 1994); boundary objects
broadly than previous S-curves related to product- or the physical (Boland et al., 2007; Randhawa et al.,
industry-specific technologies such as aircraft or wind 2011); open source hardware (Raasch
turbines. et al., 2009)
While digital technologies also impact the nature of Digital controls Manufacturing (Åstebro, 2002) and
physical innovation, that intersection remains a rarely the physical transportation systems (Zhang
explored frontier (Hanelt et al., 2015; Lee & Berente, et al., 2011); robots (Fujita, 2001;
Sturgeon, 2019)
2012; Sturgeon, 2019). When the digital meets the
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 533

3. Digital controls the physical: digital technologies con- mention issues related to digital technologies, regulation
trol activities in the physical world—such as complex and legal issues, or the need to find new business models.
manufacturing systems (Zhang et al., 2011), self- Rarely considered issues are arising from the physical
driving cars (Lyytinen, 2022), and personal or indus- aspects of the CPS itself.
trial robots (Fujita, 2001; Sturgeon, 2019). The focus on the cyber part of a CPS has often been
for the control of updated versions of existing
While all three patterns apply to 3D printing, here we physical objects, such as home appliances, thermostats,
focus on the technological challenges of how the digital automobiles or factory equipment (Borgia, 2014; Hanelt
controls the physical. By its nature, the final results of et al., 2015; Lee et al., 2015). Despite this focus on mature
such this sort of system will be delivered (and con- technologies and equipment, there are examples of CPSs
strained) by the physical processes that are digitally that rely on physical components that are yet immature.
controlled. An emerging example is the so-called metaverse, which
requires affordable, lightweight, durable and high-
definition virtual reality headsets. Another example is 3D
2.2 | Cyber-physical systems printing, which has remained an emerging technology
for decades, and can thus help us understand the role
The most commonly studied examples of innovations similar limitations can have for other CPSs.
where the digital controls the physical are CPSs, first
defined as “integrations of computation with physical
processes” (Lee, 2008, p. 363). The name captures how 2.3 | Reverse salients in physical
the “cyber-world of computations and communications” and CPSs
integrated with “real-world physical activities” (Lien
et al., 2012). A more recent definition states that CPS Innovation in physical technologies often follows a punc-
“comprise interacting digital, analog, physical, and tuated equilibrium model, which assumes a long period
human components engineered for function through of incremental innovation within a given technological
integrated physics and logic” (NIST, 2020). paradigm, interrupted by infrequent discontinuities that
Examples of CPS research have included the Inter- start a new paradigm (Dosi, 1982; Garcia & Calantone,
net of Things (IoT) (Baars et al., 2014; Mahmood & 2002; Nelson & Winter, 1982). Technological progress
Mubarik, 2020) and Industry 4.0 (Greer et al., 2019; within a given paradigm will follow a specific trajectory
Lee et al., 2015). While innovation is often mentioned (Dosi, 1982; Nelson & Winter, 1982) that is only clear
in the academic literature related to CPS, the innova- after the fact. The revealed trajectory depends on a com-
tion processes are rarely studied. Those who do eso bination of market, technical and organizational factors
xplicitly, typically raise the issue of CPS as opportuni- (Bijker & Law, 1994, p. 18; Kim & Kogut, 1996, p. 285).
ties for Open Innovation (Mubarak & Petraite, 2020; Complex innovations are often delivered as systems
Zheng et al., 2019), service and product innovation that integrate a range of components and technologies.
(Li, 2018; Li et al., 2021; Mubarak & Petraite, 2020; While such component reuse offers powerful economies
Zheng et al., 2018), and business model innovation of scale, it provides challenges for the firm both in attain-
(Li et al., 2018; Mahmood & Mubarik, 2020). ing economic success, and also for managing the com-
However, the innovation potential of CPS identified plexity and interdependencies of the various components
in the literature chiefly relates to progress made and (Davies & Brady, 2000; Majidpour, 2016; Prencipe
brought about by the digital side, as with studies of AI, et al., 2003; West, 2006).
IoT or blockchain (Bongo et al., 2020; Li, 2018; Li et al., Within these systems, the technological progress on one
2021). In these studies, the physical aspects of the CPS or more of these components will lag the rest of the system
appear to have been largely overlooked. and thus be the limiting factor of improving overall system
This emphasis on the digital side is prominent in performance. By analogy to a military front, Hughes (1983,
research that bridges theories of digital innovation and 1987) terms such a performance-limiting component a
CPSs, which emphasize the new capabilities and relaxing reverse salient, which Lehenkari & Miettinen (2002, p. 111)
of design constraints made possible by digital control of later summarized as “when one system component—
existing systems (Hendler, 2019; Herterich et al., 2016; technical, organizational or economic—has fallen behind
Lyytinen, 2022; Sandberg et al., 2020). In fact, articles or is out of phase with the others, and impedes the further
investigating either success factors or challenges and bar- development of the system.”
riers to CPS adoption (e.g., Bongo et al., 2020; Leitão Here our interest is in the former case, specifically that
et al., 2020; Sinclair et al., 2019) appear to chiefly subset of technological bottlenecks for components that
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
534 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

limit the performance improvement of the entire system. changes to data exchange interfaces that constrained
Hughes (1983, 1987) suggests that the image of a “bottle- performance.
neck” is a less accurate metaphor for such limits than a Other reverse salients can only be resolved through
salient within a broad technological front. Some have trea- changes to sociotechnical institutions. For example,
ted the overlapping concepts of bottleneck and reverse when compatibility standards both provide interoperabil-
salient as being related issues in technological change ity but constrain innovation, improvements to the shared
(Baldwin, 2015; Windrum et al., 2017) while others have standard can become the limiting factor for improving
used the terms synonymously (e.g., Davies, 1996). performance, such as in digital telecommunications
For Hughes, the solution to such a salient comes from (Dedehayir & Hornsby, 2008; West, 2008). When a
entrepreneurs, inventors and other scientists who define reverse salient is deeply embedded in society, overcoming
“the reverse salient as a set of ‘critical problems,’” and that salient is often more complex to overcome and thus
then work to correct those problems (Hughes, 1983, requires a more comprehensive approach to resolve
p. 80). This definition is fueled by the assumption that (Mathews et al., 2018; Shen, 2019).
“such problems are believed to be solvable. The belief Overall, studies of specific reverse salients have tended
that a reverse salient might be corrected increases dra- to emphasize successfully overcoming the salient. Such
matically by its articulation in a set of critical problems” research includes salients that are “solved” (Christiansen &
(Mulder & Knot, 2001, p. 266). While studies by Hughes Buen, 2002; Windrum et al., 2017) or “resolved” (Sandberg
and many others typically emphasize absolute perfor- et al., 2020) to achieve the system's “potential” (Daim
mance, the critical problem can also be one of efficiency et al., 2014; Dedehayir & Mäkinen, 2011b; Sandberg et al.,
or cost effectiveness, particularly for industries that 2020); this may reflect a long-remarked tendency toward
require economies of scale and scope (Chandler, 1990; success bias in studies of innovation (Coopersmith, 2020;
Davies, 1996). Rogers & Shoemaker, 1971). Thus there is a gap in research
The original theory was developed from Hughes' stud- that considers salients that are not resolved or overcome.
ies of the electrical power infrastructure, but the concept
subsequently demonstrated its relevance in a wide array
of sectors and applications, mainly railroads and rail- 3 | REVERSE SALIENTS IN 3D
ways, aviation, electric vehicles, ICTs, production, tele- PRIN TING
phone, and wireless services (Dedehayir, 2009). In
particular, subsequent studies of physical systems have We are interested in how the digital and physical
examined renewable power generation (Christiansen & components of a CPS—and the interaction between these
Buen, 2002; Hoppmann et al., 2014) and airplane engines components—serve as reverse salients for that system's
(Bowles & Dawson, 1998). While many studies have technological trajectory. We do so in the context of a spe-
focused on a specific salient, a few have examined the cific CPS: 3D printing. Studying 3D printing is an
impact of multiple salients on technological progress extreme case of CPS due to the perfect integration
(Dedehayir & Hornsby, 2008; Sandberg et al., 2020; required between the digital and physical to realize the
Windrum et al., 2017). longstanding promise of a one-to-one correspondence
More recent studies have examined reverse salients of between the digital and physical worlds. Such a “bits to
digital innovation. Since the 1980s, CPU speed has been atoms” solution would fulfill the science fiction dream of
the limiting factor for most general- and special-purpose materializing any digital object as a physical one (Millard
computing systems (Daim et al., 2014; Dedehayir & et al., 2018; Nyman & Sarlin, 2014; Söderberg & Daoud,
Mäkinen, 2011a), while specialized applications have 2012). While originally the best-known subset of a broader
been limited by graphic performance (Dedehayir & category of technologies termed additive manufacturing,
Mäkineif, 2008) or portability (Windrum et al., 2017). In the term 3D printing has since been commonly used to
fact, the PC industry suggests that different types of refer to the entire category (e.g., Wohlers, 2005).
reverse salient may emerge over time, not only between Consistent with other reverse salient studies
different hardware components (Dedehayir & Mäkinen, (e.g., Lehenkari & Miettinen, 2002; Lenfle & Söderlund,
2011a), but also between hardware and software compo- 2022; Sandberg et al., 2020), we identify reverse salients
nents (Dedehayir & Mäkinen, 2011b), making identifying using a longitudinal case study that includes economic
and forecasting reverse salients particularly critical and and technological factors that shaped the evolution of
difficult (Han & Shin, 2014). In the Sandberg et al. (2020) this technology over the past 35 years. Much like these
study of digitalization of previously physical-only process earlier studies, we used a combination of (a) research on
control systems—a rare study of reverse salients in the history of the science and innovation in this field;
CPSs—tight coupling of system components limited (b) scientific literature, in this case from manufacturing
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 535

and materials science journals; (c) economic and business material (such as metal or wood) is cut down to final shape
analyses of 3D printing technology and applications; and using cutting devices such as a saw, lathe or mill. While
(d) comprehensive industry reports—particularly the the CNC process worked with materials that had been
Wohlers Report—about industry statistics and trends. Our used for centuries, as in the pre-CNC era it remained lim-
analysis was informed by previous interviews from our ited by the mechanics of the cutting devices. Initial CNC
respective earlier studies of 3D printing. adoption was also limited by the quality and availability of
The predicted rapid adoption of 3D printing digital Computer Aided-Design (CAD) software, as well as
technologies has failed to materialize due to how they dif- the employee skills necessary to use this software to create
fer from other digital technologies. Indeed, the optimistic digital designs (Åstebro, 2002; Tilbury, 2019).
predictions for 3D printing all implicitly assumed that it Thus nascent additive manufacturing technologies of
would follow the same trends, adoption and use patterns the 1980s could build upon the CAD software developed
as purely digital technologies. But as a CPS, 3D printing for CNC applications. Early 3D printers were first con-
systems require a complex integration of a broad range of trolled by dedicated CAD workstations and later using
both digital and physical technologies. Because improve- personal computers (Jacobs, 1992; Shane, 2000). They ini-
ments in physical component technologies have not kept tially faced the same limitations of CAD software and
up with the digital components, they act as reverse designer skills, but digitally enabled advances in CAD
salients for the overall performance—in terms of speed, software improved the usability of these tools (de Jong &
quality, cost and degree of automation—of how 3D print- de Bruijn, 2013).
ing can create physical artifacts. Table 2 summarizes six
reverse salients of 3D printing, of which five are tied to
the physical side of 3D printing. 3.1.2 | Consumer design tools
First, we examine the digital and physical reverse
salients limiting improvements in 3D printing perfor- The original CAD software was solely intended for profes-
mance, and we then consider the effects of those limits sionals, usually with a drafting or engineering background.
upon the societal and economic transformations Even with improvements in usability for professionals in
predicted to occur through widespread adoption of the early 21st century, further improvements are still
3D printing. needed to make 3D designing as usable for ordinary con-
sumers as 2D design tools (Rayna et al., 2015). For exam-
ple, Digital Forming, a London startup founded in 2008 to
3.1 | Digital components of 3D printing address this problem, promised that “our software har-
nesses the power of mass customization”; instead, the
3.1.1 | Design tools company was declared insolvent in 2019.

To design physical objects, 3D printing build upon one of

the earliest CPSs, Computer Numerical Control. With the 3.1.3 | Rendering layers
invention of the microprocessor, starting in 1972, CNC sys-
tems provided digital control of the mature and reliable Compared to CNC and other subtractive manufacturing,
subtractive model of manufacturing, where a raw block of 3D printing allows the design of arbitrary shapes made by

TABLE 2 Limits and reverse salients in 3D printing.

Category Limitation Current status Salient

Digital Professional design tools Mature
Consumer design tools Slowly improving X
Rendering layers Mature
Distribution of digital designs Mature
Physical Replicability: repeatability and reproducibility Ongoing challenges X
Indirect replicability: digitization Slowly improving X
Costs Slowly improving X
Time Slowly improving X
Skills and knowledge Ongoing challenges X
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
536 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

adding one layer upon top of another. However, it required 3.2.1 | Direct replicability of artifacts
invention of entirely new material science technologies for
fabricating physical objects, as well as digital algorithms to Replicability is a serious and persistent challenge for 3D
translate a 3D digital design into a series of 2D layers, printing systems. Aggregating molecules of a particular
printed one at a time. While the basic layer printing issues material layer by layer never happens the exact same way
were addressed within a decade, repeatability requires twice. These inconsistencies in printing the same digital
developing precision solutions for each new manufacturing object arise even with similar printer technologies, and
material (Gibson et al., 2014; Jacobs, 1992). sometimes with two successive prints from the same
printer and materials (Dowling et al., 2020; Kim
et al., 2018). Efforts to accurately predict the output (a 3D
3.1.4 | Digital distribution print preview) depend on physical phenomena taking
place at molecular level and today require computations
The digital designs for 3D printing are digital artifacts, by a supercomputer (Goh et al., 2020; Sahini et al., 2020).
which by their nature are stored and replicated without a The direct replicability of 3D printing assumed in
loss of information (Quah, 2003; Rayna, 2008). Since the prior forecasts require both repeatability and reproduc-
advent of digitalization in the late 20th century, the stor- ibility (Khorasani et al., 2021; Mussatto et al., 2022; Salmi
age, duplication and eventual dissemination of informa- et al., 2013). Repeatability means creating identical
tion goods has moved from the analog to the digital objects using the same operator, environment, and equip-
and from the physical to the electronic (Rayna & ment (e.g., 3D printer, software, materials) in a short
Striukova, 2021; Shapiro & Varian, 1999). Online com- interval of time. Reproducibility means obtaining the
munities were created to share or sell these 3D printable same results with different operators, materials, and
designs, of which the best known and most often studied equipment. The latter is still a dream, but even the for-
is Thingiverse (Claussen & Halbinger, 2021; Kyriakou mer remains unsolved (Dowling et al., 2020).
et al., 2017; Naik et al., 2021; Stanko & Allen, 2022; A final failure of utopian forecasts for materializing
West & Kuk, 2016). physical objects is that 3D printed objects typically
While such digital designs can be cheaply, reliably require manual post-processing, such as sanding, polish-
and quickly replicated, the same is not true for the physi- ing or even painting; sometimes these steps can only be
cal object 3D printed from such design. In principle, this done by hand (Sgarbossa et al., 2021; Thürer et al., 2021).
distinction applies to the materialization of other digital Some designs also require post-processing assembly.
goods such as a photo or sound recording, but such infor- Beyond the increase in time and cost, this nonautomated
mation goods can be viewed or played on a computer post-processing reduces the consistency of replicating 3D
without being materialized. printed artifacts.

3.2 | How the physical limits digital 3.2.2 | Digitalization and indirect
innovation replicability

Printing arbitrary 3D objects required developing new Whether by CAD design or digitization, digital files used
materials and material science. During the industry's for 3D printing typically only describe the outer shape of
first 15 years, 10 new patent families (by 10 different the object. The internal structure (including wall thick-
inventors) were introduced, each creating a different ness) needed to instantiate the object is generated prior
materials technological trajectory and seeking to over- to each print, depending on the software, the printer, the
come the limitations of other 3D printing technologies; materials and user settings. While solid objects are the
however, all had significant limitations and none cap- norm with traditional fabrication processes such as injec-
tured a dominant market position (West & Kuk, 2016). tion molding and milling, the opposite is true for 3D
The proliferation of technologies, patents and firms con- printing, where printing solid objects is slow, expensive
tinued during the initial decades of the 21st century, and failure prone (Rouf et al., 2022).
with a common core of patents shaping the entire indus- For printing digital objects, indirect replicability
try until 2008 (Wang et al., 2020). Despite this high rate would mean that copies of copies are not only identical
of innovation and new entry, these technologies have to each other, but also to the original (Quah, 2003;
failed to overcome the cost and speed advantages that Rayna, 2008). Unlike digitizing media streams or 2D
high-volume manufacturing provides for most applica- objects, commercial and consumer 3D scanning tools are
tions (Jiang et al., 2017). decades behind 2D counterparts (Conkle et al., 2018):
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 537

widely available 3D scanners can copy the outer layer of technologies. Just as physical limits mean that semicon-
the object, but not its internal structure, leaving unre- ductor manufacturing faces inextricable challenges at the
solved how the object's structure should be replicated nanoscale (Khan et al., 2018), the lack of progress in
(Javaid et al., 2021). While the photograph-scanning 3D printing speed for at least some of the technology
model has been effective for digitizing what objects look branches is due to physical constraints that will be diffi-
like, it will not tell designers how to replicate those cult to resolve, in part due to the unpredictable nature of
aspects of an object's internal structure that determine its the melting/congealing processes required by most 3D
strength, weight and other key properties.3 Even with a printing technologies (Böckin & Tillman, 2019).
3D scan that produces an accurate copy of the original
artifact, the information embedded in this scanned blue-
print would most likely be radically differ from the origi- 3.2.5 | Human skill and knowledge
nal blueprint due to inherently missing information.
Replication of digital objects has been enabled by auto-
mated processes that minimize the required human
3.2.3 | Costs knowledge and skills. Extrapolated to 3D printing, this
brought predictions that consumers would be able to
In addition to high quality, purely digital artifacts can be manufacture objects at home instead of purchasing them
replicated at a negligible cost by reusing digital storage in a store, as soon as 3D printers became widely available
for other purposes. However, 3D printing requires raw (Bogers et al., 2016; Jiang et al., 2017; Rayna &
materials that cannot be reused and are difficult to recy- Striukova, 2015). However, after decades 3D printing still
cle (Ford & Despeisse, 2016). Not only does 3D printing remains a long way from providing fully automated fabrica-
an object have a significant cost, but that cost is tion and “zero skill manufacturing” (Eyers & Potter, 2017).
irrecoverable. Because of the lack of consistency and reliability, the use
Unlike smartphones and other general-purpose com- of 3D printers requires specific craft-like knowledge for both
puting devices, 3D printing also requires specialized up effective design and manufacturing (Despeisse & Minshall,
front capital investments. Not only does 3D printing 2017; Pradel et al., 2018). Different settings are required for
require a 3D printer, but today such printer is only suit- each combination of printer and material—and often for
able for printing a particular type of object with a specific specific objects. As noted in Rouhani (2018, p. 17):
material such as metal, ceramics, or plastic. In contrast to
the multifunctionality of digital devices, the idea of a Mastering FDM 3D printing requires a lot of
single 3D printer that can automatically manufacture tinkering and patience. There are many print
arbitrary objects remains a utopian fantasy for the fore- quality issues common to FDM such as
seeable future. warping, step lines, stringing, under/over
extrusion, layer shifting, and many, many
more. While there are solutions to all of
3.2.4 | Time these problems, it can take multiple attempts
in order to get the desired finished look.
Often overlooked is that the replication of 3D printed FDM machines have a lot of moving parts
objects also requires significant time. Replicating a digital and settings which need to be adjusted
artifact is usually a matter of seconds. 3D printing time according to the geometry of the part, the
depends on the shape, size, geometry and material, but environment the printer is in, and the
typically a small object will require 1–3 h, a medium expected print quality.
sized object 12–48 h, while the largest may take up to a
week (Avinson, 2018). Some improvements have been made—for example,
While print speed for some materials has improved in self-calibrating printers—but 3D printing is unlikely to
recent years, overall speed of 3D printing technologies become “plug and play” in the foreseeable future.
has not followed a similar pattern as other digital Instead, the modest successes of 3D printing have been
enabled by tacit knowledge developed and applied by
3
In principle, the issue can be solved with a medical grade CT or MRI skilled artisans. Just as in 19th century machine tools
machine that can render (and thus replicate) a single item made of
and 20th-century CNC machines, the 21st-century opera-
homogeneous materials such as a replacement bone implant
(Randhawa et al., 2021). However, even this technology is unable to
tion of additive manufacturing systems requires skilled
automatically detect and replicate the structure (let alone materials) of a workers who know their machine and can tell what will
complex assembled product. or will not work. And while digital knowledge is reliably
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
538 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

and costlessly replicated, such tacit knowledge is inher- manufactured by any subtractive (e.g., milling) or
ently neither (Nonaka & Takeuchi, 2007) and is also diffi- transformative (e.g., injection molding) manufacturing
cult to transfer: solving one 3D printing problem does not technology.
directly solve another, although experience provides a
starting point for diagnosing and addressing problems. These factors are tied to the application of 3D printing,
whether building prototypes (Rapid Prototyping), casts,
molds and jigs (Rapid Tooling), end-user parts or objects
3.3 | Effect of 3D printing reverse (Direct Manufacturing), or for Distributed Manufacturing
salients (Rayna & Striukova, 2016).

Due to these limits, after three decades of optimistic

predictions, additive manufacturing has failed to displace 4 | A S S E S S I N G PR E D I C T I O N S
potential niche applications addressed by CNC O F 3 D P R I N T I N G - E N A B LE D
machines—let alone the broader market for high volume TR ANSFORMATIONS
formative manufacturing technologies such as injection
molding or casting. As one frustrated potential user said Using the framework a for understanding the reverse
in 2016: “At this stage, AM machines are a waste of salients of 3D printing developed in Section 3, to reassess
money due to immature technology. These machines extant knowledge and generate new insights for under-
have very high operation and maintenance costs. There- standing the phenomenon we conducted a problematiz-
fore, we are not investing any money on AM technology, ing review of research on 3D printing applications
and the bulk of our income is from CNC machining” (as detailed in Appendix A). From the corpus of 84 jour-
(Wohlers, 2017, p. 170). nal articles shown in Appendix B, we identified six major
Meanwhile, 3D printing failed to realize the promise themes for predicted societal or economic impacts of 3D
of large-scale distributed replication that is today possible printing (Table 3). For each theme, we show how the
for the digital representations but not for the physical impact of 3D printing in that area has been limited by
objects created from those representations—which would one or more of the previously mentioned reverse salients
be necessary achieve predicted transformational out- in 3D printing performance.
comes. This reveals a sharp divide between the physical
side and the digital side, because the overall performance
of the technology is limited by the reverse salients of the 4.1 | Product innovation
physical side.
Given these inherent physical limitations, 3D printing Product innovation is one of the most frequently cited
is only competitive under one or more of the following transformational impact of 3D printing. In practice, this
conditions: has meant Rapid Prototyping (RP), once the only applica-
tion of 3D printing (Hull et al., 1995), and still its most
• Short lead time: despite its comparatively slow speed, prevalent use (Sculpteo, 2020; Wohlers, 2020), because
3D printing has an advantage when time constraints limitations outldeined in Section 3 do not act as reverse
are very tight, as it does not require tools to be built salients for making prototypes.
prior to manufacturing. This applies only when a small RP has had critical benefits for product innovation
number of units are needed, after which the faster pro- (Cooper, 2021; Haleem et al., 2020; Luchs et al., 2015;
duction speed of injection molding will make up for Rayna & Striukova, 2021), especially when done collec-
the time required to build tools (Achillas et al., 2015; tively (Bogers & Horst, 2014). However, these benefits are
Blakey-Milner et al., 2021; Chaudhuri et al., 2021). being eroded by advances in virtual reality (VR), aug-
• Small batches: the lack of tooling cost makes 3D print- mented reality (AR), and digital twin technologies
ing particularly attractive when small numbers of units (Coutts et al., 2019).
are produced. However, the (global) lack of economies Product innovation can also be accelerated by Rapid
of scale for 3D printing compared to traditional Tooling (Fontana et al., 2019; Wohlers, 2020)—printing
manufacturing (Baumers & Holweg, 2019), makes it molds, casts or jigs used in conventional high-volume
less competitive (ceteris paribus) when more than a manufacturing techniques (Radstok, 1999; Rayna &
few dozen to hundred units have to be manufactured Striukova, 2016). This combines advantages of estab-
(Minguella-Canela et al., 2019) lished manufacturing processes (cost, speed, and mate-
• High complexity: 3D printers enable the manufacture rials) and 3D printing (cheaper and more complex molds
of designs of such complexity that they could not be and tools). For instance, 3D printed tire molds enabled
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 539

TABLE 3 Societal impacts of 3D printing predicted by prior research.

Societal impact Prior research Main themes

Product Weller et al. (2015), Stanko (2016), Flath et al. (2017), Disruptive innovation; innovation from data; rapid
innovation Murmura and Bravi (2017), Rindfleisch et al. and iterative prototyping done collectively; real-
(2017), Steenhuis and Pretorius (2017), Kwak time customer feedback; complex tooling to
et al. (2018), Marzi et al. (2018), Candi and enable product innovation; design freedom; 4D
Beltagui (2019), Steenhuis et al. (2020), designs; new product development; no-assembly
Rayna and Striukova (2021b), Ukobitz and products; process innovation; product remixes;
Faullant (2021) technological development; open innovation;
open source
Mass Petrick and Simpson (2013), Rayna et al. (2015), Weller Customer and consumer involvement in design and
customization/ et al. (2015), Bogers et al. (2016), Rayna and Striukova manufacturing; economies of one; knowledge
co-creation (2016), Oettmeier and Hofmann (2016), Steenhuis and reuse; large-scale product customization and
Pretorius (2016), Attaran (2017), Khorram Niaki and co-creation; platforms; pricing; value creation
Nonino (2017), Kyriakou et al. (2017), Aquino et al.
(2018), Shukla et al. (2018), Chaney et al. (2021),
Lacroix et al. (2021)
Prosumption/ Fox (2014), Appleyard (2015), Rayna et al. (2015), Bogers DIY; consumer behavior and adoption; prosumption
home et al. (2016), Kothman and Faber (2016), Rayna and (consumers involved in production process); home
fabrication Striukova (2016), Rogers et al. (2016), Steenhuis and fabrication with 3D printers; intellectual property;
Pretorius (2016), West and Kuk (2016), Wang et al. open design; online design and manufacturing
(2016), Yoo et al. (2016), Despeisse et al. (2017), Flath communities; move from e-commerce to file-
et al. (2017), Jiang et al. (2017), Halassi et al. (2019), commerce; piracy; social manufacturing; user
Kleer and Piller (2019), Naghshineh et al. (2021), Rayna innovation
and Striukova (2021a)
Distributed Garrett (2014), Gress and Kalafsky (2015), Ford and Advanced manufacturing; circular economy;
manufacturing/ Despeisse (2016), Huang et al. (2016), Laplume developing countries; digital encapsulation;
global value et al. (2016), Sasson and Johnson (2016), Despeisse production ecosystems; global factory; industry
chains et al. (2017), Durão et al. (2017), Garmulewicz et al. 4.0; internationalization; value chain
(2018), Hannibal and Knight (2018), Rehnberg and reconfiguration; resource efficiency; product
Ponte (2018), Xu et al. (2018), Kleer and Piller life cycle; geographic span and density;
(2019), Li et al. (2019), Sauerwein et al. (2019), economies of scale and scope; long tail; make-
Beltagui et al. (2020a, 2020b), Hannibal (2020), to-order: manufacturing supercenters;
Rinaldi et al. (2022) organizational structure; restructuring; spare
parts; sustainability
Logistics/supply Baumers et al. (2015), Bogers et al. (2016), Glocalized production; barriers; facility location;
chains Holmström et al. (2016) Sandström (2016), hybrid manufacturing; intellectual property rights;
Oettmeier and Hofmann (2016), Rogers et al. integration; inventories; logistics management;
(2016), Durach et al. (2017), Rylands et al. (2016), SMEs; supply chain innovation; supply chain
Li et al. (2017), Ryan et al. (2017), Schniederjans flexibility supply chain management components
(2017), Chan et al. (2018), Ghadge et al. (2018), and processes; supply chain performance; supply
Eyers et al. (2018), Martinsuo and Luomaranta chain strategy; supply chain reconfiguration;
(2018), Halassi et al. (2019), Luomaranta and responsiveness; remanufacturing; reverse logistics;
Martinsuo (2019), Tziantopoulos et al. (2019), value streams
Zanoni et al. (2019), Delic and Eyers (2020),
Ghobadian et al. (2020), Strong et al. (2020),
Choudhary et al. (2021), Engelseth et al. (2021),
Hecker (2021)
Enabling new Baumers et al. (2015), Prause (2015), Bogers et al. (2016), Business model dynamics; business ecosystems;
business Ford and Despeisse (2016), Rayna and Striukova (2016), capabilities; digital entrepreneurship; disruptive
models Rogers et al. (2016), Steenhuis and Pretorius (2016), innovation; industrial sustainability; nascent
Holzmann et al. (2017), Khorram Niaki and Nonino industries; platforms; SMEs; technology adoption;
(2017), Kwak et al. (2018), Rong et al. (2018), Xu et al. user entrepreneurship; value creation; value
(2018), Holzmann et al. (2020b), Rayna and Striukova capture; value chain reconfiguration
(2021b)

Note: All references shown in this table are provided in Appendix B.

 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
540 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Michelin to develop its CrossClimate/All Season product technologies (Rayna & Striukova, 2021), this also largely
line (Molitch-Hou, 2017). failed to materialize. Despite an early enthusiasm in the
However, this prevents reaping the full benefits of 3D 2010s, consumer 3D printing faces multiple reverse
printing for product innovation. As emphasized explicitly salients—unreliability, relative high cost, post-processing,
(Rayna & Striukova, 2021) or implicitly (Bogers skills and competencies—that have hindered professional
et al., 2016; Ford & Despeisse, 2016; Kleer & Piler, 2019; direct manufacturing applications. Many experts consider
Rindfleisch et al., 2017) in the literature, the benefits only that, aside from small maker (hobbyist) communities
happen when actual products (or a significant part of (Mavri et al., 2021), consumer 3D printing was indeed a
them) are directly manufactured (DM) with 3D printers. fad (Blueprint, 2020; Wohlers, 2020).
This provides digital freedom for physical products: pre- Consumer 3D printing uptake was a key requirement
assembled designs (Haleem et al., 2020), 4D printed prod- for the predicted emergence of transformational con-
ucts (Lee et al., 2017), or products with complex internal sumer communities that would jointly innovate, remix
structure that weight a fraction of usual products (Feng and tinker, share and distribute products and artifacts
et al., 2018; Wohlers, 2020). (Bogers et al., 2016; Jiang et al., 2017; Rayna et al., 2015;
Unfortunately, the reverse salients are also most preva- Rayna & Striukova, 2016; Rayna & Striukova, 2021), as
lent with DM: risk of low replicability, more complex (and has happened with other digital technologies. Such com-
costly) the post-processing (because of the complex struc- munities have emerged, but have had a limited impact,
ture), and requiring new skills (Rylands et al., 2016; since so few people have access to a 3D printer. The
Thomas-Seale et al., 2018). Consequently, DM has majority of 3D printing platforms identified by Rayna
remained confined to a few niche markets—medical/dental et al. (2015) have disappeared or become 3D printing ser-
prosthetics, aerospace and defense (Sculpteo, 2020; vice bureaus (e.g., 3DHub, Sculpteo, i.Materialize). Even
Wohlers, 2020). Other envisaged applications have largely on the popular Thingiverse online community, objects
been abandoned, preventing a key transformation of prod- regularly appear but the top 20 most fabricated ones are
uct innovation predicted by prior research: the real-time more than 5 years old, and contrary to expectations
integration of customer feedback to continuously improve (Stanko & Allen, 2022), have led to few remixes and are
products (Rayna & Striukova, 2016; Rayna & Striukova, seldom remixes themselves.
2021; Walsh et al., 2017; West & Bogers, 2014). Likewise, the seminal RepRap Open-Source Hardware
project that gave rise to MakerBot, Ultimaker, and dozens
of other startups (de Bruijn, 2010; Greul et al., 2018; Hu &
4.2 | Consumer involvement: Sun, 2022; Stanko, 2020; Stanko & Allen, 2022; West &
Customization, co-creation, and Kuk, 2016) has proved unable to move beyond (the rela-
prosumption tively unsophisticated) plastic filament printing technol-
ogy. Limited consumer adoption has dashed widespread
The literature also widely predicted the involvement of hopes firms might leverage 3D printing user communities
consumers and other customers in the production pro- (de Jong & de Bruijn, 2013; Rayna et al., 2015; Rayna &
cess, either in design (mass customization, co-creation) Striukova, 2016; Rindfleisch et al., 2017).
or manufacturing (prosumption, home fabrication). Also unrealized is a predicted shift from e-commerce
Because of higher costs, mass customization via DM to “file-commerce” (products purchased online to be
requires a higher willingness to pay for customized prod- printed at home) and 3D printing that would dramati-
ucts (Franke & Piller, 2004). Yet, actual willingness to cally change how products are developed, distributed,
pay appear to vary greatly (Kalantari et al., 2021), so the and acquired (Jiang et al., 2017). Consequently, the once
reverse salient for DM costs make mass customization predicted IP crisis from consumer piracy of objects
mainly viable in the medical sector (Wohlers, 2020). (Birtchnell et al., 2018; Jiang et al., 2017; Mendis
Mass customization also requires specific skills for et al., 2019; Rayna & Striukova, 2016), has also failed to
both designers and users. After failed efforts to provide materialize.
easy-to-use interfaces, mass customization has generally
been restricted to pseudo-customization (choice within a
range of pre-set attributes) where 3D printing faces 4.3 | Distributed manufacturing and
cost and time disadvantages compared to regular global value chains
manufacturing.
A more radical predicted impact was that consumers Unlike other manufacturing technologies with significant
would manufacture products with their own 3D printers. economies of scale, 3D printing is characterized by a con-
While in line with the success of other digital stant unit cost, which provides far fewer reasons to
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 541

concentrate production geographically or temporally. component are key benefits of 3D printing identified in
Thus, a key impact predicted by the literature is a shift the literature (Beltagui et al., 2020a); a famous example is
toward distributed manufacturing and, consequently, a the claim made by GE to have simplified a jet engine noz-
redefinition of global value chains. zle from 20 parts to one, and, in another case 855 parts to
Digital encapsulation (Holmström et al., 2016) just 12 3D printed parts (GE, 2020). On-demand produc-
through home 3D printing (Despeisse & Minshall, 2017; tion can reduce costs through reduced waste, as well as
Jiang et al., 2017; Rayna & Striukova, 2016) would be the lower assembly, storage and transportation costs, hereby
most extreme form of distributed manufacturing, but is saving fuel (only raw materials transported) and reducing
impractical without widespread consumer 3D printing supply chain and logistics costs (since fewer parts are
uptake. Even RepRap printers, originally designed to be involved) (Ford & Despeisse, 2016).
self-replicable—by having a 3D printer make many of the Literature has hypothesized benefits of on-demand
next printer's parts (Rayna et al., 2023; West & production for sustainability (Ford & Despeisse, 2016;
Kuk, 2016)—are generally purchased as mass-produced Gebler et al., 2014; Rouf et al., 2022), chiefly focusing on
kits or fully assembled printers. spare part production. Yet, the use of 3D printing to pro-
Distributed manufacturing is theoretically possible: duce spare parts remains rate and claims of superiority of
firms could share production facilities and use service 3D printing for spare part manufacturing (Li et al., 2017)
bureaus, while large companies could internally distrib- appear to be limited to specific niches (Frandsen
ute manufacturing across different locations, which et al., 2020) where reverse salients are not as relevant;
would improve sustainability by reducing transport even with production costs four times higher, less than
costs (Ford & Despeisse, 2016; Gebler et al., 2014). Yet, 10% of spare parts are worth 3D printing (Heinen &
we were unable to find in any significant application of Hoberg, 2019). Post-processing further worsens the 3D
this concept, beside the well-publicized (but brief) pro- printing cost disadvantage (Heinen & Hoberg, 2019),
duction of medical equipment during the first months while the environmental improvement over traditional
of the COVID-19 pandemic (Budinoff et al., 2021; manufacturing appears moderate to negligible (Böckin &
Frazer, 2022). Tillman, 2019).
Recent literature suggests that even in best-case scenar- Lack of tooling (Westerweel et al., 2018) is predicted
ios, the alleged superiority of distributed manufacturing to enable shorter lead-times, but this has little impact
may not stand scrutiny. Li et al. (2019) found that, contrary overall (Knofius et al., 2019), because of the reverse
to the existing perception, distributed manufacturing does salient (comparatively slow speed of and higher
not always guarantee a quick response, and that pooling manufacturing cost) makes that, unless unique parts are
effect makes that centralized manufacturing performs bet- involved. Otherwise, 3D printing's lead-time advantage
ter when the demand rate is relatively high. Furthermore, rapidly wears off (Muir & Haddud, 2017).
Rinaldi et al. (2022) found the benefits of distributed In regard to logistics and supply chains, 3D printing's
manufacturing rather limited even when average demand ability to manufacture shapes that embed several parts
is rather low, as centralized manufacturing provides a bet- into one (consolidation) has also been emphasized. Yet,
ter capacity utilization. this often leads to a higher total cost (Knofius et al.,
This contrast between predictions and the reality can 2019)—consolidation worsens the impacts of wear or fail-
be largely attributed to the reverse salients. When physi- ure, while greater design complexity means higher
cal properties (structural or aesthetical) of a product post-processing costs (Knofius et al., 2019; Sgarbossa
matter, distributed manufacturing requires a very tight et al., 2021).
control of the manufacturing process, which makes 3D Overall, the benefits of 3D printing on supply chain
printing a QA nightmare. In a world where transport and and logistics appear chiefly limited to niches where the
logistics are rapid, efficient, and cheap, the (mainly) cen- reverse salients do not matter as much: inherently low
tralized manufacturing model is hard to beat. The physi- volumes, no post-processing, and parts that do not wear
cal limitations of 3D printing stand in the way of the out, for example, highly specialized low-volume applica-
long-predicted global value chain reconfiguration. tions, such as industrial equipment, manufacturing
at sea, or remote supply depots for combat ships
(Kenney, 2013)—or in space, as demonstrated in 2014 by
4.4 | Logistics and supply chains NASA using an FDM 3D printer to produce spare parts
on the International Space Station (Prater et al., 2016), in
On-demand production and ability of manufacture com- which case the cost of getting cargo to outer space out-
plex designs to consolidate several parts into a single weigh the limitations of 3D printing.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
542 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

4.5 | Enabling business model 4.6 | Reevaluating societal impacts of 3D

innovation printing

Beside investigating business models for the 3D printing The literature shows a prevalent belief that 3D printing
industry (e.g., Holzmann et al., 2020), literature has pro- represents the next stage of digital innovation, blurring the
jected that 3D printing would radically reconfigure busi- boundaries between digital and physical, hereby opening
ness models in physical sectors, just like what has an entire new realm of innovation (Aryan et al., 2021;
happened in the digital realm, whether by choice or Nasiri et al., 2020; Rindfleisch et al., 2017). This brought
necessity due to consumer piracy of objects (Appleyard, predictions that the physical world would witness a free-
2015; Birtchnell et al., 2018; Depoorter, 2014; Mendis dom of creation similar to the digital one, enabling new
et al., 2019). 3D printing would not only change business modes of digital innovation that would “digitize physical
models—leading to more consumer-centric business goods much like the first digital revolution digitized infor-
models involving consumers in the manufacturing supply mation goods” (Rindfleisch et al., 2017, p. 681) and trans-
chains (Bogers et al., 2016), with, over time, a prolifera- form product development and innovation, value chain,
tion of “business models that involve increased levels inventory management, customization and business model
of customer self-service” (Bogers et al., 2016, p. 12), innovation.
but business model innovation in itself (Rayna & Aside from very specific niches, the transformative
Striukova, 2016). impacts of 3D printing identified in the literature have
Yet, beside occasional business model upgrades, it is largely failed to materialize, because of the physical limi-
hard to find cases of radical business model reconfigura- tations of 3D printing that act as reverse salients. These
tion through 3D printing. The RepRap Open-Source physical reverse salients have prevented a widespread
Hardware project has enabled shifts between open adoption of 3D printing by firms and individuals—a pre-
and closed business models (Stanko, 2020; West & requisite to most of transformative impact predictions, as
Kuk, 2016), as well as business models blending these it would enable firms to gather early customer product
two modes (Appleyard & Chesbrough, 2017; Rayna et al., feedback (West & Bogers, 2014), harvest data through 3D
2023). Even in this case, only startups have carried out printing to better understand user needs (Rindfleisch
bold business model innovation (Greul et al., 2018), with et al., 2017) and leverage users and user communities to
incumbent perpetuating essentially the same business innovate derivatively through 3D printing (de Jong & de
model. Bruijn, 2013; Rindfleisch et al., 2017). Such adoption
Expectations of transformative impact on business would lead to a potential large-scale shift to online reposi-
models were largely based on the premise of a large cus- tories for product design (Jiang et al., 2017) and bypass
tomer adoption, and involved mass customization, dis- traditional manufacturing via digital encapsulation
tributed manufacturing and reconfigured global supply (Holmström et al., 2016). In turn, this would force rethink-
chains, or on-demand production. Because of the ing of the role of brands (Rindfleisch & O'Hern, 2015;
reverse salients, all have been confined to narrow Wang et al., 2016). Widespread consumer adoption would
niches. It is not hard to understand why the predicted lead to the rise of intermediaries and platforms that would,
radical business model shifts have not materialized. The on the positive side, facilitate innovation and, on the nega-
related literature generally ignores the reverse salients tive side, bring a predicted IP doomsday when IP protec-
and instead emphasizes the benefits brought about by tion of physical goods would suddenly become as
the digital side of 3D printing, with the implicit assump- challenging as for other digital goods (Jiang et al., 2017).
tion that 3D printed artifacts behave just like purely dig- As industry experts now admit (Wohlers, 2020), wide-
ital artifacts. spread adoption is unlikely to arise until reverse salients
Even in cases when 3D printing was largely adopted, have been resolved: with decades of hindsight, predic-
business models may largely remain the same. The main- tions and expectations were simply unrealistic. Conse-
stream uptake of 3D printing in the hearing aid industry quently, discussions surrounding what is the correct
in the 1990s has still not brought radical business model business model for products marketed through 3D
reconfiguration (Sandström, 2015). Despite widespread printing—such as pricing per print, subscription or open
adoption, the very same networks of firms are involved access model (Bogers et al., 2016)—remain largely
in the value creation, with no radical shift in value net- irrelevant. Even predictions that did not require a large
works, such as consumer involvement or other changes consumer adoption have largely remained niche applica-
described in Bogers et al. (2016) and Rayna and Striukova tions, such as 3D printing mass customized medical
(2016). implants (Randhawa et al., 2020).
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 543

The same reverse salients have also impeded the pre- materialize due to persistent limitations of 3D printing
dicted expansion of distributed manufacturing, and the performance. The traditional response in prior research
advent of dispersed 3D printing micro-production centers has been that those limitations would soon be solved:
located close to customers (Luchs et al., 2015), which, in
turn, would have distributed innovation beyond a single By [comparing] direct digital manufacturing
firm. Consequently, questions of best practices for collabo- with traditional tool-based manufacturing, we
rative innovation with distributed production—particularly show that direct digital manufacturing so far
the role of digital manufacturing in B2B collaboration lags behind by several orders of magnitude
between established firms and startups and SMEs, also have compared to traditional manufacturing
been pushed into the future, except for niches such as methods. Yet we also find that direct digital
defense, aerospace and industrial equipment. manufacturing clearly is on an improvement
This consistent failure to meet expectations has led to a trajectory that eventually will see it being able
reevaluation some of the most optimistic predictions of 3D to compete with traditional manufacturing on
printing use. This can be seen in the content of the influen- a unit cost basis (Holmström et al., 2016, p. 1).
tial Wohlers Report—a yearly report by industry experts
and key source of information about the 3D printing As we have shown, the manufacturing (i.e., physical)
industry. Between the Wohlers (2012) and Wohlers (2020) aspects of 3D printing have been the reverse salient for its
annual reports, most (9 of 17) of the predicted of 3D print- use and adoption for three decades: while closing this “sev-
ing applications—avatars and figurines; consumer-created eral orders of magnitude” gap has long been a priority, on
products; fashion products; furniture, home, and office the current trajectory the eventual ability to compete with
accessories; gifts, trophies, and memorials; marketing and traditional manufacturing remains decades away.
advertising; museum displays; musical instruments; and For future research, refractory reverse salients of 3D
personal 3D printers—were dropped. Applications com- printing require a new research approach. Rather than
mon across both periods are essentially those already pre- assume an imminent (but still unfulfilled) “hockey-stick”
dicted 25 years ago: aerospace, architectural modeling, art inflection in performance or adoption, researchers should
and jewelry, automotive, machinery, medical applications, instead focus on what is known to be technologically
prototyping, tooling. feasible. By doing so, they will produce more realistic
Industry experts appear to have finally realized that (and relevant) analyses that considers the impact of these
the physical reverse salients of 3D printing were not tem- limitations, and how persistent reverse salients can be cir-
porary impediments, but were instead intrinsic to 3D cumvented with more attainable technological solutions.
printing: building objects layer by layer usually requires To ground research in reality, two different paths lie
repeated but unique reactions at molecular level, leading ahead for 3D printing researchers. The first requires revi-
to complexity, unpredictability, comparatively high siting 3D printing predictions for each of the six themes
energy usage and slow speed. In fact, it can be argued of Table 3, and, accounting for the actual limitations,
that the very limitations of 3D printing lie in its very investigate which predictions remain realistic. In particu-
strength. As a result, these reverse salients are unlikely to lar, this means moving away from the focus on direct
be resolved any time soon. manufacturing, which will remain a niche use.
Mainstream uses of 3D printing remain prototyping
and tooling, so an immediate research opportunity is
5 | FUTURE RESEARCH how product design best practices can leverage both. This
is particularly important at a time when virtual prototyp-
Given the enduring reverse salients of 3D printing, how ing technologies—such as augmented or virtual reality—
does that change future research? Here we consider the are rapidly increasing in performance and use. These
implications of this for the future of digital manufactur- newer technologies may undercut prior 3D printing suc-
ing research, and more broadly for our understanding of cesses, as digital substitutes evolve more quickly than
digital innovation, CPSs, and reverse salients, concluding physical-limited processes. Likewise, fast moving high-
with managerial implications of this research. flexibility robotic automation may reduce the appeal of
rapid tooling.
A second research path would study how other digital
5.1 | Implications for 3D printing and technologies might help overcome the physical limita-
digital manufacturing tions of 3D printing. For instance, combining 3D printing
with other digital manufacturing technologies (e.g., CNC,
As analyzed in Section 4, research on 3D printing has laser cutting) could alleviate the speed and post-
brought high expectations that have repeatedly failed to processing issues that plague 3D printing, while using AI
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
544 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

in design may diminish the complexity and time needed • Business model innovation: Besides direct manufactur-
to customize 3D printed items. Augmented reality could ing and home fabrication, could rapid prototyping and
finally provide the intuitive and easy-to-use user interface tooling enable reconfigured or entirely new business
(equivalent the mouse and windows for PCs) that is models? Are there disruptive (Christensen, 1997) busi-
needed to fully leverage 3D printing technologies. ness models for direct or local manufacturing
Overall, both research directions are examples of how analogous to how lower quality MP3 music disrupted
to reduce the impact of persistent reverse salients on music consumption in the 1990s?
improving the performance and economic impact of 3D
printing. Future research and commercial applications
cannot ignore these reverse salients, but instead must 5.2 | Implications for CPSs and digital
consider those trajectories that either minimize or over- innovation
come their impact.
Considering the main themes identified in the litera- Considering the differences of the digital and physical
ture (Table 3), opportunities for further research include: models suggests that we need a more nuanced model of
innovation and technological change in today's digital
• Product innovation: Assuming direct manufacturing world, particularly with regard to reverse salients. In
will remain a niche application, how can 3D printed Table 4, we compare the rate of innovation and reverse
tooling enable complex molds and gigs that can be salients for all possible combinations of digital and physi-
used as a source of product innovation? How can com- cal innovations: the two extremes of purely digital or
bining 3D printing with virtual prototyping technolo- physical systems, and four possible configurations of
gies be leveraged to facilitate user feedback in near CPSs. We contrast the emerging technologies (where dra-
real-time? matic performance improvements are still possible) with
• Consumer involvement: If consumer adoption of 3D mature technologies that have approached the limits of
printing remains low, how can combining these tech- their technological trajectory.
nologies enable large-scale co-creation with con- The Purely Digital configuration corresponds to the
sumers? Could digital technologies supplant 3D digital innovation research of the past decade. In most
printed physical prototypes, thereby enabling online cases, the malleability and rapidly improving perfor-
communities to utilize digital twins that provide a true mance of digital systems have enabled dramatic progress
correspondence between the digital artifacts and their in such systems. A notable exception is artificial intelli-
physical materialization? Could mass-customization gence, where the rate of progress has been “patchy and
serve intermediate-sized global markets with similar unpredictable” (Stone et al., 2016, p. 4).
tastes, desires, or needs with 3D printing be used to Among CPSs, we identify four configurations.
decrease tooling cost to make economical micro-batch (1) Digital-Limited have mature physical technologies,
production? but still require dramatic improvements in the digital
• Distributed manufacturing and global value chains: technologies. (2) Emerging Digital and Physical, where
Assuming the limitations of direct manufacturing can- both modes are significant limiting factors, but neither
not be overcome, to which extent other digital technology has reached maturity. (3) Mature Digital and
manufacturing technologies (such as laser cutting or Physical technologies that are combined for an immedi-
CNC) could be combined with 3D printing in inte- ate wave of combinatorial (often digital) innovation
grated and universal manufacturing units in order to (Sandberg et al., 2020). However, once the combination
deliver the promise of distributed manufacturing, and, reaches maturity, subsequent improvements are slow and
consequently, enable more sustainable global value incremental. (4) Physical-Limited, where the rate of digi-
chain reconfigurations? tal innovation is much faster than the physical and thus
• Logistics and supply chains: Could integrated digital overall performance is limited by the physical.
manufacturing units be leveraged to deliver pseudo Finally, the Purely Physical systems are the familiar
on-demand manufacturing, possibly with the help of domain of innovation studies of the past 40 years—such as
rapid tooling enabling to produce just-in-time small railroads, power systems airplanes, automobiles and
batches production using traditional manufacturing? manufacturing processes (Anderson & Tushman, 1990;
Likewise, could rapid tooling meet very tight lead Hughes, 1983; Utterback, 1994). Today, most mechanical
times by utilizing local integrated digital manufactur- innovations have some digital component, but purely phys-
ing units? Also could hybrid production strategies ical innovation can be found in materials science—such as
manufacture parts less prone to wearing out through photovoltaic cells and modern battery materials—where
3D printing, while consumable parts remain mass the rate of progress remains slow despite billions of dollars
produced? in annual R&D spending.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 545

TABLE 4 Impact of reverse salients in digital and physical innovation.

Reverse Rate of
Domain Digital Physical salient innovation Example
Purely digital Emerging N/a Digital Typically high Artificial intelligence (Haenlein &
Kaplan, 2019)
CPS (digital-limited) Emerging Mature Digital Typically high Facial recognition, edge computing, IoT
CPS (digital and physical Emergent Emergent Both Slow Virtual reality (Zhan et al., 2020),
limited) autonomous vehicles (Bagloee
et al., 2016)
CPS (mature digital and Mature Mature Both Initially high, CNC (Ernst, 1997), Internet-enabled
physical) then slow appliances (Boutellier & Heinzen, 2014)
CPS (physical-limited) Mature Emerging Physical Slow 3D printers, wind turbines (Beiter
et al., 2021; Veers et al., 2019)
Purely physical N/a Emerging Physical Slow Photovoltaic cells, battery materials (Chu
et al., 2017)

Research on CPS innovation so far has emphasized the digital constraints on system performance, what is
digital-limited case, but future research should examine needed is a more nuanced discussion of the interaction of
CPSs that are physical-limited. Since CPS performance can this digital innovation with the inevitable physical
be limited by either digital or physical components, if constraints.
researchers overestimate the rate of improvement for the Additional research should examine other hybrid
physical components, this leads to incorrect predictions of CPSs (beyond digital manufacturing) where the perfor-
the innovation trajectory of a CPS. Future research is also mance drivers are more physical than digital. Such stud-
needed on the interdependence of component innovation ies should consider two possible technological
for CPS and their respective innovation rates and techno- trajectories. The first is when physical components limit
logical bottlenecks. the rate of innovation for digital components or the over-
At the same time, theories of digital innovation (and all system. The second is when improvements in digital
innovation more generally) have rarely considered what components work around the limitations of slowly
happens when the digital meets the physical. Thus improving physical ones, as appears to be happening
research on digital innovation should consider the rarely with self-driving cars and virtual reality displays. Theo-
explored middle ground of Table 4, between traditional ries and measures for reverse salients could help explain
(physical) and digital innovation theories. The case of 3D both technological trajectory. Finally, future research
printing reminds us that in the end, digital innovation should examine reverse salients where progress depends
must serve material beings living in a material world. on a complex interplay of digital and physical innovation;
Cyber-physical and other systems that span both domains such studies could link to the rich literature on
inherently face material constraints not applicable to co-evolutionary innovation processes (e.g., Garud &
purely digital innovations, but even the latter will eventu- Rappa, 1994; Geels, 2005; van den Ende et al., 2012).
ally face material constraints when they interface with
the material world.
Future research on digital innovation is thus at a 5.3 | Implications for reverse salients
crossroads as to how to address this tangible reality. The
easiest way would be to define the scope of the theory While Hughes' concept of a reverse salient continues to
and its predictions as bounded by purely digital influence innovation research (Kapoor & McGrath, 2014;
innovations—emphasizing the boundaries of previous Lenfle & Söderlund, 2022; Mäkitie et al., 2022; Mathews
theoretical findings and perhaps the crossover points to et al., 2018; Sandberg et al., 2020; Schubert et al., 2013),
adjacent theories. The path less traveled would be to that research continues to have a singular blind spot, by
adapt and extend theories of digital innovation to explain assuming reverse salients can and will be overcome. While
digitally-enabled physical innovation, as we have done innovation studies examine successful technologies—or
with 3D printing and others have done with automobiles occasionally failed projects within a successful technology
(Lee & Berente, 2012). While previous studies of digital- (Law & Callon, 1992)—few reverse salients studies have
physical (and CPS) systems studied have emphasized examined when the salient is not overcome.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
546 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

This could be just a methodological issue: Hughes predictions. These suggest multiple opportunities to study
(1983) did not publish his study of electric power in 1890, persistent reverse salients that shape technological trajec-
just as West (2008) did not publish his study of deep tories, problem recognition and commercialization over
space communications in 1965 and Lenfle and Söderlund decades.
(2022) did not publish their study of numerical weather
prediction in 1990. In 2000, researchers would have con-
cluded photovoltaic solar energy equipment was a tech- 5.4 | Managerial implications
nology that failed to meet its goals for decades
(West, 2014), but today such equipment generates more As Hughes emphasized, reverse salients can provide great
than $100 billion in global revenue and a double-digit market opportunities. However, 3D printing points to
compound growth rate. what happens when the time horizon for resolving them
Future research should thus consider this fundamental is beyond the time horizon of private managers. While
question: What happens to a technological trajectory when utopian predictions sell analyst reports, consulting ser-
salients are not overcome? As noted earlier, 3D printing vices and stock offerings, the longstanding salients in 3D
suggests that two possible options are to target market printing performance provide exemplars of how man-
niches or to compensate by innovating in other compo- agers can identify inherent (and difficult to resolve) tech-
nents of the system. But in both cases, the salient remains. nological limits. Although rarely identified in prior
How might this be studied? While Coopersmith research, multiple simultaneous reverse salients make it
(2020) suggests that one explanation is the lack of data even less likely that the technology will make immediate
for failed technologies, a well-documented example of and rapid progress.
utopian predictions that are unrealized due to persistent At the same time, not all hybrid digital-physical sys-
reverse salients would include artificial intelligence. tems are the same. We offer a framework for identifying
Beginning with the field's origins at a 1956 Dartmouth which systems are most capable of rapid progress and
Summer conference that included Claude Shannon, which ones are not. At the same time, digital innovation
Herb Simon and Marvin Minksy, AI missed key predic- promises rapid and malleable innovation that can deliver
tions for decades. For example, in 1973 AI experts pre- rapid initial improvements in CPSs until the physical lim-
dicted (Nilson, 2009, p. 515) the commercial availability itations are reached. Together, these suggest caution for
of automatic medical diagnosis (in 1980), automated managers investing in CPS innovation where physical
voice-to-text (in 1992) and self-driving vehicles for city innovation is required to realize the hypothesized benefit
streets (in 2000) but the actual wait was at least 50% of that system.
longer. For companies involved in 3D printing, whether as
Empirical research on reverse salients could thus providers or customers, a key implication is to take the
develop a more complete framework for the co-evolution physical limitations of 3D printing as they are, without
of salients, critical problems and technological trajecto- expecting that they will eventually disappear. Instead of
ries. Prior research has noted the difference between considering uses where 3D printing would have a major
salients that are quickly recognized and those that are impact if limitations were eliminated, they should instead
not. Future research could consider both salients that focus on uses where 3D printing already can make a dif-
take decades to resolve and those not yet resolved as of ference. Such analysis should leverage convergence with
the current writing. Even among successful innovations, other digital technologies, whether on the digital
such decades-long trajectories are more common than (AI, AR) or the physical side (robotics, CNC, IoT) to fully
often assumed: The 1987 Knowledge Navigator video that unleash the potential of 3D printing as it now is, not as
made Apple CEO John Sculley the butt of endless jokes we dream it could be.
was largely realized 24 years later when Apple added the
Siri voice assistant to the iPhone. ACKNOWLEDGMENTS
Another complex systemic innovation that has been We are grateful to the reviewers and guest editor Michael
the subject of several salient studies is the transition to Stanko for their guidance and support throughout the
renewable energy, as when Mäkitie et al. (2022) exam- development of this article; we particularly appreciate
ined how reverse salients in three RE transportation tech- the reviewer who suggested a possible connection to the
nologies influenced investment decisions. Beyond AI and influential work of Thomas Hughes.
RE, data are readily available on reverse salients for other
key technologies such as fusion power production, F U N D I N G IN F O R M A T I O N
humanoid robots or interplanetary exploration; more This research was in part funded through the Chair Tech-
recently, small-scale nuclear power plants and flying tax- nology for Change of the Institut Polytechnique de Paris,
ies are growing closer to realizing decades-long utopian as supported by the patronage of Accenture.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 547

CONFLICT OF INTEREST STATEMENT Beiter, P., A. Cooperman, E. Lantz, T. Stehly, M. Shields, R. Wiser,
The authors declare no conflicts of interest. T. Telsnig, L. Kitzing, V. Berkhout, and Y. Kikuchi. 2021.
“Wind Power Costs Driven by Innovation and Experience With
Further Reductions on the Horizon.” WIREs Energy and Envi-
E TH IC S ST A T EME N T
ronment 10(5): e398. Portico. https://doi.org/10.1002/wene.398.
The authors have read and agreed to the Committee on Ben‐Ner, A., and E. Siemsen. 2017. “Decentralization and Localiza-
Publication Ethics (COPE) international standards for tion of Production.” California Management Review 59(2): 5–23.
authors. https://doi.org/10.1177/0008125617695284.
Bijker, W. E., and J. Law, eds. 1994. Shaping Technology/Building
ORCID Society: Studies in Sociotechnical Change. Cambridge, MA: MIT
Thierry Rayna https://orcid.org/0000-0003-0144-6031 Press.
Birtchnell, T., A. Daly, T. Rayna, and L. Striukova. 2018. 3D Print-
Joel West https://orcid.org/0000-0002-7958-5229
ing and Intellectual Property Futures (Independent Report No.
2018/03). Newport, UK: UK Intellectual Property Office.
R EF E RE N C E S Blakey-Milner, B., P. Gradl, G. Snedden, M. Brooks, J. Pitot, E.
Achillas, C., D. Aidonis, E. Iakovou, M. Thymianidis, and D. Lopez, M. Leary, F. Berto, and A. du Plessis. 2021. “Metal
Tzetzis. 2015. “A Methodological Framework for the Inclusion Additive Manufacturing in Aerospace: A Review.” Materials &
of Modern Additive Manufacturing into the Production Portfo- Design 209: 110008. https://doi.org/10.1016/j.matdes.2021.
lio of a Focused Factory.” Journal of Manufacturing Systems 37: 110008.
328–39. https://doi.org/10.1016/j.jmsy.2014.07.014. Blueprint. 2020. “CES 2020 Shows Us What the End of the Consumer
Appleyard, M. M., and H. W. Chesbrough. 2017. “The Dynamics of 3D Printing Hype Cycle Looks Like.” 3DPrint.ComjThe Voice of
Open Strategy: From Adoption to Reversion.” Long Range Plan- 3D Printing/Additive Manufacturing. January 24, 2020. https://
ning 50(3): 310–21. https://doi.org/10.1016/j.lrp.2016.07.004. 3dprint.com/262622/ces-2020-shows-us-what-the-end-of-the-
Aryan, V., J. Bertling, and C. Liedtke. 2021. “Topology, Typology, consumer-3d-printing-hype-cycle-looks-like/.
and Dynamics of Commons-Based Peer Production: On Plat- Bogers, M., R. Hadar, and A. Bilberg. 2016. “Additive Manufactur-
forms, Actors, and Innovation in the Maker Movement.” Crea- ing for Consumer-Centric Business Models: Implications for
tivity and Innovation Management 30(1): 63–79. https://doi.org/ Supply Chains in Consumer Goods Manufacturing.” Technolog-
10.1111/caim.12392. ical Forecasting and Social Change 102: 225–39. https://doi.org/
Åstebro, T. 2002. “Noncapital Investment Costs and the Adoption 10.1016/j.techfore.2015.07.024.
of CAD and CNC in U.S. Metalworking Industries.” RAND Bogers, M., and W. Horst. 2014. “Collaborative Prototyping:
Journal of Economics 33(4): 672–88. https://doi.org/10.2307/ Cross-Fertilization of Knowledge in Prototype-Driven Problem
3087480. Solving.” Journal of Product Innovation Management 31(4):
Autio, E., R. Mudambi, and Y. Yoo. 2021. “Digitalization and Glob- 744–64. https://doi.org/10.1111/jpim.12121.
alization in a Turbulent World: Centrifugal and Centripetal Boland, R. J., K. Lyytinen, and Y. Yoo. 2007. “Wakes of Innovation
Forces.” Global Strategy Journal 11(1): 3–16. https://doi.org/10. in Project Networks: The Case of Digital 3‐D Representations
1002/gsj.1396. in Architecture, Engineering, and Construction.” Organization
Avinson. 2018. “How Long Does 3D Printing Take?” September Science 18(4): 631–47. https://doi.org/10.1287/orsc.1070.0304.
20, 2018. https://3space.com/blog/how-long-does-3d-printing- Borgia, E. 2014. “The Internet of Things Vision: Key Features,
take/. Applications and Open Issues.” Computer Communications 54:
Böckin, D., and A.-M. Tillman. 2019. “Environmental Assessment 1–31. https://doi.org/10.1016/j.comcom.2014.09.008.
of Additive Manufacturing in the Automotive Industry.” Jour- Boutellier, R., and M. Heinzen. 2014. Growth Through Innovation:
nal of Cleaner Production 226: 977–87. https://doi.org/10.1016/ Managing the Technology-Driven Enterprise. Cham, Switzer-
j.jclepro.2019.04.086. land: Springer Science & Business Media.
Baars, H., C. Felden, P. Gluchowski, A. Hilbert, H.-G. Kemper, and Bowles, M. D., and V. P. Dawson. 1998. “The Advanced Turboprop
S. Olbrich. 2014. “Shaping the Next Incarnation of Business Project: Radical Innovation in a Conservative Environment.” In
Intelligence.” Business & Information Systems Engineering 6(1): From Engineering Science to Big Science—The NACA and NASA
11–6. https://doi.org/10.1007/s12599-013-0307-z. Collier Trophy Research Project Winners, edited by Pamela Etter
Bagloee, S. A., M. Tavana, M. Asadi, and T. Oliver. 2016. “Autono- Mack, 321–43. Honolulu: University Press of the Pacific.
mous Vehicles: Challenges, Opportunities, and Future Implica- Budinoff, H. D., J. Bushra, and M. Shafae. 2021. “Community-
tions for Transportation Policies.” Journal of Modern Driven PPE Production Using Additive Manufacturing During
Transportation 24(4): 284–303. https://doi.org/10.1007/s40534- the COVID-19 Pandemic: Survey and Lessons Learned.” Jour-
016-0117-3. nal of Manufacturing Systems 60: 799–810. https://doi.org/10.
Baldwin, C. Y. 2015. “Bottlenecks, Modules, and Dynamic Architec- 1016/j.jmsy.2021.07.010.
tural Capabilities.” In The Oxford Handbook of Dynamic Capa- Campbell, T., C. Williams, O. Ivanova, and B. Garrett. 2011. Could
bilities, Online ed., edited by David J. Teece and Sohvi Heaton. 3D Printing Change the World? Washington, DC: Atlantic
Oxford: Oxford University Press. https://doi.org/10.1093/ Council.
oxfordhb/9780199678914.013.011. Chandler, A. D. 1990. Scale and Scope: The Dynamics of Industrial
Baumers, M., and M. Holweg. 2019. “On the Economics of Additive Capitalism. Cambridge, MA: Belknap Press.
Manufacturing: Experimental Findings.” Journal of Operations Chaudhuri, A., H. A. Gerlich, J. Jayaram, A. Ghadge, J. Shack,
Management 65(8): 794–809. https://doi.org/10.1002/joom. B. H. Brix, L. H. Hoffbeck, and N. Ulriksen. 2021. “Selecting
1053. Spare Parts Suitable for Additive Manufacturing: A Design
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
548 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Science Approach.” Production Planning & Control 32(8): 670– Dedehayir, O., and S. J. Mäkineif. 2008. “Dynamics of Reverse
87. https://doi.org/10.1080/09537287.2020.1751890. Salience as Technological Performance Gap: An Empirical
Christiansen, A. C., and J. Buen. 2002. “Managing Environmental Study of the Personal Computer Technology System.” Journal
Innovation in the Energy Sector: The Case of Photovoltaic and of Technology Management & Innovation 3(3): 55–66. https://
Wave Power Development in Norway.” International Journal of doi.org/10.4067/S0718-27242008000100006.
Innovation Management 6(3): 233–56. https://doi.org/10.1142/ Dedehayir, O., and S. J. Mäkinen. 2011b. “Determining Reverse
s1363919602000586. Salient Types and Evolutionary Dynamics of Technology Sys-
Chu, S., Y. Cui, and N. Liu. 2017. “The Path Towards Sustainable tems With Performance Disparities.” Technology Analysis &
Energy.” Nature Materials 16(1): 16–22. https://doi.org/10. Strategic Management 23(10): 1095–114. https://doi.org/10.
1038/nmat4834. 1080/09537325.2011.621308.
Claussen, J., and M. A. Halbinger. 2021. “The Role of Pre- Depoorter, B. 2014. “Intellectual Property Infringements & 3D Printing:
Innovation Platform Activity for Diffusion Success: Evidence Decentralized Piracy.” Hastings Law Journal 65(6): 1483–504.
From Consumer Innovations on a 3D Printing Platform.” https://repository.uclawsf.edu/hastings_law_journal/vol65/iss6/2.
Research Policy 50(8): 103943. https://doi.org/10.1016/j.respol. Despeisse, M., and T. Minshall. 2017. “Skills and Education for
2020.103943. Additive Manufacturing: A Review of Emerging Issues.” In
Conkle, J., P. S. Suchdev, E. Alexander, R. Flores-Ayala, Advances in Production Management Systems. The Path to Intel-
U. Ramakrishnan, and R. Martorell. 2018. “Accuracy and Reli- ligent, Collaborative and Sustainable Manufacturing, edited by
ability of a Low-Cost, Handheld 3D Imaging System for Child H. Lödding, R. Riedel, K.-D. Thoben, G. von Cieminski, and D.
Anthropometry.” PLoS One 13(10): e0205320. https://doi.org/ Kiritsis, 289–97. Cham, Switzerland: Springer International
10.1371/journal.pone.0205320. Publishing. https://doi.org/10.1007/978-3-319-66923-6_34.
Cooper, R. G. 2021. “Accelerating Innovation: Some Lessons From Dosi, G. 1982. “Technological Paradigms and Technological Trajec-
the Pandemic.” Journal of Product Innovation Management tories.” Research Policy 11(3): 147–62. https://doi.org/10.1016/
38(2): 221–32. https://doi.org/10.1111/jpim.12565. 0048-7333(82)90016-6.
Coopersmith, J. 2020. “Spectacular Flops: Game‐Changing Technol- Dowling, L., J. Kennedy, S. O'Shaughnessy, and D. Trimble. 2020.
ogies That Failed by Michael Brian Schiffer.” Technology and “A Review of Critical Repeatability and Reproducibility Issues
Culture 61(4): 1235–6. https://doi.org/10.1353/tech.2020.0134. in Powder Bed Fusion.” Materials & Design 186: 108346.
Coutts, E. R., A. Wodehouse, and J. Robertson. 2019. “A Compari- https://doi.org/10.1016/j.matdes.2019.108346.
son of Contemporary Prototyping Methods.” Proceedings of the Ernst, H. 1997. Small Business Economics 9(4): 361–81. https://doi.
Design Society: International Conference on Engineering Design org/10.1023/a:1007921808138.
1(1): 1313–22. https://doi.org/10.1017/dsi.2019.137. Eyers, D. R., and A. T. Potter. 2017. “Industrial Additive
Daim, T., J. Justice, L. Hogaboam, S. J. Mäkinen, and O. Dedehayir. Manufacturing: A Manufacturing Systems Perspective.” Com-
2014. “Identifying and Forecasting the Reverse Salient in Video puters in Industry 92–93: 208–18. https://doi.org/10.1016/j.
Game Consoles: A Performance Gap Ratio Comparative Analy- compind.2017.08.002.
sis.” Technological Forecasting and Social Change 82: 177–89. Feng, J., J. Fu, Z. Lin, C. Shang, and B. Li. 2018. “A Review of the
https://doi.org/10.1016/j.techfore.2013.06.007. Design Methods of Complex Topology Structures for 3D Print-
Davies, A. 1996. “Innovation in Large Technical Systems: The Case ing.” Visual Computing for Industry, Biomedicine, and Art 1(1):
of Telecommunications.” Industrial and Corporate Change 5(4): 5. https://doi.org/10.1186/s42492-018-0004-3.
1143–80. https://doi.org/10.1093/icc/5.4.1143. Fontana, F., C. Klahn, and M. Meboldt. 2019. “Value-Driven Clus-
Davies, A., and T. Brady. 2000. “Organisational Capabilities and tering of Industrial Additive Manufacturing Applications.”
Learning in Complex Product Systems: Towards Repeatable Journal of Manufacturing Technology Management 30(2): 366–
Solutions.” Research Policy 29(7–8): 931–53. https://doi.org/10. 90. https://doi.org/10.1108/JMTM-06-2018-0167.
1016/s0048-7333(00)00113-x. Ford, S., and M. Despeisse. 2016. “Additive Manufacturing and Sus-
de Bruijn, E. 2010. On the Viability of the Open Source Development tainability: An Exploratory Study of the Advantages and Chal-
Model for the Design of Physical Objects: Lessons Learned From lenges.” Journal of Cleaner Production 137: 1573–87. https://
the RepRap Project. Unpublished MSc Thesis. Tilburg: Tilburg doi.org/10.1016/j.jclepro.2016.04.150.
University. Frandsen, C. S., M. M. Nielsen, A. Chaudhuri, J. Jayaram, and K.
de Jong, J. P., and E. de Bruijn. 2013. “Innovation Lessons From Govindan. 2020. “In Search for Classification and
3-D Printing.” MIT Sloan Management Review 54(2): 43–52. Selection of Spare Parts Suitable for Additive Manufacturing:
Dedehayir, O., and S. J. Mäkinen. 2011a. “Measuring Industry Clock- A Literature Review.” International Journal of Production
speed in the Systemic Industry Context.” Technovation 31(12): Research 58(4): 970–96. https://doi.org/10.1080/00207543.
627–37. https://doi.org/10.1016/j.technovation.2011.07.008. 2019.1605226.
Dedehayir, O. 2009. “Bibliometric Study of the Reverse Salient Con- Franke, N., and F. Piller. 2004. “Value Creation by Toolkits for User
cept.” Journal of Industrial Engineering and Management 2(3): Innovation and Design: The Case of the Watch Market.” Jour-
3. https://doi.org/10.3926/jiem.v2n3.p569-591. nal of Product Innovation Management 21(6): 401–15. https://
Dedehayir, O., and A. Hornsby. 2008. “Staying at the Technology doi.org/10.1111/j.0737-6782.2004.00094.x.
Forefront by Spotting the Reverse Salient: The Case of Digital Frazer, J. S. 2022. “The Role of Distributed Manufacturing and 3D
Video Broadcasting.” In 2008 4th IEEE International Conference Printing in Development of Personal Protective Equipment
on Management of Innovation and Technology 189–94. https:// Against COVID-19.” In Emerging Applications of 3D Printing
doi.org/10.1109/icmit.2008.4654360. During CoVID 19 Pandemic, edited by K. Sandhu, S. Singh, C.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 549

Prakash, N. R. Sharma, and K. Subburaj, 15–34. Singapore: Operations Management 65(8): 810–26. https://doi.org/10.1002/
Springer. https://doi.org/10.1007/978-981-33-6703-6_2. joom.1054.
Fujita, M. 2001. “AIBO: Toward the Era of Digital Creatures.” The Hendler, S. 2019. “Digital‐Physical Product Development: A
International Journal of Robotics Research 20(10): 781–94. Qualitative Analysis.” European Journal of Innovation Man-
https://doi.org/10.1177/02783640122068092. agement 22(2): 315–34. https://doi.org/10.1108/ejim-01-2018-
Garcia, R., and R. Calantone. 2002. “A Critical Look at Technologi- 0026.
cal Innovation Typology and Innovativeness Terminology: A Herterich, M. M., T. Buehnen, F. Uebernickel, and W. Brenner.
Literature Review.” Journal of Product Innovation Management 2016. “A Taxonomy of Industrial Service Systems Enabled by
19(2): 110–32. https://doi.org/10.1111/1540-5885.1920110. Digital Product Innovation.” In Proceedings of the 49th Hawaii
Garud, R., and M. A. Rappa. 1994. “A Socio‐Cognitive Model of International Conference on System Sciences (HICSS-49) 1236–
Technology Evolution: The Case of Cochlear Implants.” Orga- 45. https://doi.org/10.1109/HICSS.2016.157.
nization Science 5(3): 344–62. https://doi.org/10.1287/orsc.5. Hill, C. W. L., and F. T. Rothaermel. 2003. “The Performance of
3.344. Incumbent Firms in the Face of Radical Technological Innova-
GE. 2020. Better Performance From Fewer Parts. GE.com. https:// tion.” The Academy of Management Review 28(2): 257. https://
web.archive.org/web/20200522182434/https://www.ge.com/ doi.org/10.2307/30040712.
additive/additive-manufacturing/industries/aviation-aerospace. Holmström, J., M. Holweg, S. H. Khajavi, and J. Partanen. 2016.
Gebler, M., A. J. S. Uiterkamp, and C. Visser. 2014. “A Global Sus- “The Direct Digital Manufacturing (R)evolution: Definition of a
tainability Perspective on 3D Printing Technologies.” Energy Research Agenda.” Operations Management Research 9(1–2): 1–
Policy 74: 158–67. https://doi.org/10.1016/j.enpol.2014.08.033. 10. https://doi.org/10.1007/s12063-016-0106-z.
Geels, F. W. 2005. Technological Transitions and System Innova- Holzmann, P., R. J. Breitenecker, and E. J. Schwarz. 2020. “Business
tions: A Co-Evolutionary and Socio-Technical Analysis. Chelten- Model Patterns for 3D Printer Manufacturers.” Journal of
ham, UK: Elgar. Manufacturing Technology Management 31(6): 1281–300.
Ghaffar, S., and P. Mullett. 2021. “Discussion: Commentary: 3D https://doi.org/10.1108/JMTM-09-2018-0313.
Printing Set to Transform the Construction Industry.” Proceed- Hoppmann, J., J. Huenteler, and B. Girod. 2014. “Compulsive Pol-
ings of the Institution of Civil Engineers—Structures and Build- icy‐Making—The Evolution of the German Feed‐In Tariff Sys-
ings 174(4): 336–7. https://doi.org/10.1680/jstbu.21.00024. tem for Solar Photovoltaic Power.” Research Policy 43(8): 1422–
Gibson, I., D. Rosen, B. Stucker, and M. Khorasani. 2014. Additive 41. https://doi.org/10.1016/j.respol.2014.01.014.
Manufacturing Technologies. New York: Springer. Hu, B., and Z. Sun. 2022. “Managing Self-Replicating Innovative
Goh, G. D., S. L. Sing, and W. Y. Yeong. 2020. “A Review on Goods.” Management Science 68(1): 399–419. https://doi.org/10.
Machine Learning in 3D Printing: Applications, Potential, and 1287/mnsc.2020.3936.
Challenges.” Artificial Intelligence Review 54(1): 63–94. https:// Hughes, T. P. 1983. Networks of Power: Electrification in Western
doi.org/10.1007/s10462-020-09876-9. Society, 1880–1930. Baltimore: Johns Hopkins University Press.
Greer, C., M. Burns, D. Wollman, and E. Griffor. 2019. Cyber- Hughes, T. P. 1987. “The Evolution of Large Technological Sys-
Physical Systems and Internet of Things. Gaithersburg, MD: tems.” In The Social Construction of Technological Systems, edi-
National Institute of Standards and Technology. ted by W. E. Bijker, T. P. Hughes, and T. J. Pinch, 51–82.
Greul, A., J. West, and S. Bock. 2018. “Open at Birth? Why New Cambridge, MA: MIT Press.
Firms Do (or Don’t) Use Open Innovation.” Strategic Entrepre- Hull, C., M. Feygin, Y. Baron, R. Sanders, E. Sachs, A. Lightman,
neurship Journal 12(3): 392–420. Portico. https://doi.org/10. and T. Wohlers. 1995. “Rapid Prototyping: Current Technology
1002/sej.1282. and Future Potential.” Rapid Prototyping Journal 1(1): 11–9.
Haenlein, M., and A. Kaplan. 2019. “A Brief History of Artificial https://doi.org/10.1108/13552549510732026.
Intelligence: On the Past, Present, and Future of Artificial Intel- Iansiti, M., and K. R. Lakhani. 2020. Competing in the Age of AI:
ligence.” California Management Review 61(4): 5–14. https:// Strategy and Leadership When Algorithms and Networks Run
doi.org/10.1177/0008125619864925. the World. Boston: Harvard Business Review Press.
Haleem, A., M. Javaid, and S. Rab. 2020. “Impact of Additive Jacobs, P. F. 1992. Rapid Prototyping & Manufacturing: Fundamen-
Manufacturing in Different Areas of Industry 4.0.” Interna- tals of Stereolithography. Dearborn, MI: Society of Manufactur-
tional Journal of Logistics Systems and Management 37(2): 239. ing Engineers.
https://doi.org/10.1504/IJLSM.2020.110578. Javaid, M., A. Haleem, R. Pratap Singh, and R. Suman. 2021.
Han, K., and J. Shin. 2014. “A Systematic Way of Identifying and “Industrial Perspectives of 3D Scanning: Features, Roles and
Forecasting Technological Reverse Salients Using QFD, Biblio- It's Analytical Applications.” Sensors International 2: 100114.
metrics, and Trend Impact Analysis: A Carbon Nanotube Bio- https://doi.org/10.1016/j.sintl.2021.100114.
sensor Case.” Technovation 34(9): 559–70. https://doi.org/10. Jiang, R., R. Kleer, and F. T. Piller. 2017. “Predicting the Future of
1016/j.technovation.2014.05.009. Additive Manufacturing: A Delphi Study on Economic and
Hanelt, A., E. Piccinini, R. W. Gregory, B. Hildebrandt, and L. M. Societal Implications of 3D Printing for 2030.” Technological
Kolbe. 2015, March. “Digital Transformation of Primarily Physical Forecasting and Social Change 117: 84–97. https://doi.org/10.
Industries-Exploring the Impact of Digital Trends on Business 1016/j.techfore.2017.01.006.
Models of Automobile Manufacturers.” Proceedings der 12. Inter- Kalantari, H. D., L. W. Johnson, and C. R. Perera. 2021. “The Effect
nationalen Tagung Wirtschaftsinformatik: 1313–27. of Gender on Willingness to Pay for Mass Customised Running
Heinen, J. J., and K. Hoberg. 2019. “Assessing the Potential of Addi- Shoes.” Journal of Global Fashion Marketing 12(2): 161–75.
tive Manufacturing for the Provision of Spare Parts.” Journal of https://doi.org/10.1080/20932685.2020.1867878.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
550 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Kapoor, R., and P. J. McGrath. 2014. “Unmasking the Interplay Open Journal of the Industrial Electronics Society 1: 298–310.
Between Technology Evolution and R&D Collaboration: Evi- https://doi.org/10.1109/OJIES.2020.3031660.
dence From the Global Semiconductor Manufacturing Indus- Lenfle, S., and J. Söderlund. 2022. “Project‐Oriented Agency and
try, 1990–2010.” Research Policy 43(3): 555–69. https://doi.org/ Regeneration in Socio‐Technical Transition: Insights From the
10.1016/j.respol.2013.08.002. Case of Numerical Weather Prediction (1978–2015).” Research
Kenney, M. E. 2013. Cost Reduction Through the Use of Additive Policy 51(3): 104455. https://doi.org/10.1016/j.respol.2021.
Manufacturing (3D Printing) and Collaborative Product Life Cycle 104455.
Management Technologies to Enhance the Navy's Maintenance Li, L. 2018. “China's Manufacturing Locus in 2025: With a Compar-
Programs. Doctoral dissertation. Monterey, CA: Naval Postgradu- ison of “Made-in-China 2025” and “Industry 4.0”.” Technologi-
ate School. https://apps.dtic.mil/sti/citations/ADA591795. cal Forecasting and Social Change 135: 66–74. https://doi.org/
Khan, H. N., D. A. Hounshell, and E. R. H. Fuchs. 2018. “Science 10.1016/j.techfore.2017.05.028.
and Research Policy at the End of Moore's Law.” Nature Elec- Li, S. 2013. A Desktop Factory; 3-D Printers, Which Can Make Plas-
tronics 1(1): 14–21. https://doi.org/10.1038/s41928-017-0005-9. tic Objects, Are Creeping Into the Home. Los Angeles Times,
Khorasani, M., A. Ghasemi, B. Rolfe, and I. Gibson. 2021. “Additive July 28, p. B1.
Manufacturing a Powerful Tool for the Aerospace Industry.” Lien, Shao-Yu, Shin-Ming Cheng, Sung-Yin Shih, and Kwang-
Rapid Prototyping Journal 28(1): 87–100. https://doi.org/10. Cheng Chen. 2012. “Radio Resource Management for QoS
1108/RPJ-01-2021-0009. Guarantees in Cyber-Physical Systems.” IEEE Transactions on
Kim, D.‐J., and B. Kogut. 1996. “Technological Platforms and Diver- Parallel and Distributed Systems 23(9): 1752–61. https://doi.org/
sification.” Organization Science 7(3): 283–301. https://doi.org/ 10.1109/TPDS.2012.151.
10.1287/orsc.7.3.283. Luchs, M. G., K. S. Swan, and M. E. H. Creusen. 2015. “Perspective:
Kim, H., Y. Lin, and T.-L. B. Tseng. 2018. “A Review on Quality A Review of Marketing Research on Product Design With
Control in Additive Manufacturing.” Rapid Prototyping Journal Directions for Future Research.” Journal of Product Innovation
24(3): 645–69. https://doi.org/10.1108/RPJ-03-2017-0048. Management 33(3): 320–41. Portico. https://doi.org/10.1111/
Kling, R. 1996. “Hopes and Horrors: Technological Utopianism and jpim.12276.
Anti-Utopianism in Narratives of Computerization.” In Com- Lyytinen, K. 2022. “Innovation Logics in the Digital Era: A Sys-
puterization and Controversy: Value Conflicts and Social temic Review of the Emerging Digital Innovation Regime.”
Choices, edited by R. Kling, 40–58. San Diego: Academic Press. Innovation: Organization and Management 24(1): 13–34.
Knofius, N., M. C. van der Heijden, and W. H. M. Zijm. 2019. “Con- https://doi.org/10.1080/14479338.2021.1938579.
solidating Spare Parts for Asset Maintenance With Additive Mäkitie, T., J. Hanson, M. Steen, T. Hansen, and A. D. Andersen.
Manufacturing.” International Journal of Production Economics 2022. “Complementarity Formation Mechanisms in Technol-
208: 269–80. https://doi.org/10.1016/j.ijpe.2018.11.007. ogy Value Chains.” Research Policy 51(7): 104559. https://doi.
Law, J., and M. Callon. 1992. “The Life and Death of an Aircraft: A org/10.1016/j.respol.2022.104559.
Network Analysis of Technical Change.” In Shaping Mahmood, T., and M. S. Mubarik. 2020. “Balancing Innovation and
Technology/Building Society: Studies in Sociotechnical Change, Exploitation in the Fourth Industrial Revolution: Role of Intel-
edited by W. E. Bijker and J. Law, 21–52. Cambridge, MA: MIT lectual Capital and Technology Absorptive Capacity.” Techno-
Press. logical Forecasting and Social Change 160: 120248. https://doi.
Lee, E. A. 2008. “Cyber Physical Systems: Design Challenges.” In org/10.1016/j.techfore.2020.120248.
2008 11th IEEE International Symposium on Object and Compo- Majidpour, M. 2016. “Technological Catch‐Up in Complex Product
nent‐Oriented Real‐Time Distributed Computing (ISORC). Systems.” Journal of Engineering and Technology Management
https://doi.org/10.1109/isorc.2008.25. 41: 92–105. https://doi.org/10.1016/j.jengtecman.2016.07.003.
Lee, J., B. Bagheri, and H.‐A. Kao. 2015. “A Cyber‐Physical Systems Markus, M. L. 1994. “Electronic Mail as the Medium of Managerial
Architecture for Industry 4.0‐Based Manufacturing Systems.” Choice.” Organization Science 5(4): 502–27. https://doi.org/10.
Manufacturing Letters 3: 18–23. https://doi.org/10.1016/j. 1287/orsc.5.4.502.
mfglet.2014.12.001. Mathews, J. A., H. Tan, and M. C. Hu. 2018. “Moving to a Circular
Lee, J., and N. Berente. 2012. “Digital Innovation and the Division Economy in China: Transforming Industrial Parks into Eco-
of Innovative Labor: Digital Controls in the Automotive Indus- Industrial Parks.” California Management Review 60(3): 157–81.
try.” Organization Science 23(5): 1428–47. https://doi.org/10. https://doi.org/10.1177/0008125617752692.
1287/orsc.1110.0707. Mavri, M., E. Fronimaki, and A. Kadrefi. 2021. “Survey Analysis for
Lee, J., H.-C. Kim, J.-W. Choi, and I. H. Lee. 2017. “A Review on the Adoption of 3D Printing Technology: Consumers' Perspec-
3D Printed Smart Devices for 4D Printing.” International Jour- tive.” Journal of Science and Technology Policy Management 14
nal of Precision Engineering and Manufacturing-Green Technol- (2): 353–85. https://doi.org/10.1108/JSTPM-02-2020-0023.
ogy 4(3): 373–83. https://doi.org/10.1007/s40684-017-0042-x. Mendis, D. K., M. A. Lemley, and M. Rimmer, eds. 2019. 3D Print-
Lehenkari, J., and R. Miettinen. 2002. “Standardisation in the ing and Beyond: Intellectual Property and Regulation. Chelten-
Construction of a Large Technological System—The Case of the ham, UK: Elgar.
Nordic Mobile Telephone System.” Telecommunications Policy Millard, J., M. N. Sorivelle, S. Deljanin, E. Unterfrauner, and
26(3–4): 109–27. https://doi.org/10.1016/s0308-5961(02)00004-6. C. Voigt. 2018. “Is the Maker Movement Contributing to
Leitão, P., F. Pires, S. Karnouskos, and A. W. Colombo. 2020. “Quo Sustainability?” Sustainability 10(7): 7. https://doi.org/10.3390/
Vadis Industry 4.0? Position, Trends, and Challenges.” IEEE su10072212.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 551

Minguella-Canela, J., S. M. Planas, J. R. G. Ayats, and M. A. S. Peltz, J. F. . 1987. Sylmar Firm Puts Hot Idea for Model-Making
L
opez. 2019. “Study and Comparison of the Different Costs' Machine on the Market Via Canada. Los Angeles Times,
Schema Associated to Geometry, Material and Processing October 11, 1987. p. OC-D13.
Between 3D Printing, Injection Molding and Machining Petrick, I. J., and T. W. Simpson. 2013. “3D Printing Disrupts
Manufacturing Technologies.” Procedia Manufacturing 41: Manufacturing: How Economies of One Create New Rules of
280–7. https://doi.org/10.1016/j.promfg.2019.09.010. Competition.” Research‐Technology Management 56(6): 12–6.
Molitch-Hou, M. 2017. “Michelin and Fives' AddUp: From Making https://doi.org/10.5437/08956308x5606193.
Tires to Making Metal 3D Printers.” January 30, 2017. https:// Pradel, P., Z. Zhu, R. Bibb, and J. Moultrie. 2018. “Investigation of
www.engineering.com/story/michelin-and-fives-addup-from- Design for Additive Manufacturing in Professional Design Prac-
making-tires-to-making-metal-3d-printers. tice.” Journal of Engineering Design 29(4–5): 165–200. https://
Moore, G. E. 2006. “Moore's Law at 40.” In Understanding Moore's doi.org/10.1080/09544828.2018.1454589.
Law: Four Decades of Innovation, edited by D. C. Brock, 67–84. Prater, T. J., Q. A. Bean, R. D. Beshears, T. D. Rolin, N. J.
Philadelphia: CHC Press. Werkheiser, E. A. Ordonez, R. M. Ryan, and F. E. Ledbetter,
Mubarak, M. F., and M. Petraite. 2020. “Industry 4.0 Technologies, III. 2016. “Summary Report on Phase I Results From the 3D
Digital Trust and Technological Orientation: What Matters in Printing in Zero G Technology Demonstration Mission.” https://
Open Innovation?” Technological Forecasting and Social Change ntrs.nasa.gov/citations/20160008972.
161: 120332. https://doi.org/10.1016/j.techfore.2020.120332. Prencipe, A., A. Davies, and M. Hobday, eds. 2003. The Business of
Muir, M., and A. Haddud. 2017. “Additive Manufacturing in the Systems Integration. Oxford: Oxford.
Mechanical Engineering and Medical Industries Spare Parts Sup- Quah, D. 2003. “Digital Goods and the New Economy.” In New
ply Chain.” Journal of Manufacturing Technology Management Economy Handbook, edited by D. Jones, 289–321. San Diego:
29(2): 372–97. https://doi.org/10.1108/JMTM-01-2017-0004. Academic Press.
Mulder, K., and M. Knot. 2001. “PVC Plastic: A History of Systems Raasch, C., C. Herstatt, and K. Balka. 2009. “On the Open Design of
Development and Entrenchment.” Technology in Society 23(2): Tangible Goods.” R&D Management 39(4): 382–93. https://doi.
265–86. https://doi.org/10.1016/s0160-791x(01)00013-6. org/10.1111/j.1467-9310.2009.00567.x.
Mussatto, A., R. Groarke, R. K. Vijayaraghavan, C. Hughes, M. A. Radstok, E. 1999. “Rapid Tooling.” Rapid Prototyping Journal 5(4):
Obeidi, M. N. Dogu, M. A. Yalçin, P. J. McNally, Y. Delaure, 164–9. https://doi.org/10.1108/13552549910295488.
and D. Brabazon. 2022. “Assessing Dependency of Part Proper- Rajkumar, R. ( Raj), I. Lee, L. Sha, and J. Stankovic. 2010. “Cyber‐
ties on the Printing Location in Laser-Powder Bed Fusion Metal Physical Systems.” In Proceedings of the 47th Design Automation
Additive Manufacturing.” Materials Today Communications 30: Conference. https://doi.org/10.1145/1837274.1837461.
103209. https://doi.org/10.1016/j.mtcomm.2022.103209. Randhawa, K., J. West, K. Skellern, and E. Josserand. 2020. “Evolv-
Naik, H. S., A. Fritzsche, and K. M. Moeslein. 2021. “Modularity in ing a Value Chain to an Open Innovation Ecosystem: Cognitive
Making: Simplifying Solution Space for User Innovation.” R&D Engagement of Stakeholders in Customizing Medical
Management 51(1): 57–72. https://doi.org/10.1111/radm.12427. Implants.” California Management Review 63(2): 101–34.
Nasiri, M., M. Saunila, J. Ukko, T. Rantala, and H. Rantanen. 2020. https://doi.org/10.1177/0008125620974435.
“Shaping Digital Innovation Via Digital-Related Capabilities.” Rayna, T., and L. Striukova. 2015. “Open Innovation 2.0: Is Co‐crea-
Information Systems Frontiers. https://doi.org/10.1007/s10796- tion the Ultimate Challenge?” International Journal of Technol-
020-10089-2. ogy Management 69(1): 38. https://doi.org/10.1504/ijtm.2015.
Negroponte, N. 1996. Being Digital. New York: Vintage. 071030.
Nelson, R. R., and S. G. Winter. 1982. An Evolutionary Theory of Rayna, T., and L. Striukova. 2016. “From Rapid Prototyping to Home
Economic Change. Cambridge, MA: Belknap Press. Fabrication: How 3D Printing is Changing Business Model Inno-
Nerlich, B. 2005. “From Nautilus to Nanobo(a)ts: the visual con- vation.” Technological Forecasting and Social Change 102: 214–
struction of nanoscience.” AZojono: Journal of Nanotechnology. 24. https://doi.org/10.1016/j.techfore.2015.07.023.
https://doi.org/10.2240/azojono0109. Rayna, T., and L. Striukova. 2021. “Involving Consumers: The Role
NIST. 2020. “Cyber-Physical Systems, National Institute of Standards of Digital Technologies in Promoting ‘Prosumption’ and User
and Technology.” U.S. Department of Commerce, March Innovation.” Journal of the Knowledge Economy 12(1): 218–37.
16, 2020. https://web.archive.org/web/20200316224708/https:// https://doi.org/10.1007/s13132-016-0390-8.
www.nist.gov/el/cyber-physical-systems. Rayna, T. 2008. “Understanding the Challenges of the Digital Econ-
Nonaka, I., and H. Takeuchi. 2007. “The Knowledge-Creating Com- omy: The Nature of Digital Goods.” Communications and Strat-
pany.” Harvard Business Review 85(7/8): 162. egies 71: 13–6.
Nylén, D., and J. Holmström. 2015. “Digital Innovation Strategy: A Rayna, T., L. Striukova, and J. Darlington. 2015. “Co-Creation and
Framework for Diagnosing and Improving Digital Product and User Innovation: The Role of Online 3D Printing Platforms.”
Service Innovation.” Business Horizons 58(1): 57–67. https:// Journal of Engineering and Technology Management 37: 90–102.
doi.org/10.1016/j.bushor.2014.09.001. https://doi.org/10.1016/j.jengtecman.2015.07.002.
Nyman, H. J., and P. Sarlin. 2014. “From Bits to Atoms: 3D Printing Rindfleisch, A., M. O'Hern, and V. Sachdev. 2017. “The Digital
in the Context of Supply Chain Strategies.” In Proceedings of Revolution, 3D Printing, and Innovation as Data.” Journal of
the 47th Hawaii International Conference on System Sciences Product Innovation Management 34(5): 681–90. Portico. https://
4190–9. https://doi.org/10.1109/HICSS.2014.518. doi.org/10.1111/jpim.12402.
Osterwalder, A., and Y. Pigneur. 2004. “An Ontology for E-Business Rindfleisch, A., and M. O’Hern. 2015. “Brand Remixing: 3D Print-
Models.” In Value Creation From E-Business Models, edited by ing the Nokia Case.” Review of Marketing Research: 53–81.
W. Currie, 65–97. Oxford: Butterworth-Heinemann. https://doi.org/10.1108/s1548-643520150000012003.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
552 JOURNAL OF PRODUCT INNOVATION MANAGEMENT

Rogers, E. M., and F. F. Shoemaker. 1971. Communication of Inno- Stanko, M. A. 2020. “Building an Understanding of How Winning
vations: A Cross-Cultural Approach. New York: FreePress. Products Emerge When Open and Proprietary Products Coex-
Rouf, S., A. Raina, M. Irfan Ul Haq, N. Naveed, S. Jeganmohan, ist: Evidence From the RepRap Community.” Creativity and
and A. Farzana Kichloo. 2022. “3D Printed Parts and Mechani- Innovation Management 29(3): 398–412. https://doi.org/10.
cal Properties: Influencing Parameters, Sustainability Aspects, 1111/caim.12376.
Global Market Scenario, Challenges and Applications.” Stanko, M. A., and B. J. Allen. 2022. “Disentangling the Collective
Advanced Industrial and Engineering Polymer Research. 5: Motivations for User Innovation in a 3D Printing Community.”
143–58. https://doi.org/10.1016/j.aiepr.2022.02.001. Technovation 111: 102387. https://doi.org/10.1016/j.
Rouhani, C. . 2018. “Three BIG Challenges for Using FDM 3D Print- technovation.2021.102387.
ing at Scale.” Sculpteo.com. August 2, 2018. https://www. Stone, P., R. Brooks, E. Brynjolfsson, R. Calo, O. Etzioni, G. Hager,
sculpteo.com/blog/2018/08/02/three-big-challenges-for-using- J. Hirschberg, et al. 2016. Artificial Intelligence and Life in 2030:
fdm-3d-printing-at-scale/. The One Hundred Year Study on Artificial Intelligence. Stanford,
Söderberg, J., and A. Daoud. 2012. “Atoms Want to Be Free Too! CA: Stanford University.
Expanding the Critique of Intellectual Property to Physical Sturgeon, T. J. 2019. “Upgrading Strategies for the Digital Econ-
Goods.” TripleC: Communication, Capitalism & Critique. omy.” Global Strategy Journal 11(1): 34–57. Portico. https://doi.
Open Access Journal for a Global Sustainable Information org/10.1002/gsj.1364.
Society 10(1): 66–76. https://doi.org/10.31269/triplec. Thürer, M., Y. Huang, and M. Stevenson. 2021. “Workload Control
v10i1.288. in Additive Manufacturing Shops Where Post-Processing is a
Sahini, D. K., J. Ghose, S. K. Jha, A. Behera, and A. Mandal. 2020. Constraint: An Assessment by Simulation.” International Jour-
“Optimization and Simulation of Additive Manufacturing Pro- nal of Production Research 59(14): 4268–86. https://doi.org/10.
cesses: Challenges and Opportunities—A Review.” In Additive 1080/00207543.2020.1761038.
Manufacturing Applications for Metals and Composites, edited Thomas-Seale, L. E. J., J. C. Kirkman-Brown, M. M. Attallah, D. M.
by K. R. Balasubramanian and V. Senthilkumar, 187–209. Her- Espino, and D. E. T. Shepherd. 2018. “The Barriers to the Pro-
shey, Penn: IGI Global. gression of Additive Manufacture: Perspectives From UK
Salmi, M., K.-S. Paloheimo, J. Tuomi, J. Wolff, and A. Mäkitie. Industry.” International Journal of Production Economics 198:
2013. “Accuracy of Medical Models Made by Additive 104–18. https://doi.org/10.1016/j.ijpe.2018.02.003.
Manufacturing (Rapid Manufacturing).” Journal of Cranio- Tilbury, D. M. 2019. “Cyber‐Physical Manufacturing Systems.”
Maxillofacial Surgery 41(7): 603–9. https://doi.org/10.1016/j. Annual Review of Control, Robotics, and Autonomous Systems
jcms.2012.11.041. 2(1): 427–43. https://doi.org/10.1146/annurev-control-053018-
Sandberg, J., J. Holmström, and K. Lyytinen. 2020. “Digitization 023652.
and Phase Transitions in Platform Organizing Logics: Evidence Turner, F. 2010. From Counterculture to Cyberculture: Stewart
from the Process Automation Industry.” MIS Quarterly 44(1): Brand, the Whole Earth Network, and the Rise of Digital Utopi-
129–53. https://doi.org/10.25300/MISQ/2020/14520. anism. Chicago: University of Chicago Press.
Schaller, R. R. 1997. “Moore's Law: Past, Present and Future.” IEEE van den Ende, J., G. van de Kaa, S. den Uijl, and H. J. de Vries.
Spectrum 34(6): 52–9. https://doi.org/10.1109/6.591665. 2012. “The Paradox of Standard Flexibility: The Effects of Co‐
Schubert, C., J. Sydow, and A. Windeler. 2013. “The Means of Man- evolution Between Standard and Interorganizational Network.”
aging Momentum: Bridging Technological Paths and Organisa- Organization Studies 33(5–6): 705–36. https://doi.org/10.1177/
tional Fields.” Research Policy 42(8): 1389–405. https://doi.org/ 0170840612443625.
10.1016/j.respol.2013.04.004. Veers, P., K. Dykes, E. Lantz, S. Barth, C. L. Bottasso, O. Carlson,
Sculpteo. 2020. The State of 3D Printing 2020. San Francisco: A. Clifton, et al. 2019. “Grand Challenges in the Science of
Sculpteo. Wind Energy.” Science 366(6464): eaau2027.
Sgarbossa, F., M. Peron, F. Lolli, and E. Balugani. 2021. “Conven- Walsh, G. S., J. Przychodzen, and W. Przychodzen. 2017. “Support-
tional or Additive Manufacturing for Spare Parts Management: ing the SME Commercialization Process: The Case of
An Extensive Comparison for Poisson Demand.” International 3D Printing Platforms.” Small Enterprise Research 24(3):
Journal of Production Economics 233: 107993. https://doi.org/ 257–73. https://doi.org/10.1080/13215906.2017.1396490.
10.1016/j.ijpe.2020.107993. Wang, L., S. Jiang, and S. Zhang. 2020. “Mapping Technological
Shane, S. 2000. “Prior Knowledge and the Discovery of Entrepre- Trajectories and Exploring Knowledge Sources: A Case Study
neurial Opportunities.” Organization Science 11(4): 448–69. of 3D Printing Technologies.” Technological Forecasting and
https://doi.org/10.1287/orsc.11.4.448.14602. Social Change 161: 120251. https://doi.org/10.1016/j.techfore.
Shapiro, C., and H. R. Varian. 1999. Information Rules: A Strategic 2020.120251.
Guide to the Network Economy 2163. Boston, MA: Harvard Busi- Wang, Q., X. Sun, S. Cobb, G. Lawson, and S. Sharples. 2016. “3D
ness School Press. Printing System: An Innovation for Small-Scale Manufacturing
Shen, X. 2019. “Building ‘Mass’ and ‘Momentum’—A Latecomer in Home Settings?—Early Adopters of 3D Printing Systems in
Country's Passage to Large Technological Systems—The Case China.” International Journal of Production Research 54(20):
of China.” Technology Analysis & Strategic Management 31(5): 6017–32. https://doi.org/10.1080/00207543.2016.1154211.
503–16. https://doi.org/10.1080/09537325.2018.1519184. West, J. 2008. “Commercializing Open Science: Deep Space Com-
Stanko, M. A. 2016. “Toward a Theory of Remixing in Online Inno- munications as the Lead Market for Shannon Theory, 1960–
vation Communities.” Information Systems Research 27(4): 73.” Journal of Management Studies 45(8): 1506–32. https://doi.
773–91. https://doi.org/10.1287/isre.2016.0650. org/10.1111/j.1467-6486.2008.00807.x.
 15405885, 2023, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12681 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RAYNA and WEST 553

West, J. 2014. “Too Little, Too Early: California's Transient Advan- Paradigm With Lifecycle Personalization Concerns.” Interna-
tage in the Photovoltaic Solar Industry.” The Journal of Tech- tional Journal of Production Research 57(8): 2571–84. https://
nology Transfer 39(3): 487–501. https://doi.org/10.1007/s10961- doi.org/10.1080/00207543.2018.1530475.
012-9291-6.
West, J., and G. Kuk. 2016. “The Complementarity of Openness:
How MakerBot Leveraged Thingiverse in 3D Printing.” Techno- AUTHOR BIOGRAPHIES
logical Forecasting and Social Change 102: 169–81. https://doi.
org/10.1016/j.techfore.2015.07.025. Thierry Rayna is Chaired Professor of Technology
West, J., and M. Bogers. 2014. “Leveraging External Sources of
for Change at the École Polytechnique and a Fellow
Innovation: A Review of Research on Open Innovation.” Jour-
nal of Product Innovation Management 31(4): 814–31. Portico.
of the CNRS Innovation Interdisciplinary Institute
https://doi.org/10.1111/jpim.12125. (i³), France. His research investigates the changes
West, J. 2006. “Does Appropriability Enable or Retard Open Inno- brought about by digital technologies (such as 3D
vation?” In Open Innovation: Researching a New Paradigm, edi- printing, artificial intelligence, blockchain, cloud, IoT,
ted by H. Chesbrough, W. Vanhaverbeke, and J. West, 109–33. platforms) on business strategy, organization, and
Oxford: Oxford University Press. public policy. In particular, his research focuses on the
Westerweel, B., R. J. I. Basten, and G.-J. van Houtum. 2018. “Tradi-
issues of business model innovation, ecosystem gover-
tional or Additive Manufacturing? Assessing Component
nance, and impact (environmental, social, economic),
Design Options through Lifecycle Cost Analysis.” European
Journal of Operational Research 270(2): 570–85. https://doi.org/ in relation to digital technologies. Prof. Rayna has a
10.1016/j.ejor.2018.04.015. long track record of collaboration with companies
Windrum, P., K. Frenken, and L. Green. 2017. “The Importance of (small and large) in various industries “disrupted” by
Ergonomic Design in Product Innovation. Lessons From the digital technologies, and policymakers.
Development of the Portable Computer.” Industrial and Corpo-
rate Change 26(6): 953–71. https://doi.org/10.1093/icc/dtx006. Joel West is Professor of Entrepreneurship at Hilde-
Wohlers, T. 2005. “New Trends and Developments in Additive Fab- gard College, and Professor Emeritus at the Keck
rication.” In Virtual Modeling and Rapid Manufacturing: Graduate Institute. His research has examined the
Advanced Research in Virtual and Rapid Prototyping, edited by impact of technological evolution upon industry struc-
P. B
artolo. London: Taylor & Francis. ture in personal computers, mobile phones, digital
Wohlers. 2012. Wohlers Report 2012. Ft Collins, CO: Wohlers communications, solar energy, and 3D printing. His
Associates.
has also pioneered the understanding of how firms
Wohlers. 2017. Wohlers Report 2017. Ft Collins, CO: Wohlers
Associates.
utilize strategic openness to build competitive advan-
Wohlers. 2020. Wohlers Report 2020. Ft Collins, CO: Wohlers tage through open platforms, open standards, open
Associates. communities, and open consortia. He is the co‐editor
Yoo, Y., O. Henfridsson, and K. Lyytinen. 2010. “Research Com- of Open Innovation: Researching a New Paradigm
mentary—The New Organizing Logic of Digital Innovation: An (2006), New Frontiers of Open Innovation (2014) and
Agenda for Information Systems Research.” Information Sys- the Oxford Handbook of Open Innovation (2023) and
tems Research 21(4): 724–35. https://doi.org/10.1287/isre.1100.
cofounder of the World Open Innovation Conference.
0322.
Yoo, Y., R. J. Boland, K. Lyytinen, and A. Majchrzak. 2012. “Orga-
nizing for Innovation in the Digitized World.” Organization Sci-
ence 23(5): 1398–408. https://doi.org/10.1287/orsc.1120.0771.
Zhan, T., K. Yin, J. Xiong, Z. He, and S.‐T. Wu. 2020. “Augmented SU PP O R TI N G I N F O RMA TI O N
Reality and Virtual Reality Displays: Perspectives and Chal- Additional supporting information can be found online
lenges.” IScience 23(8): 101397. https://doi.org/10.1016/j.isci. in the Supporting Information section at the end of this
2020.101397. article.
Zhang, J., F.‐Y. Wang, K. Wang, W.‐H. Lin, X. Xu, and C. Chen.
2011. “Data‐Driven Intelligent Transportation Systems: A Sur-
vey.” IEEE Transactions on Intelligent Transportation Systems How to cite this article: Rayna, Thierry, and
12(4): 1624–39. https://doi.org/10.1109/tits.2011.2158001. Joel West. 2023. “Where Digital Meets Physical
Zheng, P., T.-J. Lin, C.-H. Chen, and X. Xu. 2018. “A Systematic
Innovation: Reverse Salients and the Unrealized
Design Approach for Service Innovation of Smart Product-
Dreams of 3D Printing.” Journal of Product
Service Systems.” Journal of Cleaner Production 201: 657–67.
https://doi.org/10.1016/j.jclepro.2018.08.101. Innovation Management 40(4): 530–553. https://
Zheng, P., Y. Lin, C.-H. Chen, and X. Xu. 2019. “Smart, Connected doi.org/10.1111/jpim.12681
Open Architecture Product: An IT-Driven Co-Creation
ARTIFICIAL INTELLIGENCE
AND LIFE IN 2030
ONE HUNDRED YEAR STUDY ON ARTIFICIAL INTELLIGENCE | REPORT OF THE 2015 STUDY PANEL | SEPTEMBER 2016

PREFACE
The One Hundred Year Study on
Artificial Intelligence, launched
in the fall of 2014, is a long-
term investigation of the field of
Artificial Intelligence (AI) and
its influences on people, their
communities, and society. It
considers the science, engineering,
and deployment of AI-enabled
computing systems. As its core
activity, the Standing Committee
that oversees the One Hundred
Year Study forms a Study Panel
every five years to assess the
current state of AI. The Study
Panel reviews AI’s progress in the years following the immediately prior report, The overarching purpose
envisions the potential advances that lie ahead, and describes the technical and
societal challenges and opportunities these advances raise, including in such arenas as of the One Hundred Year
ethics, economics, and the design of systems compatible with human cognition. The
overarching purpose of the One Hundred Year Study’s periodic expert review is to
Study’s periodic expert
provide a collected and connected set of reflections about AI and its influences as the review is to provide a
field advances. The studies are expected to develop syntheses and assessments that
provide expert-informed guidance for directions in collected and connected
AI research, development, and systems design, as well as programs and policies to
help ensure that these systems broadly benefit individuals and society.1 set of reflections about
The One Hundred Year Study is modeled on an earlier effort informally known as
AI and its influences as
the “AAAI Asilomar Study.” During 2008-2009, the then president of the Association
for the Advancement of Artificial Intelligence (AAAI), Eric Horvitz, assembled a the field advances.
group of AI experts from multiple institutions and areas of the field, along with
scholars of cognitive science, philosophy, and law. Working in distributed subgroups,
the participants addressed near-term AI developments, long-term possibilities,
and legal and ethical concerns, and then came together in a three-day meeting at
Asilomar to share and discuss their findings. A short written report on the intensive
meeting discussions, amplified by the participants’ subsequent discussions with other
colleagues, generated widespread interest and debate in the field and beyond.
The impact of the Asilomar meeting, and important advances in AI that included
AI algorithms and technologies starting to enter daily life around the globe, spurred
the idea of a long-term recurring study of AI and its influence on people and society.
The One Hundred Year Study was subsequently endowed at a university to enable
1 “One Hundred Year Study on Artificial Intelligence (AI100),” Stanford University, accessed
August 1, 2016, https://ai100.stanford.edu.
 extended deep thought and cross-disciplinary scholarly investigations that could
inspire innovation and provide intelligent advice to government agencies and industry.
This report is the first in the planned series of studies that will continue for at least
a hundred years. The Standing Committee defined a Study Panel charge for the
inaugural Study Panel in the summer of 2015 and recruited Professor Peter Stone,
at the University of Texas at Austin, to chair the panel. The seventeen-member
Study Panel, comprised of experts in AI from academia, corporate laboratories
and industry, and AI-savvy scholars in law, political science, policy, and economics,
was launched in mid-fall 2015. The participants represent diverse specialties and
geographic regions, genders, and career stages.
The Standing Committee extensively discussed ways to frame the Study Panel
charge to consider both recent advances in AI and potential societal impacts on jobs,
the environment, transportation, public safety, healthcare, community engagement,
and government. The committee considered various ways to focus the study,
including surveying subfields and their status, examining a particular technology
such as machine learning or natural language processing, and studying particular
application areas such as healthcare or transportation. The committee ultimately
chose a thematic focus on “AI and Life in 2030” to recognize that AI’s various uses
and impacts will not occur independently of one another, or of a multitude of other
societal and technological developments. Acknowledging the central role cities have
played throughout most of human experience, the focus was narrowed to the large
urban areas where most people live. The Standing Committee further narrowed the
focus to a typical North American city in recognition of the great variability of urban
settings and cultures around the world, and limits on the first Study Panel’s efforts.
The Standing Committee expects that the projections, assessments, and proactive
guidance stemming from the study will have broader global relevance and is making
plans for future studies to expand the scope of the project internationally.

TABLE OF CONTENTS
PREFACE
EXECUTIVE SUMMARY 4
OVERVIEW 6
SECTION I: WHAT IS ARTIFICIAL INTELLIGENCE? 12
Defining AI 12
AI Research Trends 14
SECTION II: AI BY DOMAIN 18
Transportation18
Home/Service Robots 24
Healthcare25
Education31
Low-resource Communities 35
Public Safety and Security 36
Employment and Workplace 38
Entertainment40
SECTION III: PROSPECTS AND RECOMMENDATIONS FOR AI PUBLIC POLICY 42
AI Policy, Now and in the Future 42
2 APPENDIX I: A SHORT HISTORY OF AI 50
 As one consequence of the decision to focus on life in North American cities,
military applications were deemed to be outside the scope of this initial report. This
is not to minimize the importance of careful monitoring and deliberation about
the implications of AI advances for defense and warfare, including potentially
destabilizing developments and deployments.
The report is designed to address four intended audiences. For the general public,
it aims to provide an accessible, scientifically and technologically accurate portrayal
of the current state of AI and its potential. For industry, the report describes relevant
technologies and legal and ethical challenges, and may help guide resource allocation.
The report is also directed to local, national, and international governments to help
them better plan for AI in governance. Finally, the report can help AI researchers,
as well as their institutions and funders, to set priorities and consider the ethical and
legal issues raised by AI research and its applications.
Given the unique nature of the One Hundred Year Study on AI, we expect that
future generations of Standing Committees and Study Panels, as well as research
scientists, policy experts, leaders in the private and public sectors, and the general
public, will reflect on this assessment as they make new assessments of AI’s future. We
hope that this first effort in the series stretching out before us will be useful for both its
failures and successes in accurately predicting the trajectory and influences of AI.
The Standing Committee is grateful to the members of the Study Panel for
investing their expertise, perspectives, and significant time to the creation of this
inaugural report. We especially thank Professor Peter Stone for agreeing to serve as
chair of the study and for his wise, skillful, and dedicated leadership of the panel,
its discussions, and creation of the report.
Standing Committee of the One Hundred Year Study of Artificial Intelligence
Barbara J. Grosz, Chair Russ Altman Eric Horvitz
Alan Mackworth Tom Mitchell Deirdre Mulligan Yoav Shoham

STUDY PANEL
Peter Stone, University of Texas at Austin, Chair
Rodney Brooks, Rethink Robotics
Erik Brynjolfsson, Massachussets Institute of Technology
Ryan Calo, University of Washington
Oren Etzioni, Allen Institute for AI
Greg Hager, Johns Hopkins University
Julia Hirschberg, Columbia University
Shivaram Kalyanakrishnan, Indian Institute of Technology Bombay
Ece Kamar, Microsoft Research
Sarit Kraus, Bar Ilan University
Kevin Leyton-Brown, University of British Columbia
David Parkes, Harvard University
William Press, University of Texas at Austin
AnnaLee (Anno) Saxenian, University of California, Berkeley
Julie Shah, Massachussets Institute of Technology
Milind Tambe, University of Southern California
Astro Teller, X
© 2016 by Stanford University. Artificial
Acknowledgments: The members of the Study Panel gratefully acknowledge the Intelligence and Life in 2030 is made
support of and valuable input from the Standing Committee, especially the chair, available under a Creative Commons
Barbara Grosz, who handled with supreme grace the unenviable role of mediating Attribution-NoDerivatives 4.0 License
between two large, very passionate committees. We also thank Kerry Tremain for his (International): https://creativecommons.
tireless and insightful input on the written product during the extensive editing and org/licenses/by-nd/4.0/
polishing process, which unquestionably strengthened the report considerably. 3
 EXECUTIVE SUMMARY
Artificial Intelligence (AI) is a science and a set of computational technologies that
are inspired by—but typically operate quite differently from—the ways people use
their nervous systems and bodies to sense, learn, reason, and take action. While the
rate of progress in AI has been patchy and unpredictable, there have been significant
advances since the field’s inception sixty years ago. Once a mostly academic area of
study, twenty-first century AI enables a constellation of mainstream technologies that
Substantial increases are having a substantial impact on everyday lives. Computer vision and AI planning,
for example, drive the video games that are now a bigger entertainment industry than
in the future uses of AI Hollywood. Deep learning, a form of machine learning based on layered representations
of variables referred to as neural networks, has made speech-understanding practical
applications, including
on our phones and in our kitchens, and its algorithms can be applied widely to an
more self-driving cars, array of applications that rely on pattern recognition. Natural Language Processing
(NLP) and knowledge representation and reasoning have enabled a machine to beat
healthcare diagnostics the Jeopardy champion and are bringing new power to Web searches.
While impressive, these technologies are highly tailored to particular tasks. Each
and targeted treatment, application typically requires years of specialized research and careful, unique
and physical assistance construction. In similarly targeted applications, substantial increases in the future
uses of AI technologies, including more self-driving cars, healthcare diagnostics
for elder care can be and targeted treatments, and physical assistance for elder care can be expected. AI
and robotics will also be applied across the globe in industries struggling to attract
expected. younger workers, such as agriculture, food processing, fulfillment centers, and
factories. They will facilitate delivery of online purchases through flying drones,
self-driving trucks, or robots that can get up the stairs to the front door.
This report is the first in a series to be issued at regular intervals as a part of the
One Hundred Year Study on Artificial Intelligence (AI100). Starting from a charge
given by the AI100 Standing Committee to consider the likely influences of AI in a
typical North American city by the year 2030, the 2015 Study Panel, comprising experts
in AI and other relevant areas focused their attention on eight domains they considered
most salient: transportation; service robots; healthcare; education; low-resource
communities; public safety and security; employment and workplace; and entertainment.
In each of these domains, the report both reflects on progress in the past fifteen years
and anticipates developments in the coming fifteen years. Though drawing from a
common source of research, each domain reflects different AI influences and challenges,
such as the difficulty of creating safe and reliable hardware (transportation and service
robots), the difficulty of smoothly interacting with human experts (healthcare and
education), the challenge of gaining public trust (low-resource communities and public
safety and security), the challenge of overcoming fears of marginalizing humans
(employment and workplace), and the social and societal risk of diminishing interpersonal
interactions (entertainment). The report begins with a reflection on what constitutes
Artificial Intelligence, and concludes with recommendations concerning AI-related
policy. These recommendations include accruing technical expertise about AI in
government and devoting more resources—and removing impediments—to research
on the fairness, security, privacy, and societal impacts of AI systems.
Contrary to the more fantastic predictions for AI in the popular press, the Study
Panel found no cause for concern that AI is an imminent threat to humankind.
No machines with self-sustaining long-term goals and intent have been developed,
nor are they likely to be developed in the near future. Instead, increasingly useful
applications of AI, with potentially profound positive impacts on our society and
economy are likely to emerge between now and 2030, the period this report
4 considers. At the same time, many of these developments will spur disruptions in
how human labor is augmented or replaced by AI, creating new challenges for the
economy and society more broadly. Application design and policy decisions made in
the near term are likely to have long-lasting influences on the nature and directions
of such developments, making it important for AI researchers, developers, social
scientists, and policymakers to balance the imperative to innovate with mechanisms
to ensure that AI’s economic and social benefits are broadly shared across society. If
society approaches these technologies primarily with fear and suspicion, missteps that
slow AI’s development or drive it underground will result, impeding important work
on ensuring the safety and reliability of AI technologies. On the other hand, if society While drawing on common
approaches AI with a more open mind, the technologies emerging from the field
could profoundly transform society for the better in the coming decades. research and technologies,
AI systems are specialized
Study Panel: Peter Stone, Chair, University of Texas at Austin, Rodney Brooks,
Rethink Robotics, Erik Brynjolfsson, Massachussets Institute of Technology, Ryan to accomplish particular
Calo, University of Washington, Oren Etzioni, Allen Institute for AI, Greg Hager, Johns
Hopkins University, Julia Hirschberg, Columbia University, Shivaram Kalyanakrishnan, tasks. Each application
Indian Institute of Technology Bombay, Ece Kamar, Microsoft Research, Sarit Kraus,
Bar Ilan University. Kevin Leyton-Brown, University of British Columbia, David Parkes,
requires years of focused
Harvard University, William Press, University of Texas at Austin, AnnaLee (Anno) research and a careful,
Saxenian, University of California, Berkeley, Julie Shah, Massachussets Institute of
Technology, Milind Tambe, University of Southern California, Astro Teller, X unique construction.
Standing Committee of the One Hundred Year Study of Artificial Intelligence:
Barbara J. Grosz, Chair, Russ Altman, Eric Horvitz, Alan Mackworth, Tom Mitchell,
Deidre Mulligan, Yoav Shoham

5
 OVERVIEW
The frightening, futurist portrayals of Artificial Intelligence that dominate films and
novels, and shape the popular imagination, are fictional. In reality, AI is already
changing our daily lives, almost entirely in ways that improve human health, safety,
and productivity. Unlike in the movies, there is no race of superhuman robots on the
horizon or probably even possible. And while the potential to abuse AI technologies
must be acknowledged and addressed, their greater potential is, among other things,
Many have already grown to make driving safer, help children learn, and extend and enhance people’s lives. In
fact, beneficial AI applications in schools, homes, and hospitals are already growing
accustomed to touching at an accelerated pace. Major research universities devote departments to AI studies,
and technology companies such as Apple, Facebook, Google, IBM, and Microsoft
and talking to their
spend heavily to explore AI applications they regard as critical to their futures. Even
smart phones. People’s Hollywood uses AI technologies to bring its dystopian AI fantasies to the screen.
Innovations relying on computer-based vision, speech recognition, and Natural
future relationships with Language Processing have driven these changes, as have concurrent scientific and
technological advances in related fields. AI is also changing how people interact with
machines will become ever technology. Many people have already grown accustomed to touching and talking to
more nuanced, fluid, and their smart phones. People’s future relationships with machines will become ever more
nuanced, fluid, and personalized as AI systems learn to adapt to individual personalities
personalized. and goals. These AI applications will help monitor people’s well-being, alert them to
risks ahead, and deliver services when needed or wanted. For example, in a mere
fifteen years in a typical North American city—the time frame and scope of this report—
AI applications are likely to transform transportation toward self-driving vehicles with
on-time pickup and delivery of people and packages. This alone will reconfigure the
urban landscape, as traffic jams and parking challenges become obsolete.
This study’s focus on a typical North American city is deliberate and meant to
highlight specific changes affecting the everyday lives of the millions of people who
inhabit them. The Study Panel further narrowed its inquiry to eight domains where
AI is already having or is projected to have the greatest impact: transportation,
healthcare, education, low-resource communities, public safety and security,
employment and workplace, home/service robots, and entertainment.
Though drawing from a common source of research, AI technologies have
influenced and will continue to influence these domains differently. Each domain
faces varied AI-related challenges, including the difficulty of creating safe and reliable
hardware for sensing and effecting (transportation and service robots), the difficulty
of smoothly interacting with human experts (healthcare and education), the challenge
of gaining public trust (low-resource communities and public safety and security), the
challenge of overcoming fears of marginalizing humans (employment and workplace)
and the risk of diminishing interpersonal interaction (entertainment). Some domains
are primarily business sectors, such as transportation and healthcare, while others are
more oriented to consumers, such as entertainment and home service robots. Some
cut across sectors, such as employment/workplace and low-resource communities.
In each domain, even as AI continues to deliver important benefits, it also raises
important ethical and social issues, including privacy concerns. Robots and other AI
technologies have already begun to displace jobs in some sectors. As a society, we are now
at a crucial juncture in determining how to deploy AI-based technologies in ways that
promote, not hinder, democratic values such as freedom, equality, and transparency. For
individuals, the quality of the lives we lead and how our contributions are valued are
likely to shift gradually, but markedly. Over the next several years, AI research, systems
development, and social and regulatory frameworks will shape how the benefits of AI
6 are weighed against its costs and risks, and how broadly these benefits are spread.
 An accurate and sophisticated picture of AI—one that competes with its popular
portrayal—is hampered at the start by the difficulty of pinning down a precise
definition of artificial intelligence. In the approaches the Study Panel considered,
none suggest there is currently a “general purpose” AI. While drawing on common
research and technologies, AI systems are specialized to accomplish particular
tasks, and each application requires years of focused research and a careful, unique
construction. As a result, progress is uneven within and among the eight domains.
A prime example is Transportation, where a few key technologies have catalyzed
the widespread adoption of AI with astonishing speed. Autonomous transportation Society is now at a crucial
will soon be commonplace and, as most people’s first experience with physically
embodied AI systems, will strongly influence the public’s perception of AI. As cars juncture in determining
become better drivers than people, city-dwellers will own fewer cars, live further from
how to deploy AI-based
work, and spend time differently, leading to an entirely new urban organization. In
the typical North American city in 2030, physically embodied AI applications will technologies in ways that
not be limited to cars, but are likely to include trucks, flying vehicles, and personal
robots. Improvements in safe and reliable hardware will spur innovation over the next promote rather than hinder
fifteen years, as they will with Home/Service Robots, which have already entered
people’s houses, primarily in the form of vacuum cleaners. Better chips, low-cost 3D
democratic values such
sensors, cloud-based machine learning, and advances in speech understanding will as freedom, equality, and
enhance future robots’ services and their interactions with people. Special purpose
robots will deliver packages, clean offices, and enhance security. But technical transparency.
constraints and the high costs of reliable mechanical devices will continue to limit
commercial opportunities to narrowly defined applications for the foreseeable future.
In Healthcare, there has been an immense forward leap in collecting useful data
from personal monitoring devices and mobile apps, from electronic health records (EHR)
in clinical settings and, to a lesser extent, from surgical robots designed to assist with
medical procedures and service robots supporting hospital operations. AI-based
applications could improve health outcomes and the quality of life for millions of
people in the coming years. Though clinical applications have been slow to move from
the computer science lab to the real-world, there are hopeful signs that the pace of
innovation will improve. Advances in healthcare can be promoted via the development
of incentives and mechanisms for sharing data and for removing overbearing policy,
regulatory, and commercial obstacles. For many applications, AI systems will have to work
closely with care providers and patients to gain their trust. Advances in how intelligent
machines interact naturally with caregivers, patients, and patients’ families are crucial.
Enabling more fluid interactions between people and promising AI technologies
also remains a critical challenge in Education, which has seen considerable progress
in the same period. Though quality education will always require active engagement
by human teachers, AI promises to enhance education at all levels, especially by
providing personalization at scale. Interactive machine tutors are now being matched
to students for teaching science, math, language, and other disciplines. Natural
Language Processing, machine learning, and crowdsourcing have boosted online
learning and enabled teachers in higher education to multiply the size of their
classrooms while addressing individual students’ learning needs and styles. Over the
next fifteen years in a typical North American city, the use of these technologies in
the classroom and in the home is likely to expand significantly, provided they can be
meaningfully integrated with face-to-face learning.
Beyond education, many opportunities exist for AI methods to assist
Low-resource Communities by providing mitigations and solutions to a
variety of social problems. Traditionally, funders have underinvested in AI research
lacking commercial application. With targeted incentives and funding priorities,
7
 AI technologies could help address the needs of low-resource communities, and
budding efforts are promising. Using data mining and machine learning, for example,
AI has been used to create predictive models to help government agencies address
issues such as prevention of lead poisoning in at-risk children and distribution of
food efficiently. These budding efforts suggest more could be done, particularly if
agencies and organizations can engage and build trust with these communities.
Gaining public trust is also a challenge for AI use by Public Safety and Security
professionals. North American cities and federal agencies have already begun to
Longer term, AI may be deploy AI technologies in border administration and law enforcement. By 2030,
they will rely heavily upon them, including improved cameras and drones for
thought of as a radically surveillance, algorithms to detect financial fraud, and predictive policing. The latter
raises the specter of innocent people being unjustifiably monitored, and care must be
different mechanism
taken to avoid systematizing human bias and to protect civil liberties. Well-deployed
for wealth creation in AI prediction tools have the potential to provide new kinds of transparency about
data and inferences, and may be applied to detect, remove, or reduce human bias,
which everyone should rather than reinforcing it.
Social and political decisions are likewise at play in AI’s influences on
be entitled to a portion of Employment and Workplace trends, such as the safety nets needed to protect
the world’s AI-produced people from structural changes in the economy. AI is poised to replace people in
certain kinds of jobs, such as in the driving of taxis and trucks. However, in many
treasures. realms, AI will likely replace tasks rather than jobs in the near term, and will also
create new kinds of jobs. But the new jobs that will emerge are harder to imagine in
advance than the existing jobs that will likely be lost. AI will also lower the cost of
many goods and services, effectively making everyone better off. Longer term, AI may
be thought of as a radically different mechanism for wealth creation in which everyone
should be entitled to a portion of the world’s AI-produced treasures. It is not too soon
for social debate on how the economic fruits of AI technologies should be shared.
Entertainment has been transformed by social networks and other platforms for
sharing and browsing blogs, videos, and photos, which rely on techniques actively
developed in NLP, information retrieval, image processing, crowdsourcing, and machine
learning. Some traditional sources of entertainment have also embraced AI to compose
music, create stage performances, and even to generate 3D scenes from natural
language text. The enthusiasm with which people have already responded to AI-driven
entertainment has been surprising. As with many aspects of AI, there is ongoing debate
about the extent to which the technology replaces or enhances sociability. AI will
increasingly enable entertainment that is more interactive, personalized, and engaging.
Research should be directed toward understanding how to leverage these attributes
for individuals’ and society’s benefit.

What’s next for AI research?

The research that fuels the AI revolution has also seen rapid changes. Foremost
among them is the maturation of machine learning, stimulated in part by the rise of
the digital economy, which both provides and leverages large amounts of data. Other
factors include the rise of cloud computing resources and consumer demand for
widespread access to services such as speech recognition and navigation support.
Machine learning has been propelled dramatically forward by impressive
empirical successes of artificial neural networks, which can now be trained with huge
data sets and large-scale computing. This approach has been come to be known as
“deep learning.” The leap in the performance of information processing algorithms
has been accompanied by significant progress in hardware technology for basic
operations such as sensing, perception, and object recognition. New platforms and
8
markets for data-driven products, and the economic incentives to find new products
and markets, have also stimulated research advances. Now, as it becomes a central
force in society, the field of AI is shifting toward building intelligent systems that
can collaborate effectively with people, and that are more generally human-aware,
including creative ways to develop interactive and scalable ways for people to
teach robots. These trends drive the currently “hot” areas of AI research into both
fundamental methods and application areas:
Large-scale machine learning concerns the design of learning algorithms, as
well as scaling existing algorithms, to work with extremely large data sets. The field of AI is shifting
Deep learning, a class of learning procedures, has facilitated object recognition
toward building intelligent
in images, video labeling, and activity recognition, and is making significant inroads into
other areas of perception, such as audio, speech, and natural language processing. systems that can
Reinforcement learning is a framework that shifts the focus of machine
learning from pattern recognition to experience-driven sequential decision-making.
collaborate effectively with
It promises to carry AI applications forward toward taking actions in the real world. people, including creative
While largely confined to academia over the past several decades, it is now seeing
some practical, real-world successes. ways to develop interactive
Robotics is currently concerned with how to train a robot to interact with the
and scalable ways for
world around it in generalizable and predictable ways, how to facilitate manipulation
of objects in interactive environments, and how to interact with people. Advances in people to teach robots.
robotics will rely on commensurate advances to improve the reliability and generality
of computer vision and other forms of machine perception.
Computer vision is currently the most prominent form of machine perception.
It has been the sub-area of AI most transformed by the rise of deep learning. For the
first time, computers are able to perform some vision tasks better than people. Much
current research is focused on automatic image and video captioning.
Natural Language Processing, often coupled with automatic speech recognition,
is quickly becoming a commodity for widely spoken languages with large data sets.
Research is now shifting to develop refined and capable systems that are able to
interact with people through dialog, not just react to stylized requests. Great strides
have also been made in machine translation among different languages, with more
real-time person-to-person exchanges on the near horizon.
Collaborative systems research investigates models and algorithms to help develop
autonomous systems that can work collaboratively with other systems and with humans.
Crowdsourcing and human computation research investigates methods to
augment computer systems by making automated calls to human expertise to solve
problems that computers alone cannot solve well.
Algorithmic game theory and computational social choice draw attention
to the economic and social computing dimensions of AI, such as how systems can
handle potentially misaligned incentives, including self-interested human participants
or firms and the automated AI-based agents representing them.
Internet of Things (IoT) research is devoted to the idea that a wide array of
devices, including appliances, vehicles, buildings, and cameras, can be interconnected to
collect and share their abundant sensory information to use for intelligent purposes.
Neuromorphic computing is a set of technologies that seek to mimic
biological neural networks to improve the hardware efficiency and robustness of
computing systems, often replacing an older emphasis on separate modules for input/
output, instruction-processing, and memory.
9
 AI policy, now and in the future
The measure of success for AI applications is the value they create for human lives.
In that light, they should be designed to enable people to understand AI systems
successfully, participate in their use, and build their trust. Public policies should help
ease society’s adaptation to AI applications, extend their benefits, and mitigate their
inevitable errors and failures. Debate about how AI is deployed, including concerns
about how privacy is protected and AI’s benefits fairly shared, should be encouraged.
Given the speed with which AI technologies are being realized, and concomitant
Misunderstandings concerns about their implications, the Study Panel recommends that all layers of
government acquire technical expertise in AI. Further, research on the fairness,
about what AI is and is security, privacy, and societal implications of AI systems should be encouraged by
not could fuel opposition removing impediments and increasing private and public spending to support it.
Currently in the United States, at least sixteen separate agencies govern sectors
to technologies with of the economy related to AI technologies. Rapid advances in AI research and,
especially, its applications require experts in these sectors to develop new concepts
the potential to benefit and metaphors for law and policy. Who is responsible when a self-driven car crashes
or an intelligent medical device fails? How can AI applications be prevented from
everyone. Poorly informed
promulgating racial discrimination or financial cheating? Who should reap the gains
regulation that stifles of efficiencies enabled by AI technologies and what protections should be afforded
to people whose skills are rendered obsolete? As people integrate AI more broadly
innovation would be a and deeply into industrial processes and consumer products, best practices need to be
spread, and regulatory regimes adapted.
tragic mistake. While the Study Panel does not consider it likely that near-term AI systems will
autonomously choose to inflict harm on people, it will be possible for people to use
AI-based systems for harmful as well as helpful purposes. And though AI algorithms
may be capable of making less biased decisions than a typical person, it remains
a deep technical challenge to ensure that the data that inform AI-based decisions
can be kept free from biases that could lead to discrimination based on race, sexual
orientation, or other factors.
Faced with the profound changes that AI technologies can produce, pressure for
“more” and “tougher” regulation is probably inevitable. Misunderstandings about
what AI is and is not could fuel opposition to technologies with the potential to
benefit everyone. Inappropriate regulatory activity would be a tragic mistake. Poorly
informed regulation that stifles innovation, or relocates it to other jurisdictions, would
be counterproductive.2
Fortunately, principles that guide successful regulation of current digital technologies
provide a starting point. In privacy regulation, broad legal mandates coupled with
tough transparency requirements and meaningful enforcement—rather than strict
controls—encourage companies to develop processes and professional staff to
enforce privacy controls, engage with outside stakeholders, and adapt their practices
to technological advances. This in turn supports the development of professional
trade associations and standards committees that spread best practices. In AI, too,
regulators can strengthen a virtuous cycle of activity involving internal and external
accountability, transparency, and professionalization, rather than narrow compliance.
A vigorous and informed debate about how to best steer AI in ways that enrich
our lives and our society, while encouraging creativity in the field, is an urgent and
vital need. AI technologies could widen existing inequalities of opportunity if access
to them—along with the high-powered computation and large-scale data that fuel
many of them—is unfairly distributed across society. These technologies will improve
2 Kate Crawford, “Artificial Intelligence’s White Guy Problem,” The New York Times, June 25,
2016, accessed August 1, 2016, http://www.nytimes.com/2016/06/26/opinion/sunday/artificial-
10 intelligences-white-guy-problem.html.
the abilities and efficiency of people who have access to them. Policies should be
evaluated as to whether they foster democratic values and equitable sharing of AI’s
benefits, or concentrate power and benefits in the hands of a fortunate few.
As this report documents, significant AI-related advances have already had
an impact on North American cities over the past fifteen years, and even more
substantial developments will occur over the next fifteen. Recent advances are largely
due to the growth and analysis of large data sets enabled by the internet, advances
in sensory technologies and, more recently, applications of “deep learning.” In the
coming years, as the public encounters new AI applications in domains such as
transportation and healthcare, they must be introduced in ways that build trust and
understanding, and respect human and civil rights. While encouraging innovation,
policies and processes should address ethical, privacy, and security implications, and
should work to ensure that the benefits of AI technologies will be spread broadly and
fairly. Doing so will be critical if Artificial Intelligence research and its applications
are to exert a positive influence on North American urban life in 2030 and beyond.

11
 SECTION I: WHAT IS ARTIFICIAL
INTELLIGENCE?
This section describes how researchers and practitioners define “Artificial Intelligence,” and
the areas of AI research and application that are currently thriving. It proffers definitions of
what AI is and is not, and describes some of the currently “hot” areas of AI Research. This
section lays the groundwork for Section II, which elaborates on AI’s impacts and future in
eight domains and Section III, which describes issues related to AI design and public policy
An accurate and
and makes recommendations for encouraging AI innovation while protecting democratic values.
sophisticated picture of
DEFINING AI
AI—one that competes with
Curiously, the lack of a precise, universally accepted definition of AI probably
its popular portrayal—is has helped the field to grow, blossom, and advance at an ever-accelerating pace.
Practitioners, researchers, and developers of AI are instead guided by a rough
hampered by the difficulty sense of direction and an imperative to “get on with it.” Still, a definition remains
important and Nils J. Nilsson has provided a useful one:
of pinning down a precise
“Artificial intelligence is that activity devoted to making machines intelligent, and
definition of artificial intelligence is that quality that enables an entity to function appropriately and with
foresight in its environment.”3
intelligence. From this perspective, characterizing AI depends on the credit one is willing to
give synthesized software and hardware for functioning “appropriately” and with
“foresight.” A simple electronic calculator performs calculations much faster than
the human brain, and almost never makes a mistake.4 Is a calculator intelligent?
Like Nilsson, the Study Panel takes a broad view that intelligence lies on a multi-
dimensional spectrum. According to this view, the difference between an arithmetic
calculator and a human brain is not one of kind, but of scale, speed, degree of
autonomy, and generality. The same factors can be used to evaluate every other
instance of intelligence—speech recognition software, animal brains, cruise-control
systems in cars, Go-playing programs, thermostats—and to place them at some
appropriate location in the spectrum.
Although our broad interpretation places the calculator within the intelligence
spectrum, such simple devices bear little resemblance to today’s AI. The frontier of
AI has moved far ahead and functions of the calculator are only one among the
millions that today’s smartphones can perform. AI developers now work on improving,
generalizing, and scaling up the intelligence currently found on smartphones.
In fact, the field of AI is a continual endeavor to push forward the frontier of
machine intelligence. Ironically, AI suffers the perennial fate of losing claim to its
acquisitions, which eventually and inevitably get pulled inside the frontier, a repeating
pattern known as the “AI effect” or the “odd paradox”—AI brings a new technology
into the common fold, people become accustomed to this technology, it stops being
considered AI, and newer technology emerges.5 The same pattern will continue in the
future. AI does not “deliver” a life-changing product as a bolt from the blue. Rather,
AI technologies continue to get better in a continual, incremental way.

3 Nils J. Nilsson, The Quest for Artificial Intelligence: A History of Ideas and Achievements (Cambridge,
UK: Cambridge University Press, 2010).
4 Wikimedia Images, accessed August 1, 2016, https://upload.wikimedia.org/wikipedia/
commons/b/b6/SHARP_ELSIMATE_EL-W221.jpg.
5 Pamela McCorduck, Machines Who Think: A Personal Inquiry into the History and Prospects of
Artificial Intelligence, 2nd ed. (Natick, MA: A. K. Peters, Ltd., 2004; San Francisco: W. H. Freeman,
12 1979), Citations are to the Peters edition.
The human measure
Notably, the characterization of intelligence as a spectrum grants no special status
to the human brain. But to date human intelligence has no match in the biological
and artificial worlds for sheer versatility, with the abilities “to reason, achieve
goals, understand and generate language, perceive and respond to sensory inputs,
prove mathematical theorems, play challenging games, synthesize and summarize
information, create art and music, and even write histories.”6
This makes human intelligence a natural choice for benchmarking the progress
of AI. It may even be proposed, as a rule of thumb, that any activity computers Intelligence lies on a
are able to perform and people once performed should be counted as an instance
of intelligence. But matching any human ability is only a sufficient condition, not a
multi-dimensional
necessary one. There are already many systems that exceed human intelligence, at spectrum. According to
least in speed, such as scheduling the daily arrivals and departures of thousands of
flights in an airport. this view, the difference
AI’s long quest—and eventual success—to beat human players at the game of
chess offered a high-profile instance for comparing human to machine intelligence. between an arithmetic
Chess has fascinated people for centuries. When the possibility of building computers
calculator and a human
became imminent, Alan Turing, who many consider the father of computer science,
“mentioned the idea of computers showing intelligence with chess as a paradigm.”7 brain is not one of kind,
Without access to powerful computers, “Turing played a game in which he simulated
the computer, taking about half an hour per move.” but of scale, speed,
But it was only after a long line of improvements in the sixties and seventies—
contributed by groups at Carnegie Mellon, Stanford, MIT, The Institute for
degree of autonomy, and
Theoretical and Experimental Physics at Moscow, and Northwestern University— generality.
that chess-playing programs started gaining proficiency. The final push came
through a long-running project at IBM, which culminated with the Deep Blue
program beating Garry Kasparov, then the world chess champion, by a score of
3.5-2.5 in 1997. Curiously, no sooner had AI caught up with its elusive target than
Deep Blue was portrayed as a collection of “brute force methods” that wasn’t “real
intelligence.”8 In fact, IBM’s subsequent publication about Deep Blue, which gives
extensive details about its search and evaluation procedures, doesn’t mention the
word “intelligent” even once!9 Was Deep Blue intelligent or not? Once again, the
frontier had moved.

An operational definition
AI can also be defined by what AI researchers do. This report views AI
primarily as a branch of computer science that studies the properties of
intelligence by synthesizing intelligence.10 Though the advent of AI has depended
on the rapid progress of hardware computing resources, the focus here on
software reflects a trend in the AI community. More recently, though, progress in
building hardware tailored for neural-network-based computing11 has created a

6 Nilsson, The Quest for Artificial Intelligence.

7 Nilsson, The Quest for Artificial Intelligence, 89.
8 McCorduck, Machines Who Think, 433.
9 Murray Campbell, A. Joseph Hoane Jr., and Feng-hsiung Hsu, “Deep Blue,” Artificial
Intelligence 134, nos. 1 and 2 (2002): 57–83.
10 Herbert A. Simon, “Artificial Intelligence: An Empirical Science,” Artificial Intelligence 77, no. 2
(1995):95–127.
11 Paul Merolla John V. Arthur, Rodrigo Alvarez-Icaza, Andrew S. Cassidy, Jun Sawada, Filipp
Akopyan, Bryan L. Jackson, Nabil Imam, Chen Guo, Yutaka Nakamura, Bernard Brezzo, Ivan Vo,
Steven K. Esser, Rathinakumar Appuswamy, Brian Taba, Arnon Amir, Myron D. Flickner, William
P. Risk, Rajit Manohar, and Dharmendra S. Modha, “A Million Spiking-Neuron Integrated
Circuit with a Scalable Communication Network and Interface,” accessed August 1, 2016, http://
paulmerolla.com/merolla_main_som.pdf. 13
 tighter coupling between hardware and software in advancing AI.
“Intelligence” remains a complex phenomenon whose varied aspects have attracted
the attention of several different fields of study, including psychology, economics,
neuroscience, biology, engineering, statistics, and linguistics. Naturally, the field of AI has
benefited from the progress made by all of these allied fields. For example, the artificial
neural network, which has been at the heart of several AI-based solutions12 13 was
originally inspired by thoughts about the flow of information in biological neurons.14

Human intelligence has AI RESEARCH TRENDS

Until the turn of the millennium, AI’s appeal lay largely in its promise to deliver, but
no match in the biological
in the last fifteen years, much of that promise has been redeemed.15 AI technologies
and artificial worlds for already pervade our lives. As they becomes a central force in society, the field is
shifting from simply building systems that are intelligent to building intelligent systems
sheer versatility, with that are human-aware and trustworthy.
Several factors have fueled the AI revolution. Foremost among them is the
the abilities “to reason, maturing of machine learning, supported in part by cloud computing resources
achieve goals, understand and wide-spread, web-based data gathering. Machine learning has been propelled
dramatically forward by “deep learning,” a form of adaptive artificial neural
and generate language... networks trained using a method called backpropagation.16 This leap in the
performance of information processing algorithms has been accompanied by
create art and music, and significant progress in hardware technology for basic operations such as sensing,
perception, and object recognition. New platforms and markets for data-driven
even write histories.”
products, and the economic incentives to find new products and markets, have also
contributed to the advent of AI-driven technology.
All these trends drive the “hot” areas of research described below. This compilation
is meant simply to reflect the areas that, by one metric or another, currently receive
greater attention than others. They are not necessarily more important or valuable
than other ones. Indeed, some of the currently “hot” areas were less popular in past
years, and it is likely that other areas will similarly re-emerge in the future.
Large-scale machine learning
Many of the basic problems in machine learning (such as supervised and
unsupervised learning) are well-understood. A major focus of current efforts is to
scale existing algorithms to work with extremely large data sets. For example, whereas
traditional methods could afford to make several passes over the data set, modern
ones are designed to make only a single pass; in some cases, only sublinear methods
(those that only look at a fraction of the data) can be admitted.
Deep learning
The ability to successfully train convolutional neural networks has most benefited the
field of computer vision, with applications such as object recognition, video

12 Gerald Tesauro, “Practical Issues in Temporal Difference Learning,” Machine Learning, no. 8
(1992): 257–77.
13 David Silver, Aja Huang, Chris J. Maddison, Arthur Guez, Laurent Sifre, George van den
Driessche, Julian Schrittwieser, Ioannis Antonoglou, Veda Panneershelvam, Marc Lanctot, Sander
Dieleman, Dominik Grewe, John Nham, Nal Kalchbrenner, Ilya Sutskever, Timothy Lillicrap,
Madeleine Leach, Koray Kavukcuoglu, Thore Graepel, and Demis Hassabis, “Mastering the game
of Go with deep neural networks and tree search,” Nature 529 (2016): 484–489.
14 W. McCulloch and W. Pitts, W., “A logical calculus of the ideas immanent in nervous activity,”
Bulletin of Mathematical Biophysics, 5 (1943): 115–133.
15 Appendix I offers a short history of AI, including a description of some of the traditionally
core areas of research, which have shifted over the past six decades.
16 Backpropagation is an abbreviation for “backward propagation of errors,” a common method
of training artificial neural networks used in conjunction with an optimization method such as
gradient descent. The method calculates the gradient of a loss function with respect to all the
14 weights in the network.
labeling, activity recognition, and several variants thereof. Deep learning is also making
significant inroads into other areas of perception, such as audio, speech, and natural
language processing.
Reinforcement learning
Whereas traditional machine learning has mostly focused on pattern mining,
reinforcement learning shifts the focus to decision making, and is a technology that
will help AI to advance more deeply into the realm of learning about and executing
actions in the real world. It has existed for several decades as a framework for
experience-driven sequential decision-making, but the methods have not found great AI technologies already
success in practice, mainly owing to issues of representation and scaling. However,
the advent of deep learning has provided reinforcement learning with a “shot in the pervade our lives. As
arm.” The recent success of AlphaGo, a computer program developed by Google
Deepmind that beat the human Go champion in a five-game match, was due in large they become a central
part to reinforcement learning. AlphaGo was trained by initializing an automated
force in society, the
agent with a human expert database, but was subsequently refined by playing a large
number of games against itself and applying reinforcement learning. field is shifting from
Robotics
Robotic navigation, at least in static environments, is largely solved. Current efforts simply building systems
consider how to train a robot to interact with the world around it in generalizable
that are intelligent to
and predictable ways. A natural requirement that arises in interactive environments
is manipulation, another topic of current interest. The deep learning revolution is building intelligent
only beginning to influence robotics, in large part because it is far more difficult to
acquire the large labeled data sets that have driven other learning-based areas of AI. systems that are
Reinforcement learning (see above), which obviates the requirement of labeled data,
may help bridge this gap but requires systems to be able to safely explore a policy human-aware and
space without committing errors that harm the system itself or others. Advances in
trustworthy.
reliable machine perception, including computer vision, force, and tactile perception,
much of which will be driven by machine learning, will continue to be key enablers to
advancing the capabilities of robotics.
Computer vision
Computer vision is currently the most prominent form of machine perception. It has
been the sub-area of AI most transformed by the rise of deep learning. Until just a
few years ago, support vector machines were the method of choice for most visual
classification tasks. But the confluence of large-scale computing, especially on GPUs,
the availability of large datasets, especially via the internet, and refinements of neural
network algorithms has led to dramatic improvements in performance on benchmark
tasks (e.g., classification on ImageNet17). For the first time, computers are able to
perform some (narrowly defined) visual classification tasks better than people. Much
current research is focused on automatic image and video captioning.
Natural Language Processing
Often coupled with automatic speech recognition, Natural Language Processing is
another very active area of machine perception. It is quickly becoming a commodity
for mainstream languages with large data sets. Google announced that 20% of
current mobile queries are done by voice,18 and recent demonstrations have proven
the possibility of real-time translation. Research is now shifting towards developing
refined and capable systems that are able to interact with people through dialog, not
just react to stylized requests.

17 ImageNet, Stanford Vision Lab, Stanford University, Princeton University, 2016, accessed
August 1, 2016, www.image-net.org/.
18 Greg Sterling, “Google says 20% of mobile queries are voice searches,” Search Engine Land,
May 18, 2016, accessed August 1, 2016, http://searchengineland.com/google-reveals-20-percent-
queries-voice-queries-249917. 15
 Collaborative systems
Research on collaborative systems investigates models and algorithms to help
develop autonomous systems that can work collaboratively with other systems and
with humans. This research relies on developing formal models of collaboration,
and studies the capabilities needed for systems to become effective partners. There
is growing interest in applications that can utilize the complementary strengths of
humans and machines—for humans to help AI systems to overcome their limitations,
and for agents to augment human abilities and activities.
Natural Language Crowdsourcing and human computation
Since human abilities are superior to automated methods for accomplishing many
Processing is a very tasks, research on crowdsourcing and human computation investigates methods to
augment computer systems by utilizing human intelligence to solve problems that
active area of machine
computers alone cannot solve well. Introduced only about fifteen years ago, this
perception. Research research now has an established presence in AI. The best-known example of
crowdsourcing is Wikipedia, a knowledge repository that is maintained and updated
is now shifting towards by netizens and that far exceeds traditionally-compiled information sources, such
as encyclopedias and dictionaries, in scale and depth. Crowdsourcing focuses on
developing systems that devising innovative ways to harness human intelligence. Citizen science platforms
are able to interact with energize volunteers to solve scientific problems, while paid crowdsourcing platforms
such as Amazon Mechanical Turk provide automated access to human intelligence on
people through dialog, demand. Work in this area has facilitated advances in other subfields of AI, including
computer vision and NLP, by enabling large amounts of labeled training data and/or
not just react to stylized human interaction data to be collected in a short amount of time. Current research
efforts explore ideal divisions of tasks between humans and machines based on their
requests.
differing capabilities and costs.
Algorithmic game theory and computational social choice
New attention is being drawn to the economic and social computing dimensions of
AI, including incentive structures. Distributed AI and multi-agent systems have been
studied since the early 1980s, gained prominence starting in the late 1990s, and were
accelerated by the internet. A natural requirement is that systems handle potentially
misaligned incentives, including self-interested human participants or firms, as well
as automated AI-based agents representing them. Topics receiving attention include
computational mechanism design (an economic theory of incentive design, seeking
incentive-compatible systems where inputs are truthfully reported), computational
social choice (a theory for how to aggregate rank orders on alternatives), incentive
aligned information elicitation (prediction markets, scoring rules, peer prediction) and
algorithmic game theory (the equilibria of markets, network games, and parlor games
such as Poker—a game where significant advances have been made in recent years
through abstraction techniques and no-regret learning).
Internet of Things (IoT)
A growing body of research is devoted to the idea that a wide array of devices can
be interconnected to collect and share their sensory information. Such devices can
include appliances, vehicles, buildings, cameras, and other things. While it’s a matter
of technology and wireless networking to connect the devices, AI can process and
use the resulting huge amounts of data for intelligent and useful purposes. Currently,
these devices use a bewildering array of incompatible communication protocols. AI
could help tame this Tower of Babel.
Neuromorphic Computing
Traditional computers implement the von Neumann model of computing, which
separates the modules for input/output, instruction-processing, and memory. With
the success of deep neural networks on a wide array of tasks, manufacturers are
16
actively pursuing alternative models of computing—especially those that are inspired
by what is known about biological neural networks—with the aim of improving the
hardware efficiency and robustness of computing systems. At the moment, such
“neuromorphic” computers have not yet clearly demonstrated big wins, and are just
beginning to become commercially viable. But it is possible that they will become
commonplace (even if only as additions to their von Neumann cousins) in the
near future. Deep neural networks have already created a splash in the application
landscape. A larger wave may hit when these networks can be trained and executed
on dedicated neuromorphic hardware, as opposed to simulated on standard von A growing body of research
Neumann architectures, as they are today.
is devoted to the idea that
Overall trends and the future of AI research
a wide array of devices
The resounding success of the data-driven paradigm has displaced the traditional
paradigms of AI. Procedures such as theorem proving and logic-based knowledge can be interconnected
representation and reasoning are receiving reduced attention, in part because of the
ongoing challenge of connecting with real-world groundings. Planning, which was a to collect and share their
mainstay of AI research in the seventies and eighties, has also received less attention
of late due in part to its strong reliance on modeling assumptions that are hard
sensory information.
to satisfy in realistic applications. Model-based approaches—such as physics-based Such devices can include
approaches to vision and traditional control and mapping in robotics—have by and
large given way to data-driven approaches that close the loop with sensing the results appliances, vehicles,
of actions in the task at hand. Bayesian reasoning and graphical models, which were
very popular even quite recently, also appear to be going out of favor, having been buildings, cameras, and
drowned by the deluge of data and the remarkable success of deep learning.
other things.
Over the next fifteen years, the Study Panel expects an increasing focus on
developing systems that are human-aware, meaning that they specifically model,
and are specifically designed for, the characteristics of the people with whom they
are meant to interact. There is a lot of interest in trying to find new, creative ways
to develop interactive and scalable ways to teach robots. Also, IoT-type systems—
devices and the cloud—are becoming increasingly popular, as is thinking about
social and economic dimensions of AI. In the coming years, new perception/object
recognition capabilities and robotic platforms that are human-safe will grow, as will
data-driven products and their markets.
The Study Panel also expects a reemergence of some of the traditional forms of
AI as practitioners come to realize the inevitable limitations of purely end-to-end deep
learning approaches. We encourage young researchers not to reinvent the wheel,
but rather to maintain an awareness of the significant progress in many areas of
AI during the first fifty years of the field, and in related fields such as control theory,
cognitive science, and psychology.

17
 SECTION II: AI BY DOMAIN
Though different instances of AI research and practice share common technologies, such
as machine learning, they also vary considerably in different sectors of the economy and
society. We call these sectors “domains,” and in this section describe the different states
of AI research and implementation, as well as impacts and distinct challenges, in eight of
them: transportation; home/service robotics; healthcare; education; low-resource communities;
public safety and security; employment and workplace; and entertainment. Based on these
Autonomous analyses, we also predict trends in a typical North American city over the next fifteen years.
Contrary to AI’s typical depiction in popular culture, we seek to offer a balanced overview
transportation will soon of the ways in which AI is already beginning to transform everyday life, and how those
transformations are likely to grow by the year 2030.
be commonplace and,
as most people’s first TRANSPORTATION
experience with physically Transportation is likely to be one of the first domains in which the general public
will be asked to trust the reliability and safety of an AI system for a critical task.
embodied AI systems, Autonomous transportation will soon be commonplace and, as most people’s first
experience with physically embodied AI systems, will strongly influence the public’s
will strongly influence the perception of AI. Once the physical hardware is made sufficiently safe and robust, its
introduction to daily life may happen so suddenly as to surprise the public, which will
public’s perception of AI.
require time to adjust. As cars will become better drivers than people, city-dwellers
will own fewer cars, live further from work, and spend time differently, leading to an
entirely new urban organization. Further, in the typical North American city in 2030,
changes won’t be limited to cars and trucks, but are likely to include flying vehicles
and personal robots, and will raise social, ethical and policy issues.
A few key technologies have already catalyzed the widespread adoption of AI
in transportation. Compared to 2000, the scale and diversity of data about personal
and population-level transportation available today—enabled by the adoption of
smartphones and decreased costs and improved accuracies for variety of sensors—is
astounding. Without the availability of this data and connectivity, applications such as
real-time sensing and prediction of traffic, route calculations, peer-to-peer ridesharing
and self-driving cars would not be possible.

Smarter cars
GPS was introduced to personal vehicles in 2001 with in-car navigation devices and
has since become a fundamental part of the transportation infrastructure.19 GPS
assists drivers while providing large-scale information to technology companies and
cities about transportation patterns. Widespread adoption of smartphones with GPS
technology further increased connectivity and the amount of location data shared by
individuals.
Current vehicles are also equipped with a wide range of sensing capabilities.
An average automobile in the US is predicted to have seventy sensors including
gyroscopes, accelerometers, ambient light sensors, and moisture sensors.20 Sensors
are not new to vehicles. Automobiles built before 2000 had sensors for the
internal state of the vehicle such as its speed, acceleration, and wheel position.21

19 Mark Sullivan, “A brief history of GPS,” PCWorld, August 9, 2012, accessed August 1, 2016,
http://www.pcworld.com/article/2000276/a-brief-history-of-gps.html.
20 William J. Fleming, “New Automotive Sensors - A Review,” IEEE Sensors Journal 8, no 11,
(2008): 1900-1921.
21 Jean Jacques Meneu, ed., “Automotive Sensors: Now and in the Future,” Arrow, September 24,
2015, accessed August 1, 2016, https://www.arrow.com/en/research-and-events/articles/automotive-
18 sensors-now-and-in-the-future.
They already had a number of functionalities that combined real-time sensing with
perception and decision-making such as Anti-lock Braking Systems (ABS), airbag
control, Traction Control Systems (TCS), and Electronic Stability Control (ESC).22
Automated capabilities have been introduced into commercial cars gradually since
2003 as summarized in the following table.

Context Automated Functionality Release Date

Parking Intelligent Parking Assist System Since 200323 As cars will become better

Parking Summon Since 201624 drivers than people, city-

dwellers will own fewer
Arterial & Highway Lane departure system Since 2004 in North America25
cars, live further from
Arterial & Highway Adaptive cruise control Since 2005 in North America26
work, and spend time
Highway Blind spot monitoring 200727
differently, leading to
Highway Lane changing 201528 an entirely new urban
These functionalities assist drivers or completely take over well-defined activities organization.
for increased safety and comfort. Current cars can park themselves, perform adaptive
cruise control on highways, steer themselves during stop-and-go traffic, and alert
drivers about objects in blind spots during lane changes. Vision and radar technology
were leveraged to develop pre-collision systems that let cars autonomously brake
when risk of a collision is detected. Deep learning also has been applied to improve
automobiles’ capacity to detect objects in the environment and recognize sound.29

Self-driving vehicles
Since the 1930s, science fiction writers dreamed of a future with self-driving cars,
and building them has been a challenge for the AI community since the 1960s. By
the 2000s, the dream of autonomous vehicles became a reality in the sea and sky, and
even on Mars, but self-driving cars existed only as research prototypes in labs. Driving
in a city was considered to be a problem too complex for automation due to factors
like pedestrians, heavy traffic, and the many unexpected events that can happen
outside of the car’s control. Although the technological components required to

22 Carl Liersch, “Vehicle Technology Timeline: From Automated to Driverless,” Robert Bosch
(Australia) Pty. Ltd., 2014, accessed August 1, 2016, http://dpti.sa.gov.au/__data/assets/pdf_
file/0009/246807/Carl_Liersch_Presentation.pdf.
23 “Intelligent Parking Assist System,” Wikipedia, last modified July 26, 2016, accessed August 1,
2016, https://en.wikipedia.org/wiki/Intelligent_Parking_Assist_System.
24 The Tesla Motors Team, “Summon Your Tesla from Your Phone,” Tesla, January 10, 2016,
accessed August 1, 2016, https://www.teslamotors.com/blog/summon-your-tesla-your-phone.
25 Lane departure warning system,” Wikipedia, last modified July 24, 2016, accessed August 1,
2016, https://en.wikipedia.org/wiki/Lane_departure_warning_system.
26 “Autonomous cruise control system,” Wikipedia, last modified July 30, 2016, accessed August 1,
2016, https://en.wikipedia.org/wiki/Autonomous_cruise_control_system.
27 “Blind spot monitor,” Wikipedia, last modified April 20, 2016, accessed August 1, 2016,
https://en.wikipedia.org/wiki/Blind_spot_monitor.
28 Dana Hull, “Tesla Starts Rolling Out Autopilot Features,” Boomberg Technology, October 14,
2015, accessed August 1, 2016, http://www.bloomberg.com/news/articles/2015-10-14/tesla-
software-upgrade-adds-automated-lane-changing-to-model-s.
29 Aaron Tilley, “New Qualcomm Chip Brings Deep Learning To Cars,” Forbes, January 5, 2016,
accessed August 1, 2016, http://www.forbes.com/sites/aarontilley/2016/01/05/along-with-nvidia-
new-qualcomm-chip-brings-deep-learning-to-cars/#4cb4e9235357. 19
 make such autonomous driving possible were available in 2000—and indeed some
autonomous car prototypes existed30 31 32—few predicted that mainstream companies
would be developing and deploying autonomous cars by 2015. During the first
Defense Advanced Research Projects Agency (DARPA) “grand challenge” on
autonomous driving in 2004, research teams failed to complete the challenge in a
limited desert setting.
But in eight short years, from 2004-2012, speedy and surprising progress occurred
in both academia and industry. Advances in sensing technology and machine learning
We will see self-driving and for perception tasks has sped progress and, as a result, Google’s autonomous vehicles
and Tesla’s semi-autonomous cars are driving on city streets today. Google’s self-
remotely controlled delivery driving cars, which have logged more than 1,500,000 miles (300,000 miles without an
accident),33 are completely autonomous—no human input needed. Tesla has widely
vehicles, flying vehicles,
released self-driving capability to existing cars with a software update.34 Their cars are
and trucks. Peer-to-peer semi-autonomous, with human drivers expected to stay engaged and take over if they
detect a potential problem. It is not yet clear whether this semi-autonomous approach
transportation services is sustainable, since as people become more confident in the cars’ capabilities, they
are likely to pay less attention to the road, and become less reliable when they are
such as ridesharing are most needed. The first traffic fatality involving an autonomous car, which occurred in
also likely to utilize self- June of 2016, brought this question into sharper focus.35
In the near future, sensing algorithms will achieve super-human performance for
driving vehicles. capabilities required for driving. Automated perception, including vision, is already
near or at human-level performance for well-defined tasks such as recognition and
tracking. Advances in perception will be followed by algorithmic improvements
in higher level reasoning capabilities such as planning. A recent report predicts
self-driving cars to be widely adopted by 2020.36 And the adoption of self-driving
capabilities won’t be limited to personal transportation. We will see self-driving and
remotely controlled delivery vehicles, flying vehicles, and trucks. Peer-to-peer
transportation services such as ridesharing are also likely to utilize self-driving vehicles.
Beyond self-driving cars, advances in robotics will facilitate the creation and adoption
of other types of autonomous vehicles, including robots and drones.
It is not yet clear how much better self-driving cars need to become to encourage
broad acceptance. The collaboration required in semi-self-driving cars and its
implications for the cognitive load of human drivers is not well understood. But
if future self-driving cars are adopted with the predicted speed, and they exceed
human-level performance in driving, other significant societal changes will follow.
Self-driving cars will eliminate one of the biggest causes of accidental death and
injury in United States, and lengthen people’s life expectancy. On average, a

30 “Navlab,” Wikipedia, last updated June 4, 2016, accessed August 1, 2016, https://en.wikipedia.
org/wiki/Navlab.
31 “Navlab: The Carnegie Mellon University Navigation Laboratory,” Carnegie Mellon
University, accessed August 1, 2016, http://www.cs.cmu.edu/afs/cs/project/alv/www/.
32 “Eureka Prometheus Project,” Wikipedia, last modified February 12, 2016, accessed August 1,
2016, https://en.wikipedia.org/wiki/Eureka_Prometheus_Project.
33 “Google Self-Driving Car Project,” Google, accessed August 1, 2016, https://www.google.
com/selfdrivingcar/. 33 Molly McHugh, “Tesla’s Cars Now Drive Themselves, Kinda,” Wired,
October 14, 2015, accessed August 1, 2016, http://www.wired.com/2015/10/tesla-self-driving-over-
air-update-live/.
34 Molly McHugh, “Tesla’s Cars Now Drive Themselves, Kinda,” Wired, October 14, 2015,
accessed August 1, 2016, http://www.wired.com/2015/10/tesla-self-driving-over-air-update-live/.
35 Anjali Singhvi and Karl Russell, “Inside the Self-Driving Tesla Fatal Accident,” The New
York Times, Last updated July 12, 2016, accessed August 1, 2016, http://www.nytimes.com/
interactive/2016/07/01/business/inside-tesla-accident.html.
36 John Greenough, “10 million self-driving cars will be on the road by 2020,” Business Insider,
June 15, 2016, accessed August 1, 2016, http://www.businessinsider.com/report-10-million-self-
20 driving-cars-will-be-on-the-road-by-2020-2015-5-6.
commuter in US spends twenty-five minutes driving each way.37 With self-driving car
technology, people will have more time to work or entertain themselves during their
commutes. And the increased comfort and decreased cognitive load with self-driving
cars and shared transportation may affect where people choose to live. The reduced
need for parking may affect the way cities and public spaces are designed. Self-driving
cars may also serve to increase the freedom and mobility of different subgroups of
the population, including youth, elderly and disabled.
Self-driving cars and peer-to-peer transportation services may eliminate the need
to own a vehicle. The effect on total car use is hard to predict. Trips of empty vehicles Shared transportation
and people’s increased willingness to travel may lead to more total miles driven.
Alternatively, shared autonomous vehicles—people using cars as a service rather than may displace the need for
owning their own—may reduce total miles, especially if combined with well-constructed
public transportation—
incentives, such as tolls or discounts, to spread out travel demand, share trips, and
reduce congestion. The availability of shared transportation may displace the need or public transportation
for public transportation—or public transportation may change form towards
personal rapid transit, already available in four cities,38 which uses small capacity may change form towards
vehicles to transport people on demand and point-to-point between many stations.39
As autonomous vehicles become more widespread, questions will arise over their
personal rapid transit
security, including how to ensure that technologies are safe and properly tested that uses small capacity
under different road conditions prior to their release. Autonomous vehicles and the
connected transportation infrastructure will create a new venue for hackers to exploit vehicles to transport
vulnerabilities to attack. Ethical questions are also involved in programming cars to
act in situations in which human injury or death is inevitable, especially when there people on demand.
are split-second choices to be made about whom to put at risk. The legal systems in
most states in the US do not have rules covering self-driving cars. As of 2016, four
states in the US (Nevada, Florida, California, and Michigan), Ontario in Canada,
the United Kingdom, France, and Switzerland have passed rules for the testing
of self-driving cars on public roads. Even these laws do not address issues about
responsibility and assignment of blame for an accident for self-driving and semi-self-
driving cars.40

Transportation planning
By 2005, cities had started investing in the transportation infrastructure to develop
sensing capabilities for vehicle and pedestrian traffic.41 The sensors currently used
include inductive loops, video cameras, remote traffic microwave sensors, radars, and
GPS.42 For example, in 2013 New York started using a combination of microwave
sensors, a network of cameras, and pass readers to detect vehicle traffic in the city.43
37 Brian McKenzie and Melanie Rapino, “Commuting in the United States: 2009,” American
Community Survey Reports, United States Census Bureau, September 2011, accessed August 1, 2016,
https://www.census.gov/prod/2011pubs/acs-15.pdf.
38 Morgantown, West Virginia; Masdar City, UAE; London, England; and Suncheon, South
Korea.
39 “Personal rapid transit,” Wikipedia, Last modified July 18, 2016, accessed August 1, 2016,
https://en.wikipedia.org/wiki/Personal_rapid_transit.
40 Patrick Lin, “The Ethics of Autonomous Cars,” The Atlantic, October 8, 2013, accessed
August 1, 2016, http://www.theatlantic.com/technology/archive/2013/10/the-ethics-of-
autonomous-cars/280360/.
41 Steve Lohr, “Bringing Efficiency to the Infrastructure,” The New York Times, April 29,
2009, accessed August 1, 2016, http://www.nytimes.com/2009/04/30/business/energy-
environment/30smart.html.
42 “Intelligent transportation system,” Wikipedia, last modified July 28, 2016, accessed August 1, 2016,
https://en.wikipedia.org/wiki/Intelligent_transportation_system.
43 Access Science Editors, “Active traffic management: adaptive traffic signal control,” Access Science,
2014, accessed August 1, 2016, http://www.accessscience.com/content/active-traffic-management-
adaptive-traffic-signal-control/BR0106141. 21
 Cities use AI methods to optimize services in several ways, such as bus and subway
schedules, and tracking traffic conditions to dynamically adjust speed limits or apply
smart pricing on highways, bridges, and HOV lanes.44 45 46 Using sensors and cameras
in the road network, they can also optimize traffic light timing for improving traffic
flow and to help with automated enforcement.47 48 These dynamic strategies are aimed
at better utilizing the limited resources in the transportation network, and are made
possible by the availability of data and the widespread connectivity of individuals.
Before the 2000s, transportation planners were forced to rely on static pricing
Ethical questions arise strategies tied to particular days or times of day, to manage demand. As dynamic
pricing strategies are adopted, this raises new issues concerning the fair distribution
when programming cars to of public goods, since market conditions in high-demand situations may make
services unavailable to segments of the public.
act in situations in which
The availability of large-scale data has also made transportation an ideal domain
human injury or death is for machine learning applications. Since 2006, applications such as Mapquest, Google
Maps, and Bing Maps have been widely used by the public for routing trips, using public
inevitable, especially when transportation, receiving real-time information and predictions about traffic conditions,
49 50
and finding services around a location.51 52 Optimal search algorithms have been
there are split-second applied to the routing of vehicles and pedestrians to a given destination (i.e.,53 54).
choices to be made about Despite these advances, the widespread application of sensing and optimization
techniques to city infrastructure has been slower than the application of these
whom to put at risk. techniques to individual vehicles or people. Although individual cities have
implemented sensing and optimization applications, as yet there is no standardization
of the sensing infrastructure and AI techniques used. Infrastructure costs, differing
priorities among cities, and the high coordination costs among the parties involved
have slowed adoption, as have public concerns over privacy related to sensing. Still,

44 Kitae Jang, Koohong Chung, and Hwasoo Yeo, “A Dynamic Pricing Strategy for High
Occupancy Toll Lanes,” Transportation Research Part A: Policy and Practice 67 (2014): 69–80.
45 “Seattle Variable Tolling Study,” City of Seattle Department of Transportation, May 2009,
accessed August 1, 2016, http://www.seattle.gov/transportation/docs/FINAL%20Tolling%20
Study%20report%20revised%206.25.10.pdf.
46 James F. Peltz, “Dynamic Pricing Is Catching On in the Public and Private Sectors,”
Government Technology, March 21, 2016, accessed August 1, 2016, http://www.govtech.com/budget-
finance/Dynamic-Pricing-Is-Catching-On-in-the-Public-and-Private-Sectors.html.
47 Arthur G Sims and Kenneth W. Dobinson. “The Sydney Coordinated Adaptive Traffic
(SCAT) System Philosophy and Benefits.” IEEE Transactions on Vehicular Technology 29, no. 2 (1980):
130–137.
48 “New York City Launches Nation’s Most Sophisticated Active Traffic Management System
Powered by TransCore’s TransSuite Traffic Management Software and RFID Technology,”
Business Wire, September 27, 2009, accessed August 1, 2016, http://www.businesswire.com/news/
home/20110927005530/en/York-City-Launches-Nation%E2%80%99s-Sophisticated-Active-Traffic.
49 Eric Horvitz, Johnson Apacible, Raman Sarin, and Lin Liao, “Prediction, Expectation, and
Surprise: Methods, Designs, and Study of a Deployed Traffic Forecasting Service,” Proceedings of the
Twenty-First Conference on Uncertainty and Artificial Intelligence (2005) (Arlington, Virginia: AUAI Press,
July 2005), 275–283.
50 Timothy Hunter, Ryan Herring, Pieter Abbeel, and Alexandre Bayen, “Path and Travel Time
Inference from GPS Probe Vehicle Data,” NIPS Analyzing Networks and Learning with Graphs 12, no. 1 (2009).
51 John Krumm and Eric Horvitz, “Predestination: Inferring Destinations from Partial
Trajectories,” UbiComp 2006: Ubiquitous Computing, Proceedings of the 8th International Conference,
September 2006, (Springer Berlin, Heidelberg, 2006), 243–260.
52 Jill Duffy, “Get Organized: Using Location-Based Reminders,” PC Magazine, June 30, 2014,
accessed August 1, 2016, http://www.pcmag.com/article2/0,2817,2460207,00.asp.
53 Robert J. Szczerba, Peggy Galkowski, I. S. Glicktein, and Noah Ternullo. “Robust Algorithm
for Real-time Route Planning,” IEEE Transactions on Aerospace and Electronic Systems 36, no. 3 (2000):
869–878.
54 Matt Duckham and Lars Kulik, “Simplest” Paths: Automated Route Selection for
Navigation,” Spatial Information Theory. Foundations of Geographic Information Science,
Proceedings of the International Conference, COSIT 2003, September 2003 (Springer-Verlag
22 Berlin Heidelberg, 2003), 169-185.
AI is likely to have an increasing impact on city infrastructure. Accurate predictive
models of individuals’ movements, their preferences, and their goals are likely to
emerge with the greater availability of data. The ethical issues regarding such an
emergence are discussed in Section III of this report.
The United States Department of Transportation released a call for proposals in
2016 asking medium-size cities to imagine smart city infrastructure for transportation.55
This initiative plans to award forty million dollars to a city to demonstrate how
technology and data can be used to reimagine the movement of people as well as goods.
One vision is a network of connected vehicles that can reach a high level of safety Our Study Panel doesn’t
in driving with car-to-car communication.56 If this vision becomes reality, we expect
advances in multi-agent coordination, collaboration, and planning will have a expect drones that can
significant impact on future cars and play a role in making the transportation system
fly, swim, and drive, or
more reliable and efficient. Robots are also likely to take part in transportation by
carrying individuals and packages (c.f., Segway robot). For transportation of goods, flying quadcoptors to
interest in drones has increased, and Amazon is currently testing a delivery system using
them,57 although questions remain about the appropriate safety rules and regulations. become a common means
The increased sensing capabilities, adoption of drones, and the connected
transportation infrastructure will also raise concerns about the privacy of individuals
of transportation by 2030
and the safety of private data. In coming years, these and related transportation issues (although prototypes exist
will need to be addressed either by preemptive action on the part of industry or within
the legal framework. As noted in the Section III policy discussion, how well this is done today).
will affect the pace and scope of AI-related advances in the transportation sector.

On-demand transportation
On-demand transportation services such as Uber and Lyft have emerged as another
pivotal application of sensing, connectivity, and AI,58 with algorithms for matching
drivers to passengers by location and suitability (reputation modeling).59 60
Through dynamic pricing, these services ration access by willingness-to-pay, with
dynamic pricing also encouraging an increase in the supply of drivers, and have
become a popular method for transportation in cities. With their rapid advance have
come multiple policy and legal issues, such as competition with existing taxi services
and concerns about lack of regulation and safety. On-demand transportation services
seem likely to be a major force towards self-driving cars.
Carpooling and ridesharing have long been seen as a promising approach to
decrease traffic congestion and better utilize personal transportation resources.
Services such as Zimride and Nuride bring together people sharing similar routes for a
joint trip. But this approach to carpooling has failed to gain traction on a large scale.

55 “U.S. Department of Transportation Launches Smart City Challenge to Create a City of

the Future,” Transportation.gov, U.S. Department of Transportation, December 7, 2015, accessed
August 1, 2016, https://www.transportation.gov/briefing-room/us-department-transportation-
launches-smart-city-challenge-create-city-future.
56 Will Knight, “Car-to-Car Communication: A simple wireless technology promises to
make driving much safer.,” MIT Technology Review, accessed August 1, 2016, https://www.
technologyreview.com/s/534981/car-to-car-communication/.
57 “Amazon Prime Air,” Amazon, accessed August 1, 2016, http://www.amazon.com/
b?node=8037720011.
58 Jared Meyer, “Uber and Lyft are changing the way Americans move about their country,”
National Review, June 7, 2016, accessed August 1, 2016, http://www.nationalreview.com/
article/436263/uber-lyft-ride-sharing-services-sharing-economy-are-future.
59 Alexander Howard, “How Digital Platforms Like LinkedIn, Uber And TaskRabbit Are
Changing The On-Demand Economy,” The Huffington Post, July 14, 2015, accessed August 1, 2016,
http://www.huffingtonpost.com/entry/online-talent-platforms_us_55a03545e4b0b8145f72ccf6.
60 “Announcing UberPool,” Uber Newsroom, August 5, 2014, accessed August 1, 2016, https://
newsroom.uber.com/announcing-uberpool/. 23
 Interacting with people
For decades, people have imagined wildly different, futuristic-looking transportation
vehicles. Although future cars will be smarter and drones will be available widely, it
is unlikely that by 2030 we will have widely adopted transportation vehicles that look
and function differently than the ones we have today. Our Study Panel doesn’t expect
drones that can fly, swim, and drive, or flying quadcoptors to become a common
means of transportation in this time horizon (although prototypes exist today).
We do expect humans to become partners to self-driving cars and drones in their
Over the next fifteen years, training, execution, and evaluation. This partnering will happen both when humans
are co-located with machines and also virtually. We predict advances in algorithms to
coincident advances facilitate machine learning from human input. We also expect models and algorithms
in mechanical and AI for modeling of human attention, and to support communication and coordination
between humans and machine. This is an integral part of the development of
technologies promise to future vehicles.

increase the safe and

HOME/SERVICE ROBOTS
reliable use and utility of Robots have entered people’s homes in the past fifteen years. Disappointingly slow
growth in the diversity of applications has occurred simultaneously with increasingly
home robots in a typical sophisticated AI deployed on existing applications. AI advances are often inspired by
North American city. mechanical innovations, which in turn prompt new AI techniques to be introduced.
Over the next fifteen years, coincident advances in mechanical and AI technologies
promise to increase the safe and reliable use and utility of home robots in a typical
North American city. Special purpose robots will deliver packages, clean offices,
and enhance security, but technical constraints and the high costs of reliable
mechanical devices will continue to limit commercial opportunities to narrowly
defined applications for the foreseeable future. As with self-driving cars and other
new transportation machines, the difficulty of creating reliable, market-ready
hardware is not to be underestimated.

Vacuum cleaners
In 2001, after many years of development, the Electrolux Trilobite, a vacuum
cleaning robot, became the first commercial home robot. It had a simple control
system to do obstacle avoidance, and some navigation. A year later, iRobot introduced
Roomba, which was a tenth the price of the Trilobite and, with only 512 bytes of
RAM, ran a behavior based controller. The most intelligent thing it did was to avoid
falling down stairs. Since then, sixteen million Roombas have been deployed all over
the world and several other competing brands now exist.
As the processing power and RAM capacity of low cost embedded processors
improved from its dismal state in the year 2000, the AI capabilities of these robots
also improved dramatically. Simple navigation, self-charging, and actions for dealing
with full dust bins were added, followed by ability to deal with electrical cords and
rug tassels, enabled by a combination of mechanical improvements and sensor
based perception. More recently, the addition of full VSLAM (Visual Simultaneous
Location and Mapping)— an AI technology that had been around for twenty years—
has enabled the robots to build a complete 3D world model of a house as they clean,
and become more efficient in their cleaning coverage.
Early expectations that many new applications would be found for home robots have
not materialized. Robot vacuum cleaners are restricted to localized flat areas, while real
homes have lots of single steps, and often staircases; there has been very little research
on robot mobility inside real homes. Hardware platforms remain challenging to build,
24 and there are few applications that people want enough to buy. Perceptual algorithms
for functions such as image labeling, and 3D object recognition, while common at AI
conferences, are still only a few years into development as products.

Home robots 2030

Despite the slow growth to date of robots in the home, there are signs that this will
change in the next fifteen years. Corporations such as Amazon Robotics and Uber
are developing large economies of scale using various aggregation technologies. Also:
System in Module (SiM), with a lot of System on Chip (SoC) subsystems, are
now being pushed out the door by phone-chip makers (Qualcomm’s SnapDragon, Special purpose robots
Samsung’s Artik, etc.). These are better than supercomputers of less than ten years
ago with eight or more sixty-four-bit cores, and specialized silicon for cryptography,
will deliver packages,
camera drivers, additional DSPs, and hard silicon for certain perceptual algorithms. clean offices, and enhance
This means that low cost devices will be able to support much more onboard AI than
we have been able to consider over the last fifteen years. security, but technical
Cloud (“someone else’s computer”) is going to enable more rapid release of new
software on home robots, and more sharing of data sets gathered in many different constraints and high
homes, which will in turn feed cloud-based machine learning, and then power
costs will continue to limit
improvements to already deployed robots.
The great advances in speech understanding and image labeling enabled by deep commercial opportunities
learning will enhance robots’ interactions with people in their homes.
Low cost 3D sensors, driven by gaming platforms, have fueled work on 3D for the foreseeable future.
perception algorithms by thousands of researchers worldwide, which will speed the
development and adoption of home and service robots.
In the past three years, low cost and safe robot arms have been introduced to
hundreds of research labs around the world, sparking a new class of research on
manipulation that will eventually be applicable in the home, perhaps around 2025.
More than half a dozen startups around the world are developing AI-based robots for
the home, for now concentrating mainly on social interaction. New ethics and privacy
issues may surface as a result.

HEALTHCARE
For AI technologies, healthcare has long been viewed as a promising domain.
AI-based applications could improve health outcomes and quality of life for
millions of people in the coming years—but only if they gain the trust of doctors,
nurses, and patients, and if policy, regulatory, and commercial obstacles are removed.
Prime applications include clinical decision support, patient monitoring and
coaching, automated devices to assist in surgery or patient care, and management
of healthcare systems. Recent successes, such as mining social media to infer
possible health risks, machine learning to predict patients at risk, and robotics
to support surgery, have expanded a sense of possibility for AI in healthcare.
Improvements in methods for interacting with medical professionals and patients
will be a critical challenge.
As in other domains, data is a key enabler. There has been an immense forward
leap in collecting useful data from personal monitoring devices and mobile apps, from
electronic health records (EHR) in clinical settings and, to a lesser extent, from robots
designed to assist with medical procedures and hospital operations. But using this
data to enable more finely-grained diagnostics and treatments for both individual
patients and patient populations has proved difficult. Research and deployment have
been slowed by outdated regulations and incentive structures. Poor human-computer
interaction methods and the inherent difficulties and risks of implementing
technologies in such a large and complex system have slowed realization of AI’s 25
 promise in healthcare.61 The reduction or removal of these obstacles, combined with
innovations still on the horizon, have the potential to significantly improve health
outcomes and quality of life for millions of people in the coming years.

The clinical setting

For decades, the vision of an AI-powered clinician’s assistant has been a near cliché.
Although there have been successful pilots of AI-related technology in healthcare,62
the current healthcare delivery system unfortunately remains structurally ill-suited to
AI-based applications absorb and deploy rapid advances. Incentives provided by the Affordable Care Act
have accelerated the penetration of electronic health records (EHRs) into clinical
could improve health practice, but implementation has been poor, eroding clinicians’ confidence in their
outcomes and quality of usefulness. A small group of companies control the EHR market, and user interfaces
are widely considered substandard, including annoying pop-ups that physicians
life for millions of people routinely dismiss. The promise of new analytics using data from EHRs, including AI,
remains largely unrealized due to these and other regulatory and structural barriers.
in the coming years—but Looking ahead to the next fifteen years, AI advances, if coupled with sufficient data
and well-targeted systems, promise to change the cognitive tasks assigned to human
only if they gain the trust
clinicians. Physicians now routinely solicit verbal descriptions of symptoms from
of doctors, nurses, and presenting patients and, in their heads, correlate patterns against the clinical
presentation of known diseases. With automated assistance, the physician could
patients. instead supervise this process, applying her or his experience and intuition to guide the
input process and to evaluate the output of the machine intelligence. The literal
“hands-on” experience of the physician will remain critical. A significant challenge is to
optimally integrate the human dimensions of care with automated reasoning processes.
To achieve future advances, clinicians must be involved and engaged at the outset to
ensure that systems are well-engineered and trusted. Already, a new generation of more
tech savvy physicians routinely utilize specialized apps on mobile devices. At the same
time, workloads on primary care clinicians have increased to the point that they are
grateful for help from any quarter. Thus, the opportunity to exploit new learning methods,
to create structured patterns of inference by mining the scientific literature automatically,
and to create true cognitive assistants by supporting free-form dialogue, has never
been greater. Provided these advances are not stymied by regulatory, legal, and social
barriers, immense improvements to the value of healthcare are within our grasp.

Healthcare analytics
At the population level, AI’s ability to mine outcomes from millions of patient clinical
records promises to enable finer-grained, more personalized diagnosis and treatment.
Automated discovery of genotype-phenotype connections will also become possible
as full, once-in-a-lifetime genome sequencing becomes routine for each patient.
A related (and perhaps earlier) capability will be to find “patients like mine” as a way
to inform treatment decisions based on analysis of a similar cohort. Traditional and
non-traditional healthcare data, augmented by social platforms, may lead to the emergence
of self-defined subpopulations, each managed by a surrounding ecosystem of healthcare
providers augmented with automated recommendation and monitoring systems.
These developments have the potential to radically transform healthcare

61 LeighAnne Olsen, Dara Aisner, and J. Michael McGinnis, eds., “Institute of Medicine
(US) Roundtable on Evidence-Based Medicine,” The Learning Healthcare System: Workshop Summary.
(Washington (DC): National Academies Press (US); 2007), accessed August 1, 2016, http://www.ncbi.
nlm.nih.gov/books/NBK53500/.
62 Katherine E. Henry, David N. Hager, Peter J. Pronovost, and Suchi Saria, “A Targeted Real-
time Early Warning Score (TREWScore) for Septic Shock,” Science Translational Medicine 7, (299),
26 299ra122.
delivery as medical procedures and lifetime clinical records for hundreds of millions
of individuals become available. Similarly, the automated capture of personal
environmental data from wearable devices will expand personalized medicine. These
activities are becoming more commercially viable as vendors discover ways to engage
large populations (e.g. ShareCare)63 and then to create population-scale data that can
be mined to produce individualized analytics and recommendations.
Unfortunately, the FDA has been slow to approve innovative diagnostic software,
and there are many remaining barriers to rapid innovation. HIPAA (Health Insurance
Portability and Accountability Act) requirements for protecting patient privacy create A small group of companies
legal barriers to the flow of patient data to applications that could utilize AI technologies.
Unanticipated negative effects of approved drugs could show up routinely, sooner, and
control the EHR market,
more rigorously than they do today, but mobile apps that analyze drug interactions and user interfaces
may be blocked from pulling the necessary information from patient records. More
generally, AI research and innovation in healthcare are hampered by the lack of are widely considered
widely accepted methods and standards for privacy protection. The FDA has been
slow to approve innovative software, in part due to an unclear understanding of the substandard, including
cost/benefit tradeoffs of these systems. If regulators (principally the FDA) recognize
that effective post-marketing reporting is a dependable hedge against some safety risks,
annoying pop-ups that
faster initial approval of new treatments and interventions may become possible. physicians routinely
Automated image interpretation has also been a promising subject of study for
decades. Progress on interpreting large archives of weakly-labeled images, such as dismiss.
large photo archives scraped from the web, has been explosive. At first blush, it is
surprising that there has not been a similar revolution in interpretation of medical
images. Most medical imaging modalities (CT, MR, ultrasound) are inherently digital,
the images are all archived, and there are large, established companies with internal
R&D (e.g. Siemens, Philips, GE) devoted to imaging.
But several barriers have limited progress to date. Most hospital image archives
have only gone digital over the past decade. More importantly, the problem in
medicine is not to recognize what is in the image (is this a liver or a kidney?), but
rather to make a fine-grained judgement about it (does the slightly darker smudge
in the liver suggest a potentially cancerous tumor?). Strict regulations govern these
high-stakes judgements. Even with state-of-the-art technologies, a radiologist will still
likely have to look at the images, so the value proposition is not yet compelling. Also,
healthcare regulations preclude easy federation of data across institutions. Thus, only
very large organizations of integrated care, such as Kaiser Permanente, are able to
attack these problems.
Still, automated/augmented image interpretation has started to gain momentum.
The next fifteen years will probably not bring fully automated radiology, but initial forays
into image “triage” or second level checking will likely improve the speed and cost-
effectiveness of medical imaging. When coupled with electronic patient record systems,
large-scale machine learning techniques could be applied to medical image data. For
example, multiple major healthcare systems have archives of millions of patient scans,
each of which has an associated radiological report, and most have an associated
patient record. Already, papers are appearing in the literature showing that deep neural
networks can be trained to produce basic radiological findings, with high reliability,
by training from this data.64

63 Sharecare, accessed August 1, 2016, https://www.sharecare.com.

64 Hoo-Chang Shin, Holger R. Roth, Mingchen Gao, Le Lu, Ziyue Xu, Isabella Nogues,
Jianhua Yao, Daniel Mollura, and Ronald M. Summers, “Deep Convolutional Neural Networks for
Computer-aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning,”
IEEE Transactions on Medical Imaging 35, no. 5 (2016): 1285–1298. 27
 Healthcare robotics
Fifteen years ago, healthcare robotics was largely science fiction. One company
called Robodoc,65 a spin-out from IBM, developed robotic systems for orthopedic
surgeries, such as hip and knee replacements. The technology worked, but the
company struggled commercially, and was ultimately shut down and acquired for
its technology.66 More recently, though, the research and practical use of surgical
robotics has exploded.
In 2000 Intuitive Surgical67 introduced the da Vinci system, a novel technology
The problem in medicine initially marketed to support minimally invasive heart bypass surgery, and then gained
substantial market traction for treatment of prostate cancer and merged with its
is not to recognize what is only major competition, Computer Motion, in 2003. The da Vinci, now in its fourth
in the image—is this a liver generation, provides 3D visualization (as opposed to 2D monocular laparoscopy) and
wristed instruments in an ergonomic platform. It is considered the standard of care
or a kidney?—but rather in multiple laparoscopic procedures, and used in nearly three quarters of a million
procedures a year,68 providing not only a physical platform, but also a new data
to make a fine-grained platform for studying the process of surgery.
The da Vinci anticipates a day when much greater insight into how medical
judgement about it. Strict
professionals carry out the process of providing interventional medical care will
regulations govern these be possible. The presence of the da Vinci in day-to-day operation has also opened
the doors to new types of innovation—from new instrumentation to image fusion
high-stakes judgements. to novel biomarkers—creating its own innovation ecosystem. The success of the
platform has inspired potential competitors in robotic surgery, most notably the
Alphabet spin-off Verb, in collaboration with J&J/Ethicon.69 There are likely to be
many more, each exploring a unique niche or space and building out an ecosystem of
sensing, data analytics, augmentation, and automation.
Intelligent automation in hospital operations has been less successful. The story
is not unlike surgical robotics. Twenty years ago, one company, HelpMate, created
a robot for hospital deliveries,70 such as meals and medical records, but ultimately
went bankrupt. More recently, Aethon71 introduced TUG Robots for basic deliveries,
but few hospitals have invested in this technology to date. However, robotics in
other service industries such as hotels and warehouses, including Amazon Robotics
(formerly Kiva), are demonstrating that these technologies are practical and cost
effective in at least some large-scale settings, and may ultimately spur additional
innovation in health care.
Looking ahead, many tasks that appear in healthcare will be amenable to
augmentation, but will not be fully automated. For example, robots may be able to
deliver goods to the right room in a hospital, but then require a person to pick them
up and place them in their final location. Walking a patient down the corridor may

65 ROBODOC, accessed August 1, 2016, http://www.robodoc.com/professionals.html.

66 THINK Surgical, accessed August 1, 2016, http://thinksurgical.com/history.
67 Intuitive Surgical, accessed August 1, 2016, http://www.intuitivesurgical.com.
68 Trefis Team, “Intuitive Surgical Maintains Its Growth Momentum With Strong Growth In
Procedure Volumes,” Forbes, January 22, 2016, accessed August 1, 2016, http://www.forbes.com/
sites/greatspeculations/2016/01/22/intuitive-surgical-maintains-its-growth-momentum-with-
strong-growth-in-procedure-volumes/#22ae6b0939a1.
69 Evan Ackerman, “Google and Johnson & Johnson Conjugate to Create Verb Surgical,
Promise Fancy Medical Robots,” IEEE Spectrum, December 17, 2015, accessed August 1, 2016,
http://spectrum.ieee.org/automaton/robotics/medical-robots/google-verily-johnson-johnson-
verb-surgical-medical-robots.
70 John M. Evans and Bala Krishnamurthy, “HelpMate®, the trackless robotic courier: A
perspective on the development of a commercial autonomous mobile robot,” Lecture Notes in Control
and Information Sciences 236, June 18, 2005 (Springer-Verlag London Limited, 1998), 182–210,
accessed August 1, 2016, http://link.springer.com/chapter/10.1007%2FBFb0030806.
28 71 Aethon, accessed August 1, 2016, http://www.aethon.com.
be relatively simple once a patient is standing in a walker (though will certainly not
be trivial for patients recovering from surgery and/or elderly patients, especially in
corridors crowded with equipment and other people). Driving a needle to place a
suture is relatively straightforward once the needle is correctly placed.72 This implies
that many future systems will involve intimate interaction between people and
machines and require technologies that facilitate collaboration between them.
The growth of automation will enable new insights into healthcare processes.
Historically, robotics has not been a strongly data-driven or data-oriented science.
This is changing as (semi)automation infiltrates healthcare. As the new surgical, Specialized motion
delivery, and patient care platforms come online, the beginnings of quantification
and predictive analytics are being built on top of data coming from these platforms.73 tracking devices... and
This data will be used to assess quality of performance, identify deficiencies, errors,
the emerging (inter)
or potential optimizations, and will be used as feedback to improve performance. In
short, these platforms will facilitate making the connection between what is done, and connectedness between
the outcome achieved, making true “closed-loop” medicine a real possibility.
the home environment and
Mobile health
To date, evidence-driven analytics on healthcare have relied on traditional healthcare
health-monitoring devices
data—mainly the electronic medical records discussed above. In the clinical setting, have created a vibrant new
there are hopeful trends towards bringing new data to bear. For example, Tele-
Language enables a human clinician to conduct language therapy sessions with sector of innovation.
multiple patients simultaneously with the aid of an AI agent trained by the clinician.
And Lifegraph, which extracts behavioral patterns and creates alerts from data
passively collected from a patient’s smartphone, has been adopted by psychiatrists in
Israel to detect early signs of distressful behavior in patients.
Looking ahead, driven by the mobile computing revolution, the astonishing growth
of “biometrics in the wild”—and the explosion of platforms and applications that use
them—is a hopeful and unanticipated trend. Thousands of mobile apps now offer
information, introduce behavior modification, or identify groups of “people like me.”
This, combined with the emerging trend of more specialized motion tracking devices,
such as Fitbit, and the emerging (inter)connectedness between the home environment
and health-monitoring devices, has created a vibrant new sector of innovation.
By combining social and healthcare data, some healthcare apps can perform data
mining, learning, and prediction from captured data, though their predictions are
relatively rudimentary. The convergence of data and functionality across applications
will likely spur new and even obvious products, such as an exercise app that not only
proposes a schedule for exercise but also suggests the best time to do it, and provides
coaching to stick to that schedule.

72 Azad Shademan, Ryan S. Decker, Justin D. Opfermann, Simon Leonard, Axel Krieger, and
Peter CW Kim, “Supervised Autonomous Robotic Soft Tissue Surgery,” Science Translational Medicine
8, no. 337 (2016): 337ra64–337ra64.
73 Carolyn Chen, Lee White, Timothy Kowalewski, Rajesh Aggarwal, Chris Lintott, Bryan
Comstock, Katie Kuksenok, Cecilia Aragon, Daniel Holst, and Thomas Lendvay, “Crowd-Sourced
Assessment of Technical Skills: a novel method to evaluate surgical performance.” Journal of Surgical
Research 187, no. 1 (2014): 65–71.
29
 Elder care
Over the next fifteen years the number of elderly in the United States will grow
by over 50%.74 The National Bureau of Labor Statistics projects that home health
aides will grow 38% over the next ten years. Despite the broad opportunities in this
domain—basic social support, interaction and communication devices, home health
monitoring, a variety of simple in-home physical aids such as walkers, and light meal
preparation—little has happened over the past fifteen years. But the coming generational
shift will accompany a change in technology acceptance among the elderly. Currently,
Better hearing aids and someone who is seventy was born in 1946 and may have first experienced some form
of personalized IT in middle age or later, while a fifty-year-old today is far more
visual assistive devices technology-friendly and savvy. As a result, there will be a growing interest and market
will mitigate the effects for already available and maturing technologies to support physical, emotional, social,
and mental health. Here are a few likely examples by category:
of hearing and vision Life quality and independence
• Automated transportation will support continued independence and expanded
loss, improving safety social horizons.
• Sharing of information will help families remain engaged with one another at a
and social connection.
distance, and predictive analytics may be used to “nudge” family groups toward
Personalized rehabilitation positive behaviors, such as reminders to “call home.”
• Smart devices in the home will help with daily living activities when needed,
and in-home therapy such as cooking and, if robot manipulation capabilities improve sufficiently,
dressing and toileting.
will reduce the need for
Health and wellness
hospital stays. • Mobile applications that monitor movement and activities, coupled with social
platforms, will be able to make recommendations to maintain mental and
physical health.
• In-home health monitoring and health information access will be able to detect
changes in mood or behavior and alert caregivers.
• Personalized health management will help mitigate the complexities associated
with multiple co-morbid conditions and/or treatment interactions.
Treatments and devices
• Better hearing aids and visual assistive devices will mitigate the effects of
hearing and vision loss, improving safety and social connection.
• Personalized rehabilitation and in-home therapy will reduce the need for
hospital or care facility stays.
• Physical assistive devices (intelligent walkers, wheel chairs, and exoskeletons) will
extend the range of activities of an infirm individual.
The Study Panel expects an explosion of low-cost sensing technologies that can
provide substantial capabilities to the elderly in their homes. In principle, social
agents with a physical presence and simple physical capabilities (e.g. a mobile robot
with basic communication capabilities) could provide a platform for new innovations.
However, doing so will require integration across multiple areas of AI—Natural
Language Processing, reasoning, learning, perception, and robotics—to create
a system that is useful and usable by the elderly.
These innovations will also introduce questions regarding privacy within various
circles, including friends, family, and care-givers, and create new challenges to
accommodate an evermore active and engaged population far past retirement.

74 Jennifer M. Ortman, Victoria A. Velkoff, and Howard Hogan, “An Aging Nation: The
Older Population in the United States: Population Estimates and Projections,” Current Population
Reports, U.S Census Bureau (May 2014), accessed August 1, 2016, https://www.census.gov/
prod/2014pubs/p25-1140.pdf.
30
EDUCATION
The past fifteen years have seen considerable AI advances in education. Applications
are in wide use by educators and learners today, with some variation between K-12
and university settings. Though quality education will always require active engagement
by human teachers, AI promises to enhance education at all levels, especially by
providing personalization at scale. Similar to healthcare, resolving how to best
integrate human interaction and face-to-face learning with promising AI technologies
remains a key challenge.
Though quality education
Robots have long been popular educational devices, starting with the early
Lego Mindstorms kits developed with the MIT Media Lab in the 1980s. Intelligent will always require active
Tutoring Systems (ITS) for science, math, language, and other disciplines match
students with interactive machine tutors. Natural Language Processing, especially when engagement by human
combined with machine learning and crowdsourcing, has boosted online learning
and enabled teachers to multiply the size of their classrooms while simultaneously
teachers, AI promises to
addressing individual students’ learning needs and styles. The data sets from large enhance education at
online learning systems have fueled rapid growth in learning analytics.
Still, schools and universities have been slow in adopting AI technologies primarily all levels, especially by
due to lack of funds and lack of solid evidence that they help students achieve
learning objectives. Over the next fifteen years in a typical North American city, the providing personalization
use of intelligent tutors and other AI technologies to assist teachers in the classroom
at scale.
and in the home is likely to expand significantly, as will learning based on virtual
reality applications. But computer-based learning systems are not likely to fully
replace human teaching in schools.

Teaching robots
Today, more sophisticated and versatile kits for use in K-12 schools are available
from a number of companies that create robots with new sensing technologies
programmable in a variety of languages. Ozobot is a robot that teaches children
to code and reason deductively while configuring it to dance or play based on
color-coded patterns.75 Cubelets help teach children logical thinking through
assembling robot blocks to think, act, or sense, depending upon the function of the
different blocks.76 Wonder Workshop’s Dash and Dot span a range of programming
capabilities. Children eight years old and older can create simple actions using a
visual programming language, Blockly, or build iOS and Android applications using
C or Java.77 PLEO rb is a robot pet that helps children learn biology by teaching
the robot to react to different aspects of the environment.78 However, while fun and
engaging for some, in order for such kits to become widespread, there will need to be
compelling evidence that they improve students’ academic performance.

Intelligent Tutoring Systems (ITS) and online learning

ITS have been developed from research laboratory projects such as Why-2 Atlas,
which supported human-machine dialogue to solve physics problems early in the
era.79 The rapid migration of ITS from laboratory experimental stages to real use is

75 Ozobot, accessed August 1, 2016, http://ozobot.com/.

76 “Cubelets,” Modular Robotics, accessed August 1, 2016, http://www.modrobotics.com/
cubelets.
77 “Meet Dash,” Wonder Workshop, accessed August 1, 2016, https://www.makewonder.com/
dash.
78 “Pleo rb,” Innvo Labs, accessed August 1, 2016, http://www.pleoworld.com/pleo_rb/eng/
lifeform.php.
79 Kurt VanLehn, Pamela W. Jordan, Carolyn P. Rosé, Dumisizwe Bhembe, Michael Böttner,
Andy Gaydos, Maxim Makatchev, Umarani Pappuswamy, Michael Ringenberg, Antonio Roque,
Stephanie Siler, and Ramesh Srivastava, “The Architecture of Why2-Atlas: A Coach for Qualitative 31
 surprising and welcome. Downloadable software and online systems such as Carnegie
Speech or Duolingo provide foreign language training using Automatic Speech
Recognition (ASR) and NLP techniques to recognize language errors and help users
correct them.80 Tutoring systems such as the Carnegie Cognitive Tutor81 have been
used in US high schools to help students learn mathematics. Other ITS have been
developed for training in geography, circuits, medical diagnosis, computer literacy
and programming, genetics, and chemistry. Cognitive tutors use software to mimic
the role of a good human tutor by, for example, providing hints when a student gets
It can be argued that AI stuck on a math problem. Based on the hint requested and the answer provided, the
tutor offers context specific feedback.
is the secret sauce that Applications are growing in higher education. An ITS called SHERLOCK82 is
has enabled instructors, beginning to be used to teach Air Force technicians to diagnose electrical systems
problems in aircraft. And the University of Southern California’s Information
particularly in higher Sciences Institute has developed more advanced avatar-based training modules to
train military personnel being sent to international posts in appropriate behavior
education, to multiply the when dealing with people from different cultural backgrounds. New algorithms for
personalized tutoring, such as Bayesian Knowledge Tracing, enable individualized
size of their classrooms
mastery learning and problem sequencing.83
by a few orders of Most surprising has been the explosion of the Massive Open Online Courses
(MOOCs) and other models of online education at all levels—including the use
magnitude—class sizes of of tools like Wikipedia and Khan Academy as well as sophisticated learning
management systems that build in synchronous as well as asynchronous education
a few tens of thousands and adaptive learning tools. Since the late 1990s, companies such as the Educational
are not uncommon. Testing Service and Pearson have been developing automatic NLP assessment tools to
co-grade essays in standardized testing.84 Many of the MOOCs which have become
so popular, including those created by EdX, Coursera, and Udacity, are making use
of NLP, machine learning, and crowdsourcing techniques for grading short-answer
and essay questions as well as programming assignments.85 Online education systems
that support graduate-level professional education and lifelong learning are also
expanding rapidly. These systems have great promise because the need for face-to-face
interaction is less important for working professionals and career changers. While not
the leaders in AI-supported systems and applications, they will become early adopters
as the technologies are tested and validated.
It can be argued that AI is the secret sauce that has enabled instructors, particularly
in higher education, to multiply the size of their classrooms by a few orders of
magnitude—class sizes of a few tens of thousands are not uncommon. In order to
continually test large classes of students, automated generation of the questions is

Physics Essay Writing,” Intelligent Tutoring Systems: Proceedings of the 6th International Conference, (Springer
Berlin Heidelberg, 2002), 158–167.
80 VanLehn et al, “The Architecture of Why2-Atlas.”
81 “Resources and Support,” Carnegie Learning, accessed August 1, 2016, https://www.
carnegielearning.com/resources-support/.
82 Alan Lesgold, Suzanne Lajoie, Marilyn Bunzo, and Gary Eggan, “SHERLOCK: A Coached
Practice Environment for an Electronics Troubleshooting Job,” in J. H. Larkin and R. W. Chabay,
eds., Computer-Assisted Instruction and Intelligent Tutoring Systems: Shared Goals and Complementary Approaches
(Hillsdale, New Jersey: Lawrence Erlbaum Associates, 1988).
83 Michael V. Yudelson, Kenneth R. Koedinger, and Geoffrey J. Gordon, (2013). “ Individualized
Bayesian Knowledge Tracing Models,” Artificial Intelligence in Education, (Springer Berlin Heidelberg,
2013), 171–180.
84 Jill Burstein, Karen Kukich, Susanne Wolff, Chi Lu, Martin Chodorow, Lisa Braden-Harder,
and Mary Dee Harris, “Automated Scoring Using a Hybrid Feature Identification Technique” in
Proceedings of the Annual Meeting of the Association of Computational Linguistics, Montreal, Canada, August
1998, accessed August 1, 2016, https://www.ets.org/Media/Research/pdf/erater_acl98.pdf.
85 EdX, https://www.edx.org/, Coursera, https://www.coursera.org/, Udacity, https://www.
32 udacity.com/, all accessed August 1, 2016.
also possible, such as those designed to assess vocabulary,86 wh (who/what/when/
where/why) questions,87 and multiple choice questions,88 using electronic resources
such as WordNet, Wikipedia, and online ontologies. With the explosion of online
courses, these techniques are sure to be eagerly adopted for use in online education.
Although the long term impact of these systems will have on the educational system
remains unclear, the AI community has learned a great deal in a very short time.

Learning analytics
Data sets being collected from massive scale online learning systems, ranging from The current absence of
MOOCs to Khan Academy, as well as smaller scale online programs, have fueled
the rapid growth of the field of learning analytics. Online courses are not only good sophisticated use of AI
for widespread delivery, but are natural vehicles for data collection and experimental
technologies in schools,
instrumentation that will contribute to scientific findings and improving the quality
of learning at scale. Organizations such as the Society for Learning Analytics colleges, and universities
Research (SOLAR), and the rise of conferences including the Learning Analytics and
Knowledge Conference89 and the Learning at Scale Conference (L@S)90 reflect this may be explained by
trend. This community applies deep learning, natural language processing, and other
AI techniques to analysis of student engagement, behavior, and outcomes.
the lack of financial
Current projects seek to model common student misconceptions, predict resources as well as the
which students are at risk of failure, and provide real-time student feedback that is
tightly integrated with learning outcomes. Recent work has also been devoted to lack of data establishing
understanding the cognitive processes involved in comprehension, writing, knowledge
acquisition, and memory, and to applying that understanding to educational practice the technologies’
by developing and testing educational technologies.
effectiveness.
Challenges and opportunities
One might have expected more and more sophisticated use of AI technologies in
schools, colleges, and universities by now. Much of its absence can be explained
by the lack of financial resources of these institutions as well as the lack of data
establishing the technologies’ effectiveness. These problems are being addressed,
albeit slowly, by private foundations and by numerous programs to train primarily
secondary school teachers in summer programs. As in other areas of AI, excessive
hype and promises about the capabilities of MOOCs have meant that expectations
frequently exceed the reality. The experiences of certain institutions, such as San Jose
State University’s experiment with Udacity,91 have led to more sober assessment of
the potential of the new educational technologies.

86 Jonathan C. Brown, Gwen A. Frishkoff , and Maxine Eskenazi, “Automatic Question

Generation for Vocabulary Assessment,” Proceedings of Human Language Technology Conference and
Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), Vancouver, October
2005, (Association for Computational Linguistics, 2005), 819–826.
87 Michael Heilman, “Automatic Factual Question Generation from Text,” PhD thesis CMU-
LTI-11-004, (Carnegie Mellon University, 2011), accessed August 1, 2016, http://www.cs.cmu.
edu/~ark/mheilman/questions/papers/heilman-question-generation-dissertation.pdf.
88 Tahani Alsubait, Bijan Parsia, and Uli Sattler, “Generating Multiple Choice Questions from
Ontologies: How Far Can We Go?,” in eds. P. Lambrix, E. Hyvönen. E. Blomqvist, V. Presutti, G.
Qi, U. Sattler, Y. Ding, and C. Ghidini, Knowledge Engineering and Knowledge Management: EKAW 2014
Satellite Events, VISUAL, EKM1, and ARCOE-Logic Linköping, Sweden, November 24–28, 2014 Revised
Selected Papers, (Switzerland: Springer International Publishing, 2015), 66–79.
89 The 6th International Learning Analytics & Knowledge Conference, accessed August 1, 2016,
http://lak16.solaresearch.org/.
90 Third Annual ACM Conference on Learning at Scale, http://learningatscale.acm.org/
las2016/.
91 Ry Rivard, “Udacity Project on ‘Pause’,” Inside Higher Ed, July 18, 2013, accessed August 1,
2016, https://www.insidehighered.com/news/2013/07/18/citing-disappointing-student-outcomes-
san-jose-state-pauses-work-udacity. 33
 In the next fifteen years, it is likely that human teachers will be assisted by AI
technologies with better human interaction, both in the classroom and in the home.
The Study Panel expects that more general and more sophisticated virtual reality
scenarios in which students can immerse themselves in subjects from all disciplines
will be developed. Some steps in this direction are being taken now by increasing
collaborations between AI researchers and researchers in the humanities and social
sciences, exemplified by Stanford’s Galileo Correspondence Project92 and Columbia’s
Making and Knowing Project.93 These interdisciplinary efforts create interactive
While formal education experiences with historical documents and the use of Virtual Reality (VR) to explore
interactive archeological sites.94 VR techniques are already being used in the natural
will not disappear, the sciences such as biology, anatomy, geology and astronomy to allow students to interact
Study Panel believes that with environments and objects that are difficult to engage with in the real world. The
recreation of past worlds and fictional worlds will become just as popular for studies
MOOCs and other forms of arts and other sciences.
AI techniques will increasingly blur the line between formal, classroom education
of online education will and self-paced, individual learning. Adaptive learning systems, for example, are going
to become a core part of the teaching process in higher education because of the
become part of learning
pressures to contain cost while serving a larger number of students and moving students
at all levels, from K-12 through school more quickly. While formal education will not disappear, the Study
Panel believes that MOOCs and other forms of online education will become part of
through university, in learning at all levels, from K-12 through university, in a blended classroom experience.
This development will facilitate more customizable approaches to learning, in which
a blended classroom students can learn at their own pace using educational techniques that work best for
experience. them. Online education systems will learn as the students learn, supporting rapid
advances in our understanding of the learning process. Learning analytics, in turn, will
accelerate the development of tools for personalized education.
The current transition from hard copy books to digital and audio media and texts
is likely to become prevalent in education as well. Digital reading devices will also
become much ‘smarter’, providing students with easy access to additional information
about subject matter as they study. Machine Translation (MT) technology will
also make it easier to translate educational material into different languages with a
fair degree of accuracy, just as it currently translates technical manuals. Textbook
translation services that currently depend only upon human translators will
increasingly incorporate automatic methods to improve the speed and affordability of
their services for school systems.
Online learning systems will also expand the opportunity for adults and working
professionals to enhance their knowledge and skills (or to retool and learn a new field)
in a world where these fields are evolving rapidly. This will include the expansion
of fully online professional degrees as well as professional certifications based on
online coursework.

Broader societal consequences

In countries where education is difficult for the broad population to obtain, online
resources may have a positive effect if the population has the tools to access them.
The development of online educational resources should make it easier for
foundations that support international educational programs to provide quality
92 Stanford University: Galileo Correspondence Project, accessed August 1, 2016, http://galileo.
stanford.edu.
93 The Making and Knowing Project: Reconstructing the 16th Century Workshop of BNF MS.
FR. 640 at Columbia University, accessed August 1, 2016, http://www.makingandknowing.org.
94 Paul James, “3D Mapped HTC Vive Demo Brings Archaeology to Life,” Road to VR, August
31, 2015, accessed August 1, 2016, http://www.roadtovr.com/3d-mapped-htc-vive-demo-brings-
34 archaeology-to-life/.
education by providing tools and relatively simple amounts of training in their use.
For example, large numbers of educational apps, many of them free, are being
developed for the iPad. On the negative side, there is already a major trend among
students to restrict their social contacts to electronic ones and to spend large amounts
of time without social contact, interacting with online programs. If education also
occurs more and more online, what effect will the lack of regular, face-to-face contact
with peers have on students’ social development? Certain technologies have even
been shown to create neurological side effects.95 On the other hand, autistic children
have benefited from interactions with AI systems already.96 With targeted incentives
and funding priorities, AI
LOW-RESOURCE COMMUNITIES
Many opportunities exist for AI to improve conditions for people in low-resource technologies could help
communities in a typical North American city—and, indeed, in some cases it
address the needs of low-
already has. Understanding these direct AI contributions may also inform potential
contributions in the poorest parts of the developing world. There has not been a resource communities.
significant focus on these populations in AI gatherings, and, traditionally, AI funders
have underinvested in research lacking commercial application. With targeted incentives Budding efforts are
and funding priorities, AI technologies could help address the needs of low-resource
communities. Budding efforts are promising. Counteracting fears that AI may contribute
promising.
to joblessness and other societal problems, AI may provide mitigations and solutions,
particularly if implemented in ways that build trust in them by the affected communities.
Machine learning, data mining approaches
Under the banner of “data science for social good,” AI has been used to create
predictive models to help government agencies more effectively use their limited
budgets to address problems such as lead poisoning,97 a major public health concern
that has been in the news due to ongoing events in Flint, Michigan. Children may
be tested for elevated lead levels, but that unfortunately means the problem is only
detected after they have already been poisoned. Many efforts are underway to use
predictive models to assist government agencies in prioritizing children at
risk, including those who may not yet have been exposed.98 Similarly, the Illinois
Department of Human Services (IDHS) uses predictive models to identify pregnant
women at risk for adverse birth outcomes in order to maximize the impact of
prenatal care. The City of Cincinnati uses them to proactively identify and deploy
inspectors to properties at risk of code violations.
Scheduling, planning
Task assignment scheduling and planning techniques have been applied by many
different groups to distribute food before it spoils from those who may have excess,
such as restaurants, to food banks, community centers and individuals.99

95 Scientist have studied, for example, the way reliance on GPS may lead to changes in the
hippocampus. Kim Tingley, “The Secrets of the Wave Pilots,” The New York Times, March 17, 2016,
accessed August 1, 2016, http://www.nytimes.com/2016/03/20/magazine/the-secrets-of-the-wave-
pilots.html.
96 Judith Newman, “To Siri, With Love: How One Boy With Autism Became BFF With Apple’s
Siri,” The New York Times, October 17, 2014, accessed August 1, 2016, http://www.nytimes.
com/2014/10/19/fashion/how-apples-siri-became-one-autistic-boys-bff.html.
97 Eric Potash, Joe Brew, Alexander Loewi, Subhabrata Majumdar, Andrew Reece, Joe Walsh,
Eric Rozier, Emile Jorgensen, Raed Mansour, and Rayid Ghani, “Predictive Modeling for Public
Health: Preventing Childhood Lead Poisoning,” Proceedings of the 21th ACM SIGKDD International
Conference on Knowledge Discovery and Data Mining (New York: Association for Computing Machinery,
2015), 2039–2047.
98 Data Science for Social Good, University of Chicago, accessed August 1, 2016,
http://dssg.uchicago.edu/.
99 Senay Solak, Christina Scherrer, and Ahmed Ghoniem, “The Stop-and-Drop Problem in
Nonprofit Food Distribution Networks,” Annals of Operations Research 221, no. 1 (October 2014): 35
 Reasoning with social networks and influence maximization
Social networks can be harnessed to create earlier, less-costly interventions involving
large populations. For example, AI might be able to assist in spreading health-
related information. In Los Angeles, there are more than 5,000 homeless youth (ages
thirteen-twenty-four). Individual interventions are difficult and expensive, and the
youths’ mistrust of authority dictates that key messages are best spread through peer
leaders. AI programs might be able to leverage homeless youth social networks to
strategically select peer leaders to spread health-related information, such as how to
One of the more successful
avoid spread of HIV. The dynamic, uncertain nature of these networks does pose
uses of AI analytics is in challenges for AI research.100 Care must also be taken to prevent AI systems from
reproducing discriminatory behavior, such as machine learning that identifies people
detecting white collar through illegal racial indicators, or through highly-correlated surrogate factors, such
crime, such as credit as zip codes. But if deployed with great care, greater reliance on AI may well result
in a reduction in discrimination overall, since AI programs are inherently more easily
card fraud. Cybersecurity audited than humans.

(including spam) is a PUBLIC SAFETY AND SECURITY

widely shared concern, Cities already have begun to deploy AI technologies for public safety and security.
By 2030, the typical North American city will rely heavily upon them. These include
and machine learning is cameras for surveillance that can detect anomalies pointing to a possible crime,
making an impact. drones, and predictive policing applications. As with most issues, there are benefits
and risks. Gaining public trust is crucial. While there are legitimate concerns that
policing that incorporates AI may become overbearing or pervasive in some contexts,
the opposite is also possible. AI may enable policing to become more targeted and
used only when needed. And assuming careful deployment, AI may also help remove
some of the bias inherent in human decision-making.
One of the more successful uses of AI analytics is in detecting white collar
crime, such as credit card fraud.101 Cybersecurity (including spam) is a widely shared
concern, and machine learning is making an impact. AI tools may also prove useful
in helping police manage crime scenes or search and rescue events by helping
commanders prioritize tasks and allocate resources, though these tools are not yet
ready for automating such activities. Improvements in machine learning in general,
and transfer learning in particular—for speeding up learning in new scenarios based
on similarities with past scenarios—may facilitate such systems.
The cameras deployed almost everywhere in the world today tend to be more
useful for helping solve crimes than preventing them.102 103 This is due to the low
quality of event identification from videos and the lack of manpower to look at
massive video streams. As AI for this domain improves, it will better assist crime
prevention and prosecution through greater accuracy of event classification and
efficient automatic processing of video to detect anomalies—including, potentially,

407–426.
100 Jordan Pearson, “Artificial Intelligence Could Help Reduce HIV Among Homeless Youths,”
Teamcore, University of Southern California, February 4. 2015, accessed August 1, 2016, http://
teamcore.usc.edu/news/motherboard_news_ai_could_help_reduce_HIV.pdf.
101 “RSA Adaptive Authentication,” RSA, accessed August 1, 2016, https://www.rsa.com/en-us/
products-services/fraud-prevention/adaptive-authentication.
102 Takeshi Arikuma and Yasunori Mochizuki, “Intelligent multimedia surveillance system for
safer cities” APSIPA Transactions on Signal and Information Processing 5 (2016): 1–8.
103 “Big Op-Ed: Shifting Opinions On Surveillance Cameras,”, Talk of the Nation, NPR, April 22,
2013, accessed August 1, 2016, http://www.npr.org/2013/04/22/178436355/big-op-ed-shifting-
36 opinions-on-surveillance-cameras.
evidence of police malpractice. These improvements could lead to even more
widespread surveillance. Some cities have already added drones for surveillance
purposes, and police use of drones to maintain security of ports, airports, coastal
areas, waterways, industrial facilities is likely to increase, raising concerns about
privacy, safety, and other issues.
The New York Police Department’s CompStat was the first tool pointing toward
predictive policing,104 and many police departments now use it.105 Machine learning
significantly enhances the ability to predict where and when crimes are more likely
to happen and who may commit them. As dramatized in the movie Minority Report, As dramatized in the
predictive policing tools raise the specter of innocent people being unjustifiably
targeted. But well-deployed AI prediction tools have the potential to actually remove movie Minority Report,
or reduce human bias, rather than reinforcing it, and research and resources should
predictive policing tools
be directed toward ensuring this effect.
AI techniques can be used to develop intelligent simulations for training law- raise the specter of
enforcement personnel to collaborate. While international criminal organizations and
terrorists from different countries are colluding, police forces from different countries innocent people being
still face difficulty in joining forces to fight them. Training international groups of law
enforcement personnel to work as teams is very challenging. The European Union,
unjustifiably targeted. But
through the Horizon 2020 program, currently supports such attempts in projects such well-deployed AI prediction
as LawTrain.106 The next step will be to move from simulation to actual investigations
by providing tools that support such collaborations. tools have the potential to
Tools do exist for scanning Twitter and other feeds to look for certain types
of events and how they may impact security. For example, AI can help in social actually remove or reduce
network analysis to prevent those at risk from being radicalized by ISIS or other
human bias.
violent groups. Law enforcement agencies are increasingly interested in trying to
detect plans for disruptive events from social media, and also to monitor activity at
large gatherings of people to analyze security. There is significant work on crowd
simulations to determine how crowds can be controlled. At the same time, legitimate
concerns have been raised about the potential for law enforcement agencies to
overreach and use such tools to violate people’s privacy.
The US Transportation Security Administration (TSA), Coast Guard, and the
many other security agencies that currently rely on AI will likely increase their
reliance to enable significant efficiency and efficacy improvements.107 AI techniques—
vision, speech analysis, and gait analysis— can aid interviewers, interrogators, and
security guards in detecting possible deception and criminal behavior. For example,
the TSA currently has an ambitious project to redo airport security nationwide.108
Called DARMS, the system is designed to improve efficiency and efficacy of airport
security by relying on personal information to tailor security based on a person’s risk
categorization and the flights being taken. The future vision for this project is a tunnel
that checks people’s security while they walk through it. Once again, developers of
this technology should be careful to avoid building in bias (e.g. about a person’s risk
level category) through use of datasets that reflect prior bias.109

104 Walter L. Perry, Brian McInnis, Carter C. Price, Susan Smith, and John S. Hollywood, “The
Role of Crime Forecasting in Law Enforcement Operations,” Rand Corporation Report 233 (2013).
105 “CompStat,” Wikipedia, last modified July 28, 2016, accessed August 1, 2016, https://
en.wikipedia.org/wiki/CompStat.
106 LAW-TRAIN, accessed August 1, 2016, http://www.law-train.eu/.
107 Milind Tambe, Security and Game Theory: Algorithms, Deployed Systems, Lessons Learned (New York:
Cambridge University Press, 2011).
108 Peter Neffenger, “TSA’s 2017 Budget—A Commitment to Security (Part I),” Department
of Homeland Security, March 1, 2016, accessed August 1, 2016, https://www.tsa.gov/news/
testimony/2016/03/01/hearing-fy17-budget-request-transportation-security-administration.
109 Crawford, “AI’s White Guy Problem.” 37
 EMPLOYMENT AND WORKPLACE
While AI technologies are likely to have a profound future impact on employment
and workplace trends in a typical North American city, it is difficult to accurately
assess current impacts, positive or negative. In the past fifteen years, employment
has shifted due to a major recession and increasing globalization, particularly with
China’s introduction to the world economy, as well as enormous changes in non-AI
digital technology. Since the 1990s, the US has experienced continued growth in
productivity and GDP, but median income has stagnated and the employment to
AI will likely replace tasks population ratio has fallen.
rather than jobs in the There are clear examples of industries in which digital technologies have had
profound impacts, good and bad, and other sectors in which automation will likely
near term, and will also make major changes in the near future. Many of these changes have been driven
strongly by “routine” digital technologies, including enterprise resource planning,
create new kinds of jobs. networking, information processing, and search. Understanding these changes should
But the new jobs that provide insights into how AI will affect future labor demand, including the shift in
skill demands. To date, digital technologies have been affecting workers more in the
will emerge are harder to skilled middle, such as travel agents, rather than the very lowest-skilled or highest
skilled work.110 On the other hand, the spectrum of tasks that digital systems can
imagine in advance than do is evolving as AI systems improve, which is likely to gradually increase the scope
of what is considered routine. AI is also creeping into high end of the spectrum,
the existing jobs that will including professional services not historically performed by machines.
likely be lost. To be successful, AI innovations will need to overcome understandable human
fears of being marginalized. AI will likely replace tasks rather than jobs in the near
term, and will also create new kinds of jobs. But the new jobs that will emerge are
harder to imagine in advance than the existing jobs that will likely be lost. Changes
in employment usually happen gradually, often without a sharp transition, a trend
likely to continue as AI slowly moves into the workplace. A spectrum of effects will
emerge, ranging from small amounts of replacement or augmentation to complete
replacement. For example, although most of a lawyer’s job is not yet automated,111
AI applied to legal information extraction and topic modeling has automated parts of
first-year lawyers’ jobs.112 In the not too distant future, a diverse array of job-holders,
from radiologists to truck drivers to gardeners, may be affected.
AI may also influence the size and location of the workforce. Many organizations
and institutions are large because they perform functions that can be scaled only by
adding human labor, either “horizontally” across geographical areas or “vertically”
in management hierarchies. As AI takes over many functions, scalability no longer
implies large organizations. Many have noted the small number of employees of
some high profile internet companies, but not of others. There may be a natural scale
of human enterprise, perhaps where the CEO can know everyone in the company.
Through the creation of efficiently outsourced labor markets enabled by AI,
enterprises may tend towards that natural size.
AI will also create jobs, especially in some sectors, by making certain tasks more
important, and create new categories of employment by making new modes of
interaction possible. Sophisticated information systems can be used to create new
110 Jeremy Ashkenas and Alicia Parlapiano, “How the Recession Reshaped the Economy, in 255
Charts,” The New York Times, June 6, 2014, accessed August 1, 2016, http://www.nytimes.com/
interactive/2014/06/05/upshot/how-the-recession-reshaped-the-economy-in-255-charts.html.
111 R Dana Remus and Frank S. Levy, “Can Robots Be Lawyers? Computers, Lawyers, and the
Practice of Law,” Social Science Research Network, last modified February 12, 2016, accessed August 1,
2016, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2701092.
112 John Markoff, “Armies of Expensive Lawyers, Replaced by Cheaper Software,” The New
York Times, March 4, 2011, accessed August 1, 2016, http://www.nytimes.com/2011/03/05/
38 science/05legal.html.
markets, which often have the effect of lowering barriers to entry and increasing
participation—from app stores to AirBnB to taskrabbit. A vibrant research
community within AI studies further ways of creating new markets and making
existing ones operate more efficiently.
While work has intrinsic value, most people work to be able to purchase goods and
services they value. Because AI systems perform work that previously required human
labor, they have the effect of lowering the cost of many goods and services, effectively
making everyone richer. But as exemplified in current political debates, job loss is
more salient to people—especially those directly affected—than diffuse economic As labor becomes a
gains, and AI unfortunately is often framed as a threat to jobs rather than a boon to
living standards. less important factor in
There is even fear in some quarters that advances in AI will be so rapid as
production as compared
to replace all human jobs—including those that are largely cognitive or involve
judgment—within a single generation. This sudden scenario is highly unlikely, but to owning intellectual
AI will gradually invade almost all employment sectors, requiring a shift away from
human labor that computers are able to take over. capital, a majority of
The economic effects of AI on cognitive human jobs will be analogous to
the effects of automation and robotics on humans in manufacturing jobs. Many
citizens may find the value
middle-aged workers have lost well-paying factory jobs and the socio-economic status of their labor insufficient
in family and society that traditionally went with such jobs. An even larger fraction
of the total workforce may, in the long run, lose well-paying “cognitive” jobs. As labor to pay for a socially
becomes a less important factor in production as compared to owning intellectual
capital, a majority of citizens may find the value of their labor insufficient to pay for acceptable standard of
a socially acceptable standard of living. These changes will require a political, rather
living.
than a purely economic, response concerning what kind of social safety nets should
be in place to protect people from large, structural shifts in the economy. Absent
mitigating policies, the beneficiaries of these shifts may be a small group at the
upper stratum of the society.113
In the short run, education, re-training, and inventing new goods and services may
mitigate these effects. Longer term, the current social safety net may need to evolve into
better social services for everyone, such as healthcare and education, or a guaranteed
basic income. Indeed, countries such as Switzerland and Finland have actively considered
such measures. AI may be thought of as a radically different mechanism of wealth
creation in which everyone should be entitled to a portion of the world’s AI-produced
treasure.114 It is not too soon for social debate on how the economic fruits of
AI-technologies should be shared. As children in traditional societies support their
aging parents, perhaps our artificially intelligent “children” should support us, the
“parents” of their intelligence.

113 For example, Brynjolfsson and McAfee, Second Machine Age, have two chapters of devoted to
this (Erik Brynjolfsson and Andrew McAfee, The Second Machine Age: Work, Progress, and Prosperity in a
Time of Brilliant Technologies, (New York: W. W. Norton & Company, Inc., 2014)) and Brynjolfsson,
McAfee, and Spence describe policy responses for the combination of globalization and digital
technology (Erik Brynjolfsson, Andrew McAfee, and Michael Spence, Foreign Affairs, July/August
2014, accessed August 1, 2016, https://www.foreignaffairs.com/articles/united-states/
2014-06-04/new-world-order).
114 GDP does not do a good job of measuring the value of many digital goods. When society
can’t manage what isn’t measured, bad policy decisions result. One alternative is to look at
consumer surplus, not just dollar flows. As AI is embodied in more goods, this issue becomes more
salient. It may look like GDP goes down but people have better well-being through access to these
digital goods. See Erik Brynjolfsson and Adam Saunders, “What the GDP Gets Wrong (Why
Managers Should Care),” Sloan Management Review, vol. 51, no. 1 (October 1, 2009): 95–96. 39
 ENTERTAINMENT
With the explosive growth of the internet over the past fifteen years, few can imagine
their daily lives without it. Powered by AI, the internet has established user-generated
content as a viable source of information and entertainment. Social networks such
as Facebook are now pervasive, and they function as personalized channels of
social interaction and entertainment—sometimes to the detriment of interpersonal
interaction. Apps such as WhatsApp and Snapchat enable smart-phone users to
remain constantly “in touch” with peers and share sources of entertainment and
AI will increasingly enable information. In on-line communities such as Second Life and role-playing games such
entertainment that is more as World of Warcraft, people imagine an alternative existence in a virtual world.115
Specialized devices, such as Amazon’s Kindle have also redefined the essentials of
interactive, personalized, long-cherished pastimes. Books can now be browsed and procured with a few swipes
of the finger, stored by the thousands in a pocket-sized device, and read in much the
and engaging. Research same way as a handheld paperback.
should be directed toward Trusted platforms now exist for sharing and browsing blogs, videos, photos, and
topical discussions, in addition to a variety of other user-generated information. To
understanding how to operate at the scale of the internet, these platforms must rely on techniques that are
being actively developed in natural language processing, information retrieval, image
leverage these attributes processing, crowdsourcing, and machine learning. Algorithms such as collaborative
filtering have been developed, for example, to recommend relevant movies, songs, or
for individuals’ and articles based on the user’s demographic details and browsing history.116
society’s benefit. Traditional sources of entertainment have also embraced AI to keep pace with the
times. As exemplified in the book and movie Moneyball, professional sport is now subjected
to intensive quantitative analysis.117 Beyond aggregate performance statistics, on-field
signals can be monitored using sophisticated sensors and cameras. Software has been
created for composing music118 and recognizing soundtracks.119 Techniques from computer
vision and NLP have been used in creating stage performances.120 Even the lay user can
exercise his or her creativity on platforms such as WordsEye, which automatically
generates 3D scenes from natural language text.121 AI has also come to the aid of
historical research in the arts, and is used extensively in stylometry and, more recently,
in the analysis of paintings.122
The enthusiasm with which humans have responded to AI-driven entertainment
has been surprising and led to concerns that it reduces interpersonal interaction
among human beings. Few predicted that people would spend hours on end
interacting with a display. Children often appear to be genuinely happier playing
at home on their devices rather than outside with their friends. AI will increasingly
enable entertainment that is more interactive, personalized, and engaging. Research
should be directed toward understanding how to leverage these attributes for
individuals’ and society’s benefit.

115 Second Life, accessed August 1, 2016, http://secondlife.com; “World of Warcraft,” Blizzard
Entertainment, Inc, accessed August 1, 2016, http://us.battle.net/wow/en/.
116 John S. Breese, David Heckerman, and Carl Kadie, “Empirical Analysis of Predictive
Algorithms for Collaborative Filtering,” Proceedings of the 14th Conference on Uncertainty in Artificial
Intelligence (July 1998), accessed August 1, 2016, http://arxiv.org/pdf/1301.7363.pdf, 43–52.
117 Michael Lewis, Moneyball: The Art of Winning an Unfair Game (New York: W. W. Norton &
Company, Inc., 2003): http://www.imdb.com/title/tt1210166/).
118 MuseScore, accessed August 1, 2016, https://musescore.org/.
119 Shazam, accessed August 1, 2016, http://www.shazam.com/.
120 Annie Dorsen, accessed August 1, 2016, http://www.anniedorsen.com/.
121 WordsEye, accessed August 1, 2016, https://www.wordseye.com/.
122 “Stylometry,” Wikipedia, last modified August 4, 2016, accessed August 1, 2016, https://
40 en.wikipedia.org/wiki/Stylometry; http://arxiv.org/pdf/1408.3218v1.pdf.
Imagining the Future
The success of any form of entertainment is ultimately determined by the individuals
and social groups that are its subjects. The modes of entertainment that people find
appealing are diverse and change over time. It is therefore hard to predict the forms
entertainment will take in the next fifteen years precisely. Nevertheless, current trends
suggest at least a few features that the future entertainment landscape is likely to contain.
To date, the information revolution has mostly unfolded in software. However,
with the growing availability of cheaper sensors and devices, greater innovation in the
hardware used in entertainment systems is expected. Virtual reality and haptics could More sophisticated tools
enter our living rooms—personalized companion robots are already being developed.123
With the accompanying improvements in Automatic Speech Recognition, the Study
and apps will become
Panel expects that interaction with robots and other entertainment systems will become available to make it
dialogue-based, perhaps constrained at the start, but progressively more human-like.
Equally, the interacting systems are predicted to develop new characteristics such as even easier to produce
emotion, empathy, and adaptation to environmental rhythms such as time of day.124
Today, an amateur with a video camera and readily-available software tools can high-quality content, for
make a relatively good movie. In the future, more sophisticated tools and apps will
example, to compose
become available to make it even easier to produce high-quality content, for example,
to compose music or to choreograph dance using an avatar. The creation and music or to choreograph
dissemination of entertainment will benefit from the progress of technologies such as
ASR, dubbing, and Machine Translation, which will enable content to be customized dance using an avatar.
to different audiences inexpensively. This democratization and proliferation of AI-
created media makes it difficult to predict how humans’ taste for entertainment, which
are already fluid, will evolve.
With content increasingly delivered digitally, and large amounts of data being
logged about consumers’ preferences and usage characteristics, media powerhouses
will be able to micro-analyze and micro-serve content to increasingly specialized
segments of the population—down to the individual.125 Conceivably the stage is set
for the emergence of media conglomerates acting as “Big Brothers” who are able to
control the ideas and online experiences to which specific individuals are exposed.
It remains to be seen whether broader society will develop measures to prevent their
emergence. This topic, along with others pertaining to AI-related policy, is treated in
more detail in the next section.

123 Emoters, accessed August 1, 2016, http://emoterbots.com/.

124 “Siri,” Apple, Inc., accessed August 1, 2016, http://www.apple.com/in/ios/siri/.
125 Ryan Calo, “Digital Market Manipulation,” George Washington Law Review 82, no. 4 (2014):
995–1051. 41
 SECTION III: PROSPECTS AND
RECOMMENDATIONS FOR AI PUBLIC POLICY
The goal of AI applications must be to create value for society. Our policy recommendations
flow from this goal, and, while this report is focused on a typical North American city in
2030, the recommendations are broadly applicable to other places over time. Strategies that
enhance our ability to interpret AI systems and participate in their use may help build trust
The measure of success and prevent drastic failures. Care must be taken to augment and enhance human capabilities
and interaction, and to avoid discrimination against segments of society. Research to encourage
for AI applications is this direction and inform public policy debates should be emphasized. Given the current
sector-specific regulation of US industries, new or retooled laws and policies will be needed
the value they create to address the widespread impacts AI is likely to bring. Rather than “more” or “stricter”
regulation, policies should be designed to encourage helpful innovation, generate and transfer
for human lives. Going
expertise, and foster broad corporate and civic responsibility for addressing critical societal
forward, the ease with issues raised by these technologies. In the long term, AI will enable new wealth creation that
will require social debate on how the economic fruits of AI technologies should be shared.
which people use and
adapt to AI applications AI POLICY, NOW AND IN THE FUTURE
Throughout history, humans have both shaped and adapted to new technologies. This
will likewise largely report anticipates that advances in AI technologies will be developed and fielded
gradually—not in sudden, unexpected jumps in the techniques themselves—and will
determine their success.
build on what exists today, making this adaptation easier. On the other hand, small
improvements to techniques, computing power, or availability of data can occasionally
lead to novel, game-changing applications. The measure of success for AI applications
is the value they create for human lives. Going forward, the ease with which people
use and adapt to AI applications will likewise largely determine their success.
Conversely, since AI applications are susceptible to errors and failures, a mark of
their success will be how users perceive and tolerate their shortcomings. As AI becomes
increasingly embedded in daily lives and used for more critical tasks, system mistakes
may lead to backlash from users and negatively affect their trust. Though accidents
in a self-driving car may be less probable than those driven by humans, for example,
they will attract more attention. Design strategies that enhance the ability of humans
to understand AI systems and decisions (such as explicitly explaining those decisions),
and to participate in their use, may help build trust and prevent drastic failures. Likewise,
developers should help manage people’s expectations, which will affect their happiness
and satisfaction with AI applications. Frustration in carrying out functions promised
by a system diminishes people’s trust and reduces their willingness to use the system
in the future.
Another important consideration is how AI systems that take over certain tasks
will affect people’s affordances and capabilities. As machines deliver super-human
performances on some tasks, people’s ability to perform them may wither. Already,
introducing calculators to classrooms has reduced children’s ability to do basic
arithmetic operations. Still, humans and AI systems have complementary abilities.
People are likely to focus on tasks that machines cannot do as well, including complex
reasoning and creative expression.
Already, children are increasingly exposed to AI applications, such as interacting
with personal assistants on cell phones or with virtual agents in theme parks. Having
early exposure will improve children’s interactions with AI applications, which
will become a natural part of their daily lives. As a result, gaps will appear in how
younger and older generations perceive AI’s influences on society.
42
 Likewise, AI could widen existing inequalities of opportunity if access to AI
technologies—along with the high-powered computation and large-scale data that
fuel many of them—is unfairly distributed across society. These technologies will
improve the abilities and efficiency of people who have access to them. A person with
access to accurate Machine Translation technology will be better able to use learning
resources available in different languages. Similarly, if speech translation technology is
only available in English, people who do not speak English will be at a disadvantage.
Further, AI applications and the data they rely upon may reflect the biases of their
designers and users, who specify the data sources. This threatens to deepen existing AI could widen
social biases, and concentrate AI’s benefits unequally among different subgroups
of society. For example, some speech recognition technologies do not work well for existing inequalities of
women and people with accents. As AI is increasingly used in critical applications,
opportunity if access to
these biases may surface issues of fairness to diverse groups in society. On the other
hand, compared to the well-documented biases in human decision-making, AI-based AI technologies—along
decision-making tools have the potential to significantly reduce the bias in critical
decisions such as who is lent money or sent to jail. with the high-powered
Privacy concerns about AI-enabled surveillance are also widespread, particularly
in cities with pervasive instrumentation. Sousveillance, the recording of an activity
computation and large-
by a participant, usually with portable personal devices, has increased as well. Since scale data that fuel
views about bias and privacy are based on personal and societal ethical and value
judgments, the debates over how to address these concerns will likely grow and resist many of them—is unfairly
quick resolution. Similarly, since AI is generating significant wealth, debates will grow
regarding how the economic fruits of AI technologies should be shared—especially as distributed across society.
AI expertise and the underlying data sets that fuel applications are concentrated in a
small number of large corporations.
To help address these concerns about the individual and societal implications
of rapidly evolving AI technologies, the Study Panel offers three general policy
recommendations:
1. Define a path toward accruing technical expertise in AI at all
levels of government. Effective governance requires more experts
who understand and can analyze the interactions between AI
technologies, programmatic objectives, and overall societal values.
Absent sufficient technical expertise to assess safety or other metrics, national or
local officials may refuse to permit a potentially promising application. Or insufficiently
trained officials may simply take the word of industry technologists and green light a
sensitive application that has not been adequately vetted. Without an understanding
of how AI systems interact with human behavior and societal values, officials will be
poorly positioned to evaluate the impact of AI on programmatic objectives.
2. Remove the perceived and actual impediments to research on the
fairness, security, privacy, and social impacts of AI systems.
Some interpretations of federal laws such as the Computer Fraud and Abuse Act
and the anti-circumvention provision of the Digital Millennium Copyright Act are
ambiguous regarding whether and how proprietary AI systems may be reverse engineered
and evaluated by academics, journalists, and other researchers. Such research is critical
if AI systems with physical and other material consequences are to be properly vetted
and held accountable.
3. Increase public and private funding for interdisciplinary studies of
the societal impacts of AI.
As a society, we are underinvesting resources in research on the societal implications
of AI technologies. Private and public dollars should be directed toward interdisciplinary
43
 teams capable of analyzing AI from multiple angles. Research questions range from
basic research into intelligence to methods to assess and affect the safety, privacy,
fairness, and other impacts of AI.
Questions include: Who is responsible when a self-driven car crashes or an
intelligent medical device fails? How can AI applications be prevented from unlawful
discrimination? Who should reap the gains of efficiencies enabled by AI technologies
and what protections should be afforded to people whose skills are rendered obsolete?
As AI becomes integrated more broadly and deeply into industrial and consumer
As a society, we are products, it enters areas in which established regulatory regimes will need to be
adapted to AI innovations or in some cases fundamentally reconfigured according to
underinvesting resources broadly accepted goals and principles.
The approach in the United States to date has been sector-specific, with oversight
in research on the
by a variety of agencies. The use of AI in devices that deliver medical diagnostics and
societal implications of treatments is subject to aggressive regulation by the Food and Drug Administration
(FDA), both in defining what the product is and specifying the methods by which it is
AI technologies. Private produced, including standards of software engineering. The use of drones in regulated
airspace falls under the authority of the Federal Aviation Administration (FAA).126 For
and public dollars should consumer-facing AI systems, regulation by the Federal Trade Commission (FTC) comes
be directed toward into play. Financial markets using AI technologies, such as in high-frequency trading,
come under regulation by the Security Exchange Commission (SEC).
interdisciplinary teams In addition to sector-specific approaches, the somewhat ambiguous and broad
regulatory category of “critical infrastructure” may apply to AI applications.127 The
capable of analyzing AI Obama Administration’s Presidential Policy Directive (PPD) 21 broadly defines critical
infrastructure as composed of “the assets, systems, and networks, whether physical or
from multiple angles.
virtual, so vital to the United States that their incapacitation or destruction would have
a debilitating effect on security, national economic security, national public health or
safety, or any combination thereof.” Today, an enterprise does not come under federal
regulation solely by falling under that broad definition. Instead, the general trend of
federal policy is to seek regulation in sixteen sectors of the economy.128
As regards AI, critical infrastructure is notably defined by the end-user application,
and not the technology or sector that actually produces AI software.129 Software

126 FAA controls the ways drones fly, requires drones to be semiautonomous as opposed to
autonomous, requires visual connection to the drone, and enforces no-fly zones close to airports.
127 “Presidential Policy Directive (PPD-21)—Critical Infrastructure Security and Resilience,” The
White House, February 12, 2013, accessed August 1, 2016, https://www.whitehouse.gov/the-press-
office/2013/02/12/presidential-policy-directive-critical-infrastructure-security-and-resil.
128 PPD 21 identifies agencies responsible in each case. Chemical: Department of Homeland
Security; Commercial Facilities: Department of Homeland Security; Communications: Department
of Homeland Security; Critical Manufacturing: Department of Homeland Security; Dams:
Department of Homeland Security; Defense Industrial Base: Department of Defense; Emergency
Services: Department of Homeland Security; Energy: Department of Energy; Financial Services:
Department of the Treasury; Food and Agriculture: U.S. Department of Agriculture and
Department of Health and Human Services; Government Facilities: Department of Homeland
Security and General Services Administration; Healthcare and Public Health: Department of
Health and Human Services; Information Technology: Department of Homeland Security;
Nuclear Reactors, Materials, and Waste: Department of Homeland Security; Transportation
Systems: Department of Homeland Security and Department of Transportation; Water and
Wastewater Systems: Environmental Protection Agency.
129 In “ICYMI- Business Groups Urge White House to Rethink Cyber Security Order,” Internet
Association, March 5, 2013, accessed August 1, 2016, https://internetassociation.org/030513gov-3/:
“Obama’s Feb. 12 order says the government can’t designate ‘commercial information technology
products’ or consumer information technology services as critical U.S. infrastructure targeted for
voluntary computer security standards,’ ... ‘Obama’s order isn’t meant to get down to the level
of products and services and dictate how those products and services behave,’ said David LeDuc,
senior director of public policy for the Software & Information Industry Association, a Washington
44 trade group that lobbied for the exclusions.”
companies such as Google, Facebook, and Amazon have actively lobbied to avoid
being designated as critical to the economy, arguing that this would open the door to
regulation that would inevitably compromise their rapid product development cycles
and ability to innovate.130 Nonetheless, as the companies creating, operating, and
maintaining critical infrastructure use AI, interest will grow in regulating that software.
Some existing regulatory regimes for software safety (for example, the FDA’s
regulation of high consequence medical software) require specific software
engineering practices at the developer level. However, modern software systems
are often assembled from library components which may be supplied by multiple Absent sufficient technical
vendors, and are relatively application-independent. It doesn’t seem feasible or
desirable to subject all such developers to the standards required for the most critical, expertise to assess
rare applications. Nor does it seem advisable to allow unregulated use of such
safety or other metrics,
components in safety critical applications. Tradeoffs between promoting innovation
and regulating for safety are difficult ones, both conceptually and in practice. At a national or local officials
minimum, regulatory entities will require greater expertise going forward in order to
understand the implications of standards and measures put in place by researchers, may refuse to permit a
government, and industry.131
potentially promising
Policy and legal considerations application—or green light
While a comprehensive examination of the ways artificial intelligence (AI) interacts
with the law is beyond the scope of this inaugural report, this much seems clear: a sensitive application that
as a transformative technology, AI has the potential to challenge any number of
has not been adequately
legal assumptions in the short, medium, and long term. Precisely how law and policy
will adapt to advances in AI—and how AI will adapt to values reflected in law and vetted.
policy—depends on a variety of social, cultural, economic, and other factors, and
is likely to vary by jurisdiction.
American law represents a mixture of common law, federal, state, and local
statutes and ordinances, and—perhaps of greatest relevance to AI—regulations.
Depending on its instantiation, AI could implicate each of these sources of law. For
example, Nevada passed a law broadly permitting autonomous vehicles and instructed
the Nevada Department of Motor Vehicles to craft requirements. Meanwhile, the
National Highway Transportation Safety Administration has determined that a self-
driving car system, rather than the vehicle occupants, can be considered the “driver”
of a vehicle. Some car designs sidestep this issue by staying in autonomous mode
only when hands are on the wheel (at least every so often), so that the human driver
has ultimate control and responsibility. Still, Tesla’s adoption of this strategy did not
prevent the first traffic fatality involving an autonomous car, which occurred in June
of 2016. Such incidents are sure to influence public attitudes towards autonomous
driving. And as most people’s first experience with embodied agents, autonomous
transportation will strongly influence the public’s perception of AI.
Driverless cars are, of course, but one example of the many instantiations of
AI in services, products, and other contexts. The legal effect of introducing AI into
the provision of tax advice, automated trading on the stock market, or generating
medical diagnoses will also vary in accordance to the regulators that govern
these contexts and the rules that apply within them. Many other examples of AI
applications fall within current non-technology-specific policy, including predictive

130 Eric Engleman, “Google Exception in Obama’s Cyber Order Questioned as Unwise Gap,”
Bloomberg Technology, March 4, 2013, accessed August 1, 2016, http://www.bloomberg.com/news/
articles/2013-03-05/google-exception-in-obama-s-cyber-order-questioned-as-unwise-gap.
131 Ryan Calo, “The Case for a Federal Robotics Commission,” Brookings Report, September 15,
2014, accessed August 1, 2016, http://www.brookings.edu/research/reports2/2014/09/case-for-
federal-robotics-commission. 45
 policing, non-discriminatory loans, healthcare applications such as eldercare and
drug delivery, systems designed to interact with children (for example, autonomous
tutoring systems are required to respect laws in regard to balanced handling of
evolution vs. intelligent design), and interactive entertainment.
Given the present structure of American administrative law, it seems unlikely
that AI will be treated comprehensively in the near term. Nevertheless, it is possible
to enumerate broad categories of legal and policy issues that AI tends to raise in
various contexts.
As AI applications engage Privacy
Private information about an individual can be revealed through decisions and
in behavior that, were it predictions made by AI. While some of the ways that AI implicates privacy mirror
those of technologies such as computers and the internet, other issues may be unique
done by a human, would
to AI. For example, the potential of AI to predict future behavior based on previous
constitute a crime, courts patterns raises challenging questions. Companies already use machine learning to
predict credit risk. And states run prisoner details through complex algorithms to
and other legal actors will predict the likelihood of recidivism when considering parole. In these cases, it is a
technical challenge to ensure that factors such as race and sexual orientation are not
have to puzzle through being used to inform AI-based decisions. Even when such features are not directly
whom to hold accountable provided to the algorithms, they may still correlate strongly with seemingly innocuous
features such as zip code. Nonetheless, with careful design, testing, and deployment,
and on what theory. AI algorithms may be able to make less biased decisions than a typical person.
Anthropomorphic interfaces increasingly associated with AI raise novel privacy
concerns. Social science research suggests people are hardwired to respond to
anthropomorphic technology as though it were human. Subjects in one study
were more likely to answer when they were born if the computer first stated
when it was built.132 In another, they skipped sensitive questions when posed by an
anthropomorphic interface.133 At a basic level lies the question: Will humans continue
to enjoy the prospect of solitude in a world permeated by apparently social agents
“living” in our houses, cars, offices, hospital rooms, and phones?134
Innovation policy
Early law and policy decisions concerning liability and speech helped ensure the
commercial viability of the Internet. By contrast, the software industry arguably
suffers today from the decision of firms to pivot from open and free software to the
more aggressive pursuit of intellectual property protections, resulting in what some
have termed patent “thickets.” Striking the proper balance between incentivizing
innovation in AI while promoting cooperation and protection against third party
harm will prove a central challenge.
Liability (civil)
As AI is organized to directly affect the world, even physically, liability for harms
caused by AI will increase in salience. The prospect that AI will behave in ways
designers do not expect challenges the prevailing assumption within tort law that
courts only compensate for foreseeable injuries. Courts might arbitrarily assign
liability to a human actor even when liability is better located elsewhere for reasons
of fairness or efficiency. Alternatively, courts could refuse to find liability because
the defendant before the court did not, and could not, foresee the harm that the AI
caused. Liability would then fall by default on the blameless victim. The role

132 Youngme Moon, “Intimate Exchanges: Using Computers to Elicit Self-Disclosure from
Consumers,” Journal of Consumer Research 26, no. 4 (March 2000): 323–339.
133 Lee Sproull, Mani Subramani, Sara Kiesler , Janet H. Walker, and Keith Waters, “When the
Interface is a Face,” Human-Computer Interaction 11, no. 2 (1996): 97–124.
134 M. Ryan Calo, “People Can Be So Fake: A New Dimension to Privacy and Technology
46 Scholarship,” Penn State Law Review 114, no. 3 (2010): 809–855.
of product liability—and the responsibility that falls to companies manufacturing
these products— will likely grow when human actors become less responsible for the
actions of a machine.
Liability (criminal)
If tort law expects harms to be foreseeable, criminal law goes further to expect that
harms be intended. US law in particular attaches great importance to the concept of
mens rea—the intending mind. As AI applications engage in behavior that, were it
done by a human, would constitute a crime, courts and other legal actors will have to
puzzle through whom to hold accountable and on what theory. AI applications could
Agency
The above issues raise the question of whether and under what circumstances an increasingly shift
AI system could operate as the agent of a person or corporation. Already regulatory
investment from payroll
bodies in the United States, Canada, and elsewhere are setting the conditions under
which software can enter into a binding contract.135 The more AI conducts legally and income to capital
salient activities, the greater the challenge to principles of agency under the law.
Certification expenditure. Depending on
The very notion of “artificial intelligence” suggests a substitution for human skill
and ingenuity. And in many contexts, ranging from driving to performing surgery or
a state budget’s reliance
practicing law, a human must attain some certification or license before performing on payroll and income
a given task. Accordingly, law and policy will have to—and already does—grapple
with how to determine competency in an AI system. For example, imagine a tax, such a shift could be
robotics company creates a surgical platform capable of autonomously removing an
appendix. Or imagine a law firm writes an application capable of rendering legal destabilizing.
advice. Today, it is unclear from a legal perspective who in this picture would have to
pass the medical boards or legal bar, let alone where they would be required to do so.136
Labor
As AI substitutes for human roles, some jobs will be eliminated and new jobs will
be created. The net effect on jobs is ambiguous, but labor markets are unlikely to
benefit everyone evenly. The demand for some types of skills or abilities will likely
drop significantly, negatively affecting the employment levels and wages of people
with those skills.137 While the ultimate effects on income levels and distribution are not
inevitable, they depend substantially on government policies, on the way companies
choose to organize work, and on decisions by individuals to invest in learning new
skills and seeking new types of work and income opportunities. People who find their
employment altered or terminated as a consequence of advances of AI may seek
recourse in the legislature and courts. This may be why Littler Mendelson LLP—
perhaps the largest employment law firm in the world—has an entire practice group
to address robotics and artificial intelligence.
Taxation
Federal, state, and local revenue sources may be affected. Accomplishing a task using AI
instead of a person can be faster and more accurate—and avoid employment taxes.
As a result, AI applications could increasingly shift investment from payroll and
income to capital expenditure. Depending on a state budget’s reliance on payroll
and income tax, such a shift could be destabilizing. AI may also display different

135 Ian R. Kerr, “Ensuring the Success of Contract Formation in Agent-Mediated Electronic
Commerce,” Electronic Commerce Research 1 (2001): 183–202.
136 Ryan Calo, “Digital Agenda’s public discussion on ‘The effects of robotics on economics,
labour and society,’ ” Ausschuss Digitale Agenda, (Deutsche Bundestag: Ausschussdrucksache A-Drs.
18(24)102), June 22, 2016, accessed August 1, 2016, https://www.bundestag.de/blob/428266/195a
1cde8d5347849accbbe60ed91865/a-drs-18-24-102-data.pdf.
137 Brynjolfsson and McAfee, “Race Against the Machine: How the Digital Revolution is
Accelerating Innovation, Driving Productivity, and Irreversibly Transforming Employment and the
Economy (2011)”; Brynjolfsson and McAfee, Second Machine Age. 47
 “habits” than people, resulting in still fewer revenue sources. The many municipalities
relying on income from speeding or parking tickets will have to find alternatives
if autonomous cars can drop people off and find distance parking, or if they are
programmed not to violate the law. As a result, government bodies trying to balance
their budgets in light of advances in AI may pass legislation to slow down or alter the
course of the technology.
Politics
AI technologies are already being used by political actors in gerrymandering and
Like other technologies, targeted “robocalls” designed to suppress votes, and on social media platforms in the
form of “bots.”138 They can enable coordinated protest as well as the ability to predict
AI has the potential to be protests, and promote greater transparency in politics by more accurately pinpointing
who said what, when. Thus, administrative and regulatory laws regarding AI can be
used for good or nefarious
designed to promote greater democratic participation or, if ill-conceived, to reduce it.
purposes. A vigorous and This list is not exhaustive and focuses largely on domestic policy in the United
States, leaving out many areas of law that AI is likely to touch. One lesson that
informed debate about might be drawn concerns the growing disconnect between the context-specific way
in which AI is governed today and a wider consideration of themes shared by AI
how to best steer AI in technologies across industries or sectors of society. It could be tempting to create new
ways that enrich our lives institutional configurations capable of amassing expertise and setting AI standards
across multiple contexts. The Study Panel’s consensus is that attempts to regulate “AI”
and our society is an in general would be misguided, since there is no clear definition of AI (it isn’t any
one thing), and the risks and considerations are very different in different domains.
urgent and vital need. Instead, policymakers should recognize that to varying degrees and over time, various
industries will need distinct, appropriate, regulations that touch on software built
using AI or incorporating AI in some way. The government will need the expertise to
scrutinize standards and technology developed by the private and public sector, and
to craft regulations where necessary.

Guidelines for the future

Faced with the profound changes that AI technologies can produce, pressure for
“more” and “tougher” regulation is inevitable. Misunderstanding about what AI is
and is not, especially against a background of scare-mongering, could fuel opposition
to technologies that could benefit everyone. This would be a tragic mistake.
Regulation that stifles innovation, or relocates it to other jurisdictions, would be
similarly counterproductive.
Fortunately, principles that guide successful regulation of current digital
technologies can be instructive. A recent multi-year study comparing privacy
regulation in four European countries and the United States, for example, yielded
counter-intuitive results.139 Those countries, such as Spain and France, with strict
and detailed regulations bred a “compliance mentality” within corporations, which
had the effect of discouraging both innovation and robust privacy protections.
Rather than taking responsibility for privacy protection internally and developing a
professional staff to foster it in business and manufacturing processes, or engaging with
privacy advocates or academics outside their walls, these companies viewed privacy as
a compliance activity. Their focus was on avoiding fines or punishments, rather than
proactively designing technology and adapting practices to protect privacy.
By contrast, the regulatory environment in the United States and Germany,
which combined more ambiguous goals with tough transparency requirements and
meaningful enforcement, were more successful in catalyzing companies to view
138 Political Bots, accessed August 1, 2016, http://politicalbots.org/.
139 Kenneth A. Bamberger and Deirdre K. Mulligan, Privacy on the Ground: Driving Corporate Behavior
48 in the United States and Europe (Cambridge, Massachusetts: MIT Press, 2015).
privacy as their responsibility. Broad legal mandates encouraged companies to
develop a professional staff and processes to enforce privacy controls, engage with
outside stakeholders, and to adapt their practices to technology advances. Requiring
greater transparency enabled civil society groups and media to become credible
enforcers both in court and in the court of public opinion, making privacy more
salient to corporate boards and leading them to further invest in privacy protection.
In AI, too, regulators can strengthen a virtuous cycle of activity involving internal
and external accountability, transparency, and professionalization, rather than narrow
compliance. As AI is integrated into cities, it will continue to challenge existing
protections for values such as privacy and accountability. Like other technologies, AI
has the potential to be used for good or nefarious purposes. This report has tried to
highlight the potential for both. A vigorous and informed debate about how to best
steer AI in ways that enrich our lives and our society, while encouraging creativity
in the field, is an urgent and vital need. Policies should be evaluated as to whether
they democratically foster the development and equitable sharing of AI’s benefits,
or concentrate power and benefits in the hands of a fortunate few. And since future
AI technologies and their effects cannot be foreseen with perfect clarity, policies will
need to be continually re-evaluated in the context of observed societal challenges and
evidence from fielded systems.
As this report documents, significant AI-related advances have already had
an impact on North American cities over the past fifteen years, and even more
substantial developments will occur over the next fifteen. Recent advances are largely
due to the growth and analysis of large data sets enabled by the Internet, advances
in sensory technologies and, more recently, applications of “deep learning.” In the
coming years, as the public encounters new AI applications in domains such as
transportation and healthcare, they must be introduced in ways that build trust and
understanding, and respect human and civil rights. While encouraging innovation,
policies and processes should address ethical, privacy, and security implications, and
should work to ensure that the benefits of AI technologies will be spread broadly and
fairly. Doing so will be critical if Artificial Intelligence research and its applications
are to exert a positive influence on North American urban life in 2030 and beyond.

49
 APPENDIX I: A SHORT HISTORY OF AI
This Appendix is based primarily on Nilsson’s book140 and written from the prevalent
current perspective, which focuses on data intensive methods and big data. However
important, this focus has not yet shown itself to be the solution to all problems. A
complete and fully balanced history of the field is beyond the scope of this document.
The field of Artificial Intelligence (AI) was officially born and christened at a 1956
workshop organized by John McCarthy at the Dartmouth Summer Research Project
on Artificial Intelligence. The goal was to investigate ways in which machines could
The field of Artificial be made to simulate aspects of intelligence—the essential idea that has continued to
Intelligence (AI) was drive the field forward. McCarthy is credited with the first use of the term “artificial
intelligence” in the proposal he co-authored for the workshop with Marvin Minsky,
officially born and Nathaniel Rochester, and Claude Shannon.141 Many of the people who attended soon
led significant projects under the banner of AI, including Arthur Samuel, Oliver
christened at a 1956 Selfridge, Ray Solomonoff, Allen Newell, and Herbert Simon.
workshop. The goal was to Although the Dartmouth workshop created a unified identity for the field and
a dedicated research community, many of the technical ideas that have come to
investigate ways in which characterize AI existed much earlier. In the eighteenth century, Thomas Bayes
provided a framework for reasoning about the probability of events.142 In the
machines could be made nineteenth century, George Boole showed that logical reasoning—dating back to
Aristotle—could be performed systematically in the same manner as solving a system
to simulate aspects of of equations.143 By the turn of the twentieth century, progress in the experimental
intelligence—the essential sciences had led to the emergence of the field of statistics,144 which enables
inferences to be drawn rigorously from data. The idea of physically engineering a
idea that has continued to machine to execute sequences of instructions, which had captured the imagination
of pioneers such as Charles Babbage, had matured by the 1950s, and resulted in the
drive the field forward. construction of the first electronic computers.145 Primitive robots, which could
sense and act autonomously, had also been built by that time.146
The most influential ideas underpinning computer science came from Alan Turing,
who proposed a formal model of computing. Turing’s classic essay, Computing Machinery
and Intelligence,147 imagines the possibility of computers created for simulating intelligence
and explores many of the ingredients now associated with AI, including how intelligence
might be tested, and how machines might automatically learn. Though these ideas
inspired AI, Turing did not have access to the computing resources needed to
translate his ideas into action.
Several focal areas in the quest for AI emerged between the 1950s and the 1970s.148
140 Nilsson, The Quest for Artificial Intelligence.
141 J. McCarthy, Marvin L. Minsky, Nathaniel Rochester, and Claude E. Shannon, “A Proposal
for the Dartmouth Summer Research Project on Artificial Intelligence,” August 31, 1955, accessed
August 1, 2016, http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html.
142 Thomas Bayes, “An Essay towards Solving a Problem in the Doctrine of Chances,”
Philosophical Transactions of the Royal Society of London 53 (January 1, 1763): 370–418, accessed August 1,
2016, http://rstl.royalsocietypublishing.org/search?fulltext=an+essay+towards+solving&submit=yes
&andorexactfulltext=and&x=0&y=0.
143 George Boole, An Investigation of the Laws of Thought on Which are Founded the Mathematical Theories
of Logic and Probabilities, (Macmillan, 1854, reprinted with corrections, Dover Publications, New
York, NY, 1958, and reissued by Cambridge University Press, 2009), accessed August 1, 2016,
http://ebooks.cambridge.org/ebook.jsf ?bid=CBO9780511693090.
144 “History of statistics,” Wikipedia, Last modified June 3, 2016, accessed August 1, 2016,
https://en.wikipedia.org/wiki/History_of_statistics.
145 Joel N. Shurkin, Engines of the Mind: The Evolution of the Computer from Mainframes to Microprocessors
(New York: W. W. Norton & Company, 1996).
146 William Grey Walter, “An Electromechanical Animal,” Dialectica 4 (1950): 42–49.
147 A. M. Turing, “Computing Machinery and Intelligence,” Mind 59, no. 236 (1950): 433–460.
148 Marvin Minsky, “Steps toward Artificial Intelligence,” MIT Media Laboratory, October 24,
50 1960, accessed August 1, 2016, http://web.media.mit.edu/~minsky/papers/steps.html.
Newell and Simon pioneered the foray into heuristic search, an efficient procedure
for finding solutions in large, combinatorial spaces. In particular, they applied this
idea to construct proofs of mathematical theorems, first through their Logic Theorist
program, and then through the General Problem Solver.149 In the area of computer
vision, early work in character recognition by Selfridge and colleagues150 laid the
basis for more complex applications such as face recognition.151 By the late sixties,
work had also begun on natural language processing.152 “Shakey”, a wheeled
robot built at SRI International, launched the field of mobile robotics. Samuel’s
Checkers-playing program, which improved itself through self-play, was one of the Although the separation
first working instances of a machine learning system.153 Rosenblatt’s Perceptron,154
a computational model based on biological neurons, became the basis for the field of AI into sub-fields
of artificial neural networks. Feigenbaum and others advocated 155the case for
has enabled deep
building expert systems—knowledge repositories tailored for specialized domains
such as chemistry and medical diagnosis.156 technical progress along
Early conceptual progress assumed the existence of a symbolic system that could
be reasoned about and built upon. But by the 1980s, despite this promising headway several different fronts,
made into different aspects of artificial intelligence, the field still could boast no
significant practical successes. This gap between theory and practice arose in part from
synthesizing intelligence
an insufficient emphasis within the AI community on grounding systems physically, with at any reasonable scale
direct access to environmental signals and data. There was also an overemphasis on
Boolean (True/False) logic, overlooking the need to quantify uncertainty. The field invariably requires many
was forced to take cognizance of these shortcomings in the mid-1980s, since interest
in AI began to drop, and funding dried up. Nilsson calls this period the “AI winter.” different ideas to be
A much needed resurgence in the nineties built upon the idea that “Good Old-
integrated.
Fashioned AI”157 was inadequate as an end-to-end approach to building intelligent
systems. Rather, intelligent systems needed to be built from the ground up, at all times
solving the task at hand, albeit with different degrees of proficiency.158 Technological
progress had also made the task of building systems driven by real-world data more
feasible. Cheaper and more reliable hardware for sensing and actuation made robots
easier to build. Further, the Internet’s capacity for gathering large amounts of data,
and the availability of computing power and storage to process that data, enabled
statistical techniques that, by design, derive solutions from data. These developments
have allowed AI to emerge in the past two decades as a profound influence on our
daily lives, as detailed in Section II.
149 Allen Newell, John C Shaw, and Herbert A Simon, “Report on a general problem-solving
program,” Proceedings of the International Conference on Information Processing, UNESCO, Paris 15-20
June 1959 (Unesco/Oldenbourg/Butterworths, 1960), 256–264.
150 O. G. Selfridge, “Pandemonium: A paradigm for learning,” Proceedings of the Symposium on
Mechanization of Thought Processes (London: H. M. Stationary Office, 1959): 511–531.
151 Woodrow W. Bledsoe and Helen Chan, “A Man-Machine Facial Recognition System: Some
Preliminary Results,” Technical Report PRI 19A (Palo Alto, California: Panoramic Research, Inc., 1965).
152 D. Raj Reddy, “Speech Recognition by Machine: A Review,” Proceedings of the IEEE 64, no.4
(April 1976), 501–531.
153 Arthur Samuel, “Some Studies in Machine Learning Using the Game of Checkers, IBM
Journal of Research and Development 3, no. 3 (1959): 210—229.
154 Frank Rosenblatt, “The Perceptron—A Perceiving and Recognizing Automaton,”
Report 85-460-1, (Buffalo, New York: Cornell Aeronautical Laboratory, 1957).
155 “Shakey the robot,” Wikipedia, last modified July 11, 2016, accessed August 1, 2016,
https://en.wikipedia.org/wiki/Shakey_the_robot.
156 Edward A. Feigenbaum and Bruce G. Buchanan, “DENDRAL and Meta-DENDRAL:
Roots of Knowledge Systems and Expert System Applications,” Artificial Intelligence 59, no. 1-2
(1993), 233–240.
157 John Haugeland, Artificial Intelligence: The Very Idea, (Cambridge, Massachusetts: MIT
Press, 1985).
158 Rodney A. Brooks, “Elephants Don’t Play Chess,” Robotics and Autonomous Systems 6, no. 1-2
(June 1990): 3–15. 51
 In summary, following is a list of some of the traditional sub-areas of AI. As described
in Section II, some of them are currently “hotter” than others for various reasons.
But that is neither to minimize the historical importance of the others, nor to say that
they may not re-emerge as hot areas in the future.
• Search and Planning deal with reasoning about goal-directed behavior. Search
plays a key role, for example, in chess-playing programs such as Deep Blue, in
deciding which move (behavior) will ultimately lead to a win (goal).
• The area of Knowledge Representation and Reasoning involves processing
information (typically when in large amounts) into a structured form that can
be queried more reliably and efficiently. IBM’s Watson program, which beat
human contenders to win the Jeopardy challenge in 2011, was largely based on
an efficient scheme for organizing, indexing, and retrieving large amounts of
information gathered from various sources.159
• Machine Learning is a paradigm that enables systems to automatically improve
their performance at a task by observing relevant data. Indeed, machine
learning has been the key contributor to the AI surge in the past few decades,
ranging from search and product recommendation engines, to systems for
speech recognition, fraud detection, image understanding, and countless other
tasks that once relied on human skill and judgment. The automation of these
tasks has enabled the scaling up of services such as e-commerce.
• As more and more intelligent systems get built, a natural question to consider
is how such systems will interact with each other. The field of Multi-Agent
Systems considers this question, which is becoming increasingly important in
on-line marketplaces and transportation systems.
• From its early days, AI has taken up the design and construction of systems that
are embodied in the real world. The area of Robotics investigates fundamental
aspects of sensing and acting—and especially their integration—that enable
a robot to behave effectively. Since robots and other computer systems share
the living world with human beings, the specialized subject of Human Robot
Interaction has also become prominent in recent decades.
• Machine perception has always played a central role in AI, partly in developing
robotics, but also as a completely independent area of study. The most commonly
studied perception modalities are Computer Vision and Natural Language
Processing, each of which is attended to by large and vibrant communities.
• Several other focus areas within AI today are consequences of the growth of the
Internet. Social Network Analysis investigates the effect of neighborhood relations
in influencing the behavior of individuals and communities. Crowdsourcing is
yet another innovative problem-solving technique, which relies on harnessing
human intelligence (typically from thousands of humans) to solve hard
computational problems.
Although the separation of AI into sub-fields has enabled deep technical progress
along several different fronts, synthesizing intelligence at any reasonable scale invariably
requires many different ideas to be integrated. For example, the AlphaGo program160 161
that recently defeated the current human champion at the game of Go used multiple
machine learning algorithms for training itself, and also used a sophisticated search
procedure while playing the game.
159 David A. Ferrucci, “Introduction to ‘This is Watson,’” IBM Journal of Research and Development,
56, no. 3-4 (2012): 1.
160 David Silver et al., “Mastering the Game of Go with Deep Neural Networks and Tree Search.”
161 Steven Borowiec and Tracey Lien, “AlphaGo beats human Go champ in milestone for
artificial intelligence,” Los Angeles Times, March 12, 2016, accessed August 1, 2016, http://www.
latimes.com/world/asia/la-fg-korea-alphago-20160312-story.html.
52
 Journal of Global Fashion Marketing
Bridging Fashion and Marketing

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/rgfm20

The effect of gender on willingness to pay for mass

customised running shoes

Hassan Daronkola Kalantari, Lester W Johnson & Chamila R. Perera

To cite this article: Hassan Daronkola Kalantari, Lester W Johnson & Chamila R. Perera (2021)
The effect of gender on willingness to pay for mass customised running shoes, Journal of
Global Fashion Marketing, 12:2, 161-175, DOI: 10.1080/20932685.2020.1867878

To link to this article: https://doi.org/10.1080/20932685.2020.1867878

View supplementary material

Published online: 19 Feb 2021.

Submit your article to this journal

Article views: 528

View related articles

View Crossmark data

Citing articles: 4 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rgfm20
JOURNAL OF GLOBAL FASHION MARKETING
2021, VOL. 12, NO. 2, 161–175
https://doi.org/10.1080/20932685.2020.1867878

The effect of gender on willingness to pay for mass

customised running shoes
Hassan Daronkola Kalantari , Lester W Johnson and Chamila R. Perera
Department of Management and Marketing, Swinburne University of Technology, Hawthorn, Australia

ABSTRACT ARTICLE HISTORY

We investigate whether gender and different levels of customisation Received 10 August 2020
have a significant effect on consumers’ willingness to purchase mass Revised 27 November 2020
customised running shoes. An online panel survey was utilised to Accepted 18 December 2020
collect data from 353 Australian adults who wear running shoes at KEYWORDS
least once a month. Three attributes – degree of customisation, price, Mass customisation; conjoint
and delivery time were utilised in the data analysis to explore will­ analysis; gender; purchase
ingness to pay for customised shoes. Informed by conjoint analysis and intention; willingness to Pay
t-tests, the study makes a significant theoretical contribution by
extending the understanding of inconveniences of mass customisation
from the perspective of customers’ willingness across genders. It is
found that for women, degree of customisation and delivery time are
the most important attributes, while for men, price and degree of
customisation are the more crucial attributes. Female customers are
more willing to purchase the product than male customers. The study
addresses the research gap that is how differently females and males
who usually have different sensory perceptions respond to mass-
customised products. Further, the study provides valuable strategic
insights for both manufacturers and marketers to cater fragmented
consumer markets through mass customisation by identifying subtle
differences in customer readiness among their target group of
customers.

1. Introduction
One size fit all, the age-old Henry Ford philosophy of production has been clearly taken
over by mass customisation, a computer-aided manufacturing and marketing technique
which produces custom outputs. Previous studies show the relationships between mass
customisation and consumer behaviour in general, such as experience (Liang & Liu,
2019), value (Adel & Younis, 2019; Lang et al., 2020; Winters & Ha, 2012), satisfaction
(Kasiri et al., 2017) and identity-expressions (Ling et al., 2020; Ribeiro et al., 2017).
Focusing on more specific areas of consumer behaviour, Jessie et al. (2020) find an
association between male consumers’ characteristics (age, education, and household
income) and purchase intentions of mass-customised apparel. Clearly, these findings
provide opportunities for businesses to market mass customised products more effec­
tively. However, to better inform strategies of mass customisation, it is essential to

CORRESPONDENCE TO Hassan Daronkola Kalantari [email protected] Swinburne University of

Technology, Hawthorn, VIC 3122, Australia
Supplemental data for this article can be accessed here.
© 2021 Korean Scholars of Marketing Science
162 H. D. KALANTARI ET AL.

investigate how different consumer groups respond to mass customised products and
services, albeit, gender differences have been largely overlooked in this field of study
except for a few investigations (e.g. Seo & Lang, 2019; Walcher et al., 2016). Our study
wishes to address this research gap.
Among the few recent investigations, Seo and Lang (2019) find that while both
genders seek for uniqueness when purchasing mass customised products, males are
more driven by social identity and females are more inspired by self-promotion.
Considering the differences, the effect of gender on mass customised product purchase
behaviour needs to be further investigated.
Footwear is among the top product categories in the retail sector that customers are
willing to purchase in a customised form (Aichner & Coletti, 2013; YouGov, 2018).
However, using Adidas as a case study, Zhao et al. (2019) found that individual differ­
ences is a more determinant factor which stops consumers from purchasing mass
customised products. Mass customised footwear firms seem not to be at their best in
Australia (e.g. failure of Shoes of Prey). Given this background, this study investigates to
what extent gender differences influence consumers’ willingness to purchase mass cus­
tomised products. In other words, how does gender and different levels of customisation
influence consumers’ willingness to purchase mass customised running shoes? Highly
fragmented markets such as fashions and shoes increasingly welcome mass customised
products. The investigation provides firms operate in the markets with strategic insights
to develop their marketing strategies in line with consumers’ willingness to purchase
mass customised products.

2. Literature review
2.1. Theoretical background
The concept of mass customisation was first developed by Davis (1987) who proposed
that mass customised products and services can be economically feasible to mass
produce. Later, Pine (1993) claimed mass customisation is an ability to provide uniquely
designed products and services to individual customers by using high process flexibility
and integration. Today, with the development of new technologies such as 3D printers,
computer numerical control (CNC) machines, robots, virtual reality and machine learn­
ing, anything that can be digitalised can be customised. This mega-trend is increasingly
embraced by automobile, food, and fashion manufacturers alike (PWC, 2019).
This study draws theoretical insights from the framework developed by Bardakci and
Whitelock (2004) who propose three main inconveniences of mass customisation that
can be used to examine the extent to which customers are ready for mass customisation;
premium price, delivery time, and product designing time. Overall, the framework
suggests that customers’ readiness may be determined by estimating their willingness
to pay a premium price, wait to receive the customised product and invest time in
designing the product.
Customers usually expect faster delivery, reasonable price, and higher product varia­
bility (Purnomo & Shinta, 2019) and longer delivery time and premium prices are found
to be the greatest disadvantage of mass customisation (Jost & Süsser, 2019). Moreover,
consumers are willing to pay extra to have their products customised (Protolabs., 2019)
 JOURNAL OF GLOBAL FASHION MARKETING 163

as it makes their products special and unique (Jafari et al., 2015). Kalantari and Johnson
(2018) find that consumers are even willing to trade-off between price, delivery waiting
time and type of customisation. These debates in the literature clearly call for further
investigations into mass customisation.
According to Bardakci and Whitelock (2004), customers’ readiness (i.e. willing to pay,
wait and spend time to design) is a significant determinant of evincing a positive will­
ingness to put up with the inconveniences (i.e. premium price, delivery time, and product
designing time). Further investigation into the effect of gender on willingness to pay for
different levels of mass customised products can extend the theoretical parameters of
Bardakci and Whitelock (2004) framework, thus it is considered the focus of the present
investigation.

2.2. Mass customisation and the level of customisation

In 1993, Pine proposes theory of mass customisation which explains how affordable
products can be made through a combination of production efficiency and customisa­
tion. However, Levitt and Levitt (1998) argues that consumers are willing to sacrifice
their preferences towards features of a product for a lower-priced product. As confirmed
by more recent research, consumers even look for products that fit into their needs and
requirements and are willing to pay extra for those products (e.g. Jessie et al., 2020;
Kalantari & Johnson, 2018).
In addition to meeting consumer preferences, previous research finds an association
between consumers’ satisfaction and the interactivity of firms’ web sites and visual
aspects that provide them with an opportunity to customise the products they purchase
(e.g. H.-H. Lee et al., 2011; Zhao et al., 2019). Clearly, consumers who purchase mass
customised products seek for more opportunities of customisations.
Following the well cited researchers, Gilmore and Pine (1997), the present study
outlines three levels of customisation. The first level refers to a standard mass-
produced product and it is called pure standardisation, a strategy which can still be
useful in many industrial segments. When making a shoe purchase, the shoes that
consumers buy in normal shoe stores are a classic example of a standard product.
The second level refers to a customised product wherein the customer can choose one
out of several options, such as the colour of the product (e.g. from blue, green, or red).
NikeiD is a notable example of this category. NikeiD is a service provided by Nike which
allows customers to customise their shoes and effectively become co-designers of their
apparel.
The third level refers to full customisation (fully customised) where the product is
based on the customer’s requirements and measurements in addition to customisation
features. That is, not only is the customer able to choose the colour and fabric of the
shoes, but also the shoes are made to perfectly fit the customer’s measured feet. Shoes of
Prey is a standout example of fully customised shoes. Here, customers can purchase
custom-made shoes at non-luxury prices. They can choose the shape, colour, height of
the shoes, size, and multiple widths.
From both manufacturers’ and marketers’ point of view, mass customisation is an
opportunity to reach multiple segments in a market with a lower cost of production
(Jablonsky, 2018). Moreover, based on mass customisation preferences, Pallant et al.
164 H. D. KALANTARI ET AL.

(2020) profile four distinct segments of consumers: Non-Customisers, New Customisers,

Active Customisers, and Lapsing Customisers. This is, however, debatable as some recent
studies claim that due to the early development of mass customisation and the limitations
of some recent manufacturing technologies, most customised products are more expen­
sive than mass produced products (Jost & Süsser, 2019). Therefore, it can be argued that
mass customisation may fail to deliver what it promised at its inception. A further
exploration of this field of study is therefore necessary.

2.3. Gender specific decision-making process

Consumer decision making processes involve a problem recognition, information search,
alternative evaluation, purchase, and post-purchase evaluation (Belch & Belch, 2016).
Sanz de Acedo et al. (2009) deconstruct the essence of decision making into categories
such as uncertainty, time/money constraints, information/goals, consequences of deci­
sion, motivation, self-regulation, emotions, cognition, social pressure, and work pressure.
All these factors coalesce, leading a person to decide on whether to make a purchase.
Previous research often finds evidence that decision-making can be gender specific
(Alonso-Almeida & Bremser, 2015; Bae & Miller, 2009). Unlike females, males are
reported to adopt a simper decision-making style, spend little time deciding on the
products they buy, visit the same stores each time they shop and usually buy the lower-
priced products (Bakewell & Mitchell, 2004, 2006). Confirmed by a more recent study
(Lee, 2020), male shoppers can be more shopping goal-oriented than female shoppers.
Many studies find that females perceive greater risks than males in making decisions
regarding finance, health, and the environment (Bacha & Azouzi, 2019; Wolf & Zhang,
2016). On the contrary, recent research in a corporate setting finds that females are more
ethical but not more risk-averse decision makers than their male counterparts (Doan &
Iskandar-Datta, 2020). Nevertheless, in online purchase settings, gender differences
clearly exist to influence interactivity, vividness, diagnosticity, and perceived risk and
online purchase intentions (Lin et al., 2019).
The notion of perceived control has been discussed in the consumer behaviour
literature for a long time. According to White (1959), perceived control shows
a person’s competence, mastery and superiority over their environment. To this end,
three types of control are proposed by Averill (1973). These include cognitive, decisional,
and behavioural control. Based on Schutz (1966), control is one of the three kinds of
interpersonal needs which drive human social behaviours. That is, having a feeling of
control is an important determinant of satisfactory interactions with other humans.
Perception of control increases dramatically by increasing the number of choices
available (Botti et al., 2003). Therefore, given that mass customisation arguably offers
more choices, it can be assumed that allowing consumers to customise their products
can significantly increase their perceived levels of control, and in turn increase levels of
buying customised products (Marathe & Sundar, 2011) and satisfaction with the
customised product (Dellaert & Dabholkar, 2009). It is, therefore, intriguing to inves­
tigate whether customised, and fully customised products are more appealing to
females. This is an interesting area to investigate especially when inconsistent results
co-exist in the literature regarding how females deal with uncertainty and control (e.g.
Doan & Iskandar-Datta, 2020; Lin et al., 2019; Singh et al., 2020) while some studies
 JOURNAL OF GLOBAL FASHION MARKETING 165

find that gender differences trigger different consumer behaviours in males and females
(e.g. Arsel et al., 2015; Maclaran & Kravets, 2018) especially in online purchasing (Lin
et al., 2019).

2.4. Willingness to pay

Paying more for a customised product is a common tendency shared among online and
offline consumers alike (Kalantari & Johnson, 2018; Gilal et al., 2018). Several studies
investigate how much extra customers are willing to pay for a customised product (e.g.
J. Chen-Yu & Yang, 2018; Franke et al., 2010; Li & Unger, 2012; Seiji & Doris, 2008).
A survey of 200 respondents reveals that customers are willing to pay a premium price
for customised news and financial services (Li & Unger, 2012). According to a series of
different experiments on the value customisation created for consumers in the diverse
goods categories of newspapers, cereals, kitchens, skis, and fountain pens, consumers are
willing to pay 15% more for a customised newspaper, 40% more for a customised
fountain pen, 37% more for a customised kitchen, 34% more for customised skis and
50% more for customised cereals (Franke et al., 2010). However, the researchers focused
only on customers’ willingness to pay for mass customisation. They did not consider the
effect of the delivery time and time spent on designing a customised product on
customers’ willingness to buy customised products.
More recently, several other studies (e.g. Aichner & Coletti, 2013; Kalantari &
Johnson, 2019), assess customer willingness to choose, pay more and wait for mass
customised products. Although these researchers examine two important customer
readiness factors for mass customisation, they apparently do not consider the interaction
between price and delivery time as well as customer’s willingness to spend time on
designing customised products.
Some studies (e.g. Bardakci & Whitelock, 2004; Y. Lee et al., 2012) investigate mass
customised products while considering all three customer readiness factors (i.e. estimat­
ing their willingness to pay a premium price, wait to receive the customised product and
invest time in designing the product). Y. Lee et al. (2012) find that clothing fit is the most
important consideration for females in making mass produced apparel purchase deci­
sions and younger consumers are willing to wait longer for delivery of the customised
products. Overall, it can be assumed that many consumers are willing to spend time to
design their mass customised products.

3. Methodology
The footwear industry is reported to have an enormous potential for mass customisation.
Globenewswire (2019) forecasts that the 3D printed footwear industry can represent
1.5% of overall global footwear revenues by 2029. Mass customisation has been a key area
in the recent literature in the shoe fashion industry (Ribeiro et al., 2017; Runfola et al.,
2018). Footwear is one of the top product categories in which customers are willing to
purchase in a customised form (Aichner & Coletti, 2013). Thus, running shoes are
considered a suitable category in which to investigate the effect of gender on willingness
to pay for a different level of mass customised product.
166 H. D. KALANTARI ET AL.

An online survey was run using a panel sample provided by Opinio that comprised
a national sample from Australia. Email invitations to participate in the study were
individually sent. Individual panel members were screened for participation in the survey
by two questions requiring them to be adults over the age of 18 and who wear running
shoes at least once a month. Qualified individuals were then invited to participate in the
survey via a link. A total of 353 participants completed the survey.
The data were analysed using conjoint analysis (Green & Rao, 1971; Rao, 2014).
Conjoint analysis is a well-known method which allows inclusion and combinations of
a number of attributes to describe a hypothetical situation in which participants evaluate
the situation rather than evaluating each attribute individually. This method makes
preference statements more realistic. Conjoint analysis first was proposed by Green
and Rao (1971). There is growing interest in the use of conjoint analysis as a tool for
estimating customer preferences within the economic evaluation of different products
and services. It has been suggested that where cost is included as one of the attributes
within the exercise, conjoint analysis can be used to estimate willingness to pay
(Kulshreshtha et al., 2018). There are many studies that have used conjoint analysis to
estimate the willingness to pay (e.g. Tokimatsu et al., 2016).

3.1. Questionnaire
A short introduction explained the terms and concept of the survey and used to place
participants in the context of purchasing running shoes. Then, four questions were asked
to address the following socio-demographic information: gender, age, income, and state
of residence.
Attributes in this study were chosen according to customer readiness for mass
customisation. To determine customer willingness to purchase a customised product
compared to a standard one, the study considered the degree of customisation as an
attribute in the conjoint analysis. The second and third attributes were the price of the
product and the delivery time. Based on Bardakci and Whitelock (2004), to estimate the
customer readiness to buy a customised product, first, the study must estimate consu­
mers’ willingness to pay for the product, the length of time waiting to receive the item
and the time invested in designing the product. Hence, the study used three attributes for
this study: price, delivery time and the degree of customisation (standard, customised
and fully customised). As the time for designing the customised and fully customised
products were constant, the study defined a designing time for each customised and fully
customised product and put it in the description of each product in the questionnaire. It
was assumed that the participants were aware of the designing time and willing to spend
that amount of time for designing their customised product. Then, through brainstorm­
ing, reviewing the literature, and web-browsing, the study provided the most relevant
levels for each attribute. Table 1 illustrates the attributes and levels for this conjoint study.

Table 1. Attributes and levels for conjoint study.

Attribute Levels
Degree of Customisation Standard, Customised, Fully-Customised
Price $120, $140, $160, $180
Delivery time 4 Days, 1.5 Weeks, 3 Weeks, 4.5 Weeks
 JOURNAL OF GLOBAL FASHION MARKETING 167

Based on the number of attributes and levels, 48 possible profiles (3 × 4 × 4) can be

created. To make the task more viable for respondents, one way to reduce the number of
profiles is to use a fractional factorial design (in our case a one-third fraction) such as an
orthogonal design, which uses a fraction of all possible profiles (Rao, 2014). Orthogonal
designs consider only the main effects of the attribute levels and consider their interac­
tions unimportant. In this study we used SPSS – Orthogonal Design procedure
(ORTHOPLAN) – automatically generates main-effects orthogonal fractional factorial
plans, known as orthogonal arrays (Kuzmanovic et al., 2011).
By using an orthogonal design, the number of profiles that respondents need to rate is
significantly reduced to 19 profiles where 3 of them were holdouts (Appendix 1).
Holdouts help to determine the predictive power of the model and validate the result
of the conjoint analysis (Rao, 2014). As per previous research (e.g. Wittink et al., 1990), it
is possible that 100-point rating shows better result, hence a scale of 0 to 100, with 1 being
“Definitely do not purchase” and 100 being “Definitely purchase”, was presented for each
profile, asking participants to indicate their preferred estimated rating. A pilot study was
conducted to pre-test the questionnaire, sampling 40 students from an Australian uni­
versity. The pre-test confirmed that the questions and instructions were clear and under­
standable. The feedback from participants also showed that the survey was easy to follow
and complete. In conclusion, there were no major changes to the survey, its questions, or
instructions.

3.2. Data analysis

The part-worth model in conjoint method estimates three functions of U1(X1), U2(X2)
and U3(X3) respectively, for the three attributes of X1 (degree customisation), X2 (price)
and X3 (delivery time) in such a way that the sum of various realizations of U1, U2 and U3
best represents the judged evaluations for the 16 alternatives.

Yi ¼ U1 ðxi1 Þ þ U2 ðxi2 Þ þ U3 ðxi3 Þ; i ¼ 1; 2; . . . ; 16

where:
xi1 = level of the degree customisation attribute for the ith alternative,
xi2 = level of the price attribute for the ith alternative,
xi3 = level of the delivery time attribute for the ith alternative,
Yi = rating given to the ith alternative,
U1 (*) = part-worth function for level of customisation attribute,
U2 (*) = part-worth function for price attribute, and
U3 (*) = part-worth function for delivery time attribute.
One of the main features of conjoint analysis, is that the estimated functions can be
used to predict the utility score of a new profile which was not used in the data collected
process. The dependent variable in the conjoint model represent consumer’s trade-off
among the attributes of the alternative. A conjoint analysis was run for both groups,
males, and females. As we hypothesize, the trade-off and part-worth differ by gender. We
tested our hypothesis by running a conjoint analysis for both group and comparing them.
168 H. D. KALANTARI ET AL.

4. Results
Participants comprised 52% female and 48% male with an average age of 45.32
(SD = 17.51). More specifically, the average age of female participants was 43.24
(SD = 17.47) and average age of males were 47.57 (SD = 17.35). The median income
level of both females and males were 50,000 USD to 74,999. USD Many participants were
from Victoria, Australia (male = 40%, female = 43%). By using ordinary least-squares
regression, the full profile conjoint model of the participants’ decision making was
estimated. The method calculated the weighting placed on each level of each attribute,
which is also called a utility score or part-worth, for each participant. Hence, this method
was used to estimate the full profile conjoint model of the participants’ decision making.
Based on these part-worths, predictions were generated for how each participant rates
new profiles. In this method, the goodness-of-fit is also calculated and assessed based on
the strength of the Pearson’s correlation between the rating that participants gave to each
profile and the rating that the conjoint model predicted.
Before presenting the results, the two scores (importance scores and utility scores or
part-worths) that are produced by conjoint analysis are explained. Importance scores
indicate the influence of the attributes. In other words, importance scores reveal the
importance of the attributes to the participants, indicating which attributes most influ­
ence their purchase decision-making process. The importance score is calculated based
on the utility scores, with a larger range in utility scores for the levels of one attribute
ranking higher in importance. The higher importance scores imply a higher influence in
the participants’ decision-making process. The relative importance of attributes is pre­
sented as a percentage based on the importance scores. As the calculated importance
scores are ratio-scaled data and based on the current study measures, an attribute with an
importance score of 40% can be considered twice as important as an attribute with a 20%
importance. According to Orme (2001), in conjoint analysis, the attributes’ importance
can be compared with one another, notwithstanding the fact that the importance is
always relative to the other attributes in the research.
Utility scores (or part-worths), another measured score in conjoint analysis, provide
details about a participant’s desirability of the levels. For one attribute, the level with the
highest positive part-worth is the most preferred level, while the level with the lowest
negative part-worth is the least preferred level among the attribute’s levels. As the utility
score is based on an interval scale, the utility score value of an attributes level cannot be
compared with another attribute level utility score. The multiple linear regressions were
performed for each participant individually and the average of part worth utility were
then calculated for male and female participants separately. The results are presented in
Table 2.
The results indicate that female participants exhibited more concern about the level of
customisation and the delivery time. In contrast, male participants indicated more
concern about the price of the product. Figure 1 shows how important each attribute
was across the male and female participants.
Following Kulshreshtha et al. (2017); (2019), the validity and reliability of conjoint
analysis is assessed by calculating Kendall’s tau and Pearson’s r. The results reported in
Table 3 demonstrate a very high goodness-of-fit for the available data for both female and
male groups. A very high correlation was found between the holdout set’s rating and the
 JOURNAL OF GLOBAL FASHION MARKETING 169

Table 2. Part-worth utilities (PWU) and relative importance (RI).

Male Female
Attribute/Levels RI %/PWU RI %/PWU
Degree of Customisation 29.60% 37.97%
Fully-Customised 5.838 9.945
Customised −1.039 −1.769
Standard −4.799 −8.176
Price 44.10% 28.41%
$120 13.962 6.279
$140 3.257 1.480
$160 −5.193 −1.562
$180 −12.026 −6.197
Delivery time 26.30% 33.62%
4 Days 5.854 8.318
1.5 Weeks 1.030 1.284
3 Weeks −1.785 −2.538
4.5 Weeks −5.099 −7.064
RI: Relative Importance, PWU: Part-Worth Utility

50.000
45.000
40.000
Relative Importance

35.000
30.000
25.000
Male
20.000
15.000 Female
10.000
5.000
0.000
Level of Customisation Price Delivery Time

Attributes
Figure 1. Relative importance for male and female participants.

Table 3. Correlations between observed and estimated preferences.

Correlations a
Males Females
Value Sig. Value Sig.
Pearson’s r 0.994 0.000 0.990 0.000
Kendall’s tau 0.950 0.000 0.979 0.000
Kendall’s tau for Holdouts 1.000 0.059 1.000 0.059
a. Correlations between observed and estimated preferences

predicted model. The correlations were significant at a level of 10%, which implies strong
confidence in the main effect model’s suitability (Table 3).
Conjoint analysis for each of the participants calculated the relative importance of each
attribute at an individual level. An independent-samples t-test was conducted to compare
the relative importance of each attribute for male and female participants (Table 4). There
was a significant difference in the relative importance of all three attributes (p < 0.001).
170 H. D. KALANTARI ET AL.

Table 4. Group Statistics.

Group Statistics
Attributes Gender N Mean (RI) Std. Deviation t (p-values)
Degree of Customisation Male 170 0.296 0.1937 −3.97 (0.000)
Female 183 0.379 0.2003
Price Male 170 0.441 0.1996 8.822 (0.000)
Female 183 0.284 0.1291
Delivery time Male 170 0.263 0.1610 −4.013 (0.000)
Female 183 0.336 0.1803

These results suggest that degree of customisation was the most important attribute
for females (Relative importance = 0.379) and price was the most important attribute for
males (Relative importance = 0.441). The outcome also revealed that female participants
were more likely to purchase a fully customised product rather than customised and
standard products, compared to male participants.

5. Discussion
The study showed evidence that gender influences willingness to pay for different levels
of mass customised products. Previous studies generally agree that consumers are willing
to pay extra for mass customised products (Kalantari & Johnson, 2018; Gilal et al., 2018).
Extending this knowledge, the present study found that female customers are more
concerned about the level of customisation and the delivery time when making mass
customised running shoe purchases. Male customers are, however, more concerned
about the price of the products. These findings are consistent with previous literature
on gender and decision making (e.g. Moss, 2017).
In addition, the study showed that female customers are more likely to purchase a fully
customised product rather than customised and standard products, compared to male
customers. Moreover, marketing strategies that aim at promoting customised products
(e.g. NikeiD) may not be preferred by female consumers, rather they may prefer an
option to purchase a fully customised product at a higher price. While the finding
addresses the research gap of this field of study where gender differences have been
largely overlooked (e.g. Seo & Lang, 2019; Walcher et al., 2016), it also informs the
marketing and promotion strategies of mass customised products.
Addressing the some of the inconsistent findings available in the literature, regarding
risk-averse preference of female consumers (e.g. Bacha & Azouzi, 2019; Doan &
Iskandar-Datta, 2020; Lin et al., 2019), this study found that female customers perceive
greater risks than male customers do, and they are willing to try a new kind of product at
the risk of premium price. Female customers’ decision-making process is based on
uncertainty, doubts and validity; hence, they are more likely to purchase a fully custo­
mised product. On the other hand, men are less ready for mass customised products and
lower prices are more applicable to them. Therefore, as is found in previous literature
(Wu et al., 2012), although product uniqueness has a stronger effect than assumed price
premiums of limited editions of products on product purchase intention, in the mass
customised product purchase context, this may not necessarily be the case for male
consumers who prefer fully customised products as a low-price. As reported in more
recent research (e.g. Lee, 2020), male consumers can be more shopping goal-oriented
 JOURNAL OF GLOBAL FASHION MARKETING 171

than their female counterparts. Overall, the current study extends the understanding of
gender-specific consumer responses towards mass customised products.

6. Implications and conclusion

This study draws several theoretical and practical implications that can be considered by
the future researchers as well as firms operate in mass customised product markets.
Extending the Bardakci and Whitelock (2004) framework of customer readiness of mass
customisation, this study assessed customer readiness to purchase a customised product
by considering the degree of customisation, price, delivery time and the design time
simultaneously. It provides valuable theoretical insights on the interaction between price
and delivery time as well as customer willingness to spend time on designing customised
products. In addition, the study extends the customer readiness for purchasing customi­
sation by examining the role of gender in the consumer decision-making process across
standard products, customised products, and fully customised products as three degrees
of customisation with different levels of perceived control.
The findings of the study suggest that firms who wish to invest in fragmented market
through mass customisation should first focus more on female customers as they are
shown to be more open to mass customised products and willing to pay extra for mass
customised products. Therefore, premium pricing strategy can be recommended to reach
those female target markets. Regarding male customers, firms should keep mass custo­
mised product prices relatively lower using standard delivery methods. Those firms are
cautioned that this strategy could increase to product delivery time which could dis­
satisfied the consumers eventually.
It should be noted that the study findings are limited to the running shoe product-
context. Therefore, they cannot be generalised to investigate the gender effect of all (or
many) other mass customisation contexts as these effects may depend on the identity
relevance of a product or service domain to a specific gender. This requires future research
to examine gender effects on other product contexts. Further future research can also
investigate the optimum price and delivery time for mass customised products.

Disclosure statement
No potential conflict of interest was reported by the authors.

ORCID
Hassan Daronkola Kalantari http://orcid.org/0000-0003-3972-0788
Lester W Johnson http://orcid.org/0000-0003-3577-2804
Chamila R. Perera http://orcid.org/0000-0002-4206-0955

References
Adel, H. M., & Younis, R. A. A. (2019). Using co-creating mass-customisation and innovation
climate for enhanced value. Journal of Humanities and Applied Social Sciences, 1(1), 25–42.
https://doi.org/10.1108/JHASS-05-2019-002
172 H. D. KALANTARI ET AL.

Aichner, T., & Coletti, P. (2013). Customers’ online shopping preferences in mass customization.
Journal of Direct Data and Digital Marketing Practice, 15(1), 20–35. https://doi.org/10.1057/
dddmp.2013.34
Alonso-Almeida, M., & Bremser, K. (2015). Does gender specific decision making exist? EuroMed
Journal of Business, 10(1), 47–65. https://doi.org/10.1108/EMJB-02-2014-0008
Arsel, Z., Eräranta, K., & Moisander, J. (2015). Theorising gender and gendering theory in
marketing and consumer research. Journal of Marketing Management, 31(15–16), 1553–1558.
https://doi.org/10.1080/0267257X.2015.1078396
Averill, J. R. (1973). Personal control over aversive stimuli and its relationship to stress.
Psychological Bulletin, 80(4), 286–303. https://doi.org/10.1037/h0034845
Bacha, S., & Azouzi, M. A. (2019). How gender and emotions bias the credit decision-making in
banking firms. Journal of Behavioral and Experimental Finance, 22, 183–191. https://doi.org/10.
1016/j.jbef.2019.03.004
Bae, S., & Miller, J. (2009). Consumer decision-making styles for sport apparel: Gender compar­
isons between college consumers. ICHPER-SD. Journal of Research, 4(1), 40–45. https://eric.ed.
gov/?id=EJ903481
Bakewell, C., & Mitchell, V. W. (2004). Male consumer decision-making styles. The International
Review of Retail, Distribution and Consumer Research, 14(2), 223–240. https://doi.org/10.1080/
0959396042000178205
Bakewell, C., & Mitchell, V.-W. (2006). Male versus female consumer decision making styles.
Journal of Business Research, 59(12), 1297–1300. https://doi.org/10.1016/j.jbusres.2006.09.008
Bardakci, A., & Whitelock, J. (2004). How “ready” are customers for mass customisation? An
exploratory investigation. European Journal of Marketing, 38(11/12), 1396–1416. https://doi.
org/10.1108/03090560410560164
Belch, G. E., & Belch, M. A. (2016). Advertising & promotion. The McGraw-Hill/Irwin Series in
Marketing.
Botti, S., McGill, A. L., & Iyengar, S. S. (2003). Preference for control and its effect on the
evaluation of consumption experiences. Advances in Consumer Research, 30(1), 127–128.
https://www.acrwebsite.org/volumes/8749/volumes/v30/NA-30
Chen-Yu, J., & Yang, J.-H. (2018). Predictors of purchase intention, price willing to pay, and
willingness to pay a premium for men’s mass-customized dress shirts. Paper presented at the
International Textile and Apparel Association (ITAA), Ames, Iowa, United States.
Davis, S. M. (1987). Future perfect. Addison-Wesley.
Dellaert, B. G., & Dabholkar, P. A. (2009). Increasing the attractiveness of mass customization: The
role of complementary on-line services and range of options. International Journal of Electronic
Commerce, 13(3), 43–70. https://doi.org/10.2753/JEC1086-4415130302
Doan, T., & Iskandar-Datta, M. (2020). Are female top executives more risk-averse or more
ethical? Evidence from corporate cash holdings policy. Journal of Empirical Finance, 55,
161–176. https://doi.org/10.1016/j.jempfin.2019.11.005
Franke, N., Keinz, P., & Steger, C. J. (2010). Customization: A goldmine or a wasteland? GfK
Marketing Intelligence Review, 2(2), 26–33. https://doi.org/10.2478/gfkmir-2014-0060
Gilal, F. G., Zhang, J., Gilal, N. G., & Gilal, R. G. (2018). Integrating self-determined needs into the
relationship among product design, willingness-to-pay a premium, and word-of-mouth: A
cross-cultural gender-specific study. Psychology Research and Behavior Management, 11,
227–241. https://doi.org/10.2147/PRBM.S161269
Gilmore, J. H., & Pine, B. J. (1997). The four faces of mass customization. Harvard Business Review,
75(1), 91–101.
Globenewswire, 2019. https://www.globenewswire.com/news-release/2019/04/22/1807194/0/en/
5-9-Billion-3D-Printed-Footwear-Market-2019-2029-10-Year-Forecasts-and-Analysis-of-the-
Market-Potential.html
Green, P. E., & Rao, V. R. (1971). Conjoint measurement for quantifying judgmental data. Journal
of Marketing Research, 8(3), 355–363. https://doi.org/10.2307/3149575
Jablonsky, C. (2018). Mass customization: A key to improving customer outcomes. https://www.
aspentech.com/en/blog/blog/Mass_Customization_Key_to_Improving_Customer_Outcomes
 JOURNAL OF GLOBAL FASHION MARKETING 173

Jafari, H., Nyberg, A., Osnes, T.-L., & Schmitz, A. (2015). Customization in bicycle retailing. Journal of
Retailing and Consumer Services, 23, 77–90. https://doi.org/10.1016/j.jretconser.2014.12.004
Jessie H.Chen-Yu, & Jung-Ha Yang. (2020). Consumer characteristics as predictors of purchase
intentions and willingness to pay a premium for men’s mass-customized apparel. Journal of
Global Fashion Marketing, 11(2), 154–170.
Jost, P.-J., & Süsser, T. (2019). Company-customer interaction in mass customization.
International Journal of Production Economics, 220, 107454.
Kalantari, H. D., & Johnson, L. (2018). Australian customer willingness to pay and wait for
mass-customised products. Asia Pacific Journal of Marketing and Logistics, 30(1), 106–120.
https://doi.org/10.1108/APJML-01-2017-0006
Kalantari, H. D., & Johnson, L. W. (2019). Willingness to pay and wait for mass customised
products. International Journal of Mass Customisation, 5(2), 111–129. https://doi.org/10.1504/
IJMASSC.2019.103632
Kasiri, L. A., Guan Cheng, K. T., Sambasivan, M., & Sidin, S. M. (2017). Integration of standardi­
zation and customization: Impact on service quality, customer satisfaction, and loyalty. Journal
of Retailing and Consumer Services, 35, 91–97. https://doi.org/10.1016/j.jretconser.2016.11.007
Kulshreshtha, K., Bajpai, N., Tripathi, V., & Sharma, G. (2019). Cause-related marketing: An
exploration of new avenues through conjoint analysis. Benchmarking: An International Journal,
26(6), 2017–2050. https://doi.org/10.1108/BIJ-05-2018-0131
Kulshreshtha, K., Tripathi, V., & Bajpai, N. (2018). 1971–2017: Evolution, exploration and test of
time of conjoint analysis. Quality & Quantity, 52(6), 2893–2919. https://doi.org/10.1007/
s11135-018-0717-6
Kulshreshtha, K., Tripathi, V., Bajpai, N., & Dubey, P. (2017). Discriminating market segments
using preferential green shift: A conjoint approach. Foresight, 19(4), 386–408. https://doi.org/10.
1108/FS-02-2017-0007
Kuzmanovic, M., Martic, M., Vujosevic, M., & Panic, B. (2011). Construction of efficient conjoint
experimental designs using MCON procedure. International Journal of Physical Sciences, 6(24),
5659–5670. https://doi.org/10.5897/IJPS11.450
Lang, C., Xia, S., & Liu, C. (2020). Style and fit customization: A web content mining approach to
evaluate online mass customization experiences. Journal of Fashion Marketing and
Management: An International Journal. https://doi.org/10.1108/JFMM-12-2019-0288
Lee, E. J. (2020). Gender influence in the effect of design aesthetics on perceived product value of
wearables. International Journal of Advanced Culture Technology, 8(2), 129–138. https://doi.org/
10.17703/IJACT.2020.8.2.129
Lee, -H.-H., Damhorst, M. L., Campbell, J. R., Loker, S., & Parsons, J. L. (2011). Consumer
satisfaction with a mass customized Internet apparel shopping site. International Journal of
Consumer Studies, 35(3), 316–329. https://doi.org/10.1111/j.1470-6431.2010.00932.x
Lee, Y., Damhorst, M. L., Lee, M.-S., Kozar, J. M., & Martin, P. (2012). Older women’s clothing fit
and style concerns and their attitudes toward the use of 3D body scanning. Clothing and Textiles
Research Journal, 30(2), 102–118. https://doi.org/10.1177/0887302X11429741
Levitt, M., & Levitt, T. (1998). Thinking about management. The Free Press. (pp. 103–112).
Li, T., & Unger, T. (2012). Willing to pay for quality personalization? Trade-off between quality
and privacy. European Journal of Information Systems, 21(6), 621–642. https://doi.org/10.1057/
ejis.2012.13
Liang, Y., & Liu, C. (2019). Comparison of consumers’ acceptance of online apparel mass
customization across web and mobile channels. Journal of Global Fashion Marketing, 10(3),
228–245. https://doi.org/10.1080/20932685.2019.1619469
Lin, X., Featherman, M., Brooks, S. L., & Hajli, N. (2019). Exploring gender differences in online
consumer purchase decision making: An online product presentation perspective. Information
Systems Frontiers, 21(5), 1187–1201. https://doi.org/10.1007/s10796-018-9831-1
Ling, I.-L., Liu, Y.-F., Lin, C.-W., & Shieh, C.-H. (2020). Exploring IKEA effect in self-expressive
mass customization: Underlying mechanism and boundary conditions. Journal of Consumer
Marketing, 37(4), 365–374. https://doi.org/10.1108/JCM-09-2017-2373
174 H. D. KALANTARI ET AL.

Maclaran, P., & Kravets, O. (2018). Feminist perspectives in marketing: Past, present and future.
forthcoming in the Routledge Companion to Critical Marketing. In M. Tadjewski, M. Higgins,
J. Denegri-Knott, R. Varman, & N. Dholakia (Eds.), (pp. 64–82). Oxon, Routledge.
Marathe, S., & Sundar, S. S. (2011).What drives customization? Control or identity? In Proceedings
of the SIGCHI conference on human factors in computing systems, 781–790. ACM Press.
Moss, G. (2017). Gender, design and marketing: How gender drives our perception of design and
marketing. Routledge.
Orme, B. K. (2001). Assessing the monetary value of attribute levels with conjoint analysis: Warnings
and suggestions. Sawtooth Solutions Customer Newsletter (Spring) Sawtooth Software Inc
Pallant, J., Sands, S., & Karpen, I. (2020). Product customization: A profile of consumer demand. Journal
of Retailing and Consumer Services, 54, 102030. https://doi.org/10.1016/j.jretconser.2019.102030
Pine, B. J. (1993). Mass customization: The new frontier in business competition. Harvard Business Press.
Protolabs. (2019). Smart manufacturing enables mass customization trend. https://www.protolabs.
com/resources/blog/smart-manufacturing-enables-mass-customization-trend/
Purnomo, M. R. A., & Shinta, R. C. (2019). Simulation model development for determination of
components production quantity and lead time reduction in mass customisation of single
production stage. IOP conference series: Materials science and engineering, 528, 012065.
Makasar, South Sulawesi, Indonesia. https://doi.org/10.1088/1757-899x/528/1/012065
PWC. (2019). Digital pulse. https://www.digitalpulse.pwc.com.au/trends-brand-personalisation-cx/
Rao, V. R. (2014). Applied conjoint analysis. Springer.
Ribeiro, L. S., Duarte, P. A. O., & Miguel, R. (2017). Online consumer behaviour of mass-customised
apparel products: A hierarchy of traits approach. Journal of Fashion Marketing and Management, 21
(2), 158–171. https://doi.org/10.1108/JFMM-07-2016-0068
Runfola, A., Temperini, V., & Gregori, G. L. (2018). Problematizing customization and IT in the
fashion industry: A case study of an Italian shoemaker AU - Perna, Andrea. Journal of Global
Fashion Marketing, 9(1), 73–86. https://doi.org/10.1080/20932685.2017.1399081
Sanz de Acedo Lizarraga, M. L., Sanz de Acedo Baquedano, M. T., Soria Oliver, M., & Closas, A.
(2009). Development and validation of a decision-making questionnaire. British Journal of
Guidance & Counselling, 37(3), 357–373.
Schutz, W. (1966). The interpersonal world. Science and Behavior Books.
Seiji, E., & Doris, K. (2008). Mass customization for long-term relationship development: Why
consumers purchase mass customized products again. Qualitative Market Research: An
International Journal, 11(3), 275–294. https://doi.org/10.1108/13522750810879011
Seo, S., & Lang, C. (2019). Psychogenic antecedents and apparel customization: Moderating effects
of gender. Fashion and Textiles, 6(1), 1–19. https://doi.org/10.1186/s40691-019-0175-3
Singh, V., Schiebener, J., Müller, S. M., Liebherr, M., Brand, M., & Buelow, M. T. (2020). Country
and sex differences in decision making under uncertainty and risk. Frontiers in Psychology, 11,
486. https://doi.org/10.3389/fpsyg.2020.00486
Tokimatsu, K., Aicha, M., Yoshida, K., Nishio, M., Endo, E., Sakagami, M., . . . Itsubo, N. (2016).
Measuring marginal willingness to pay using conjoint analysis and developing benefit transfer
functions in various Asian cities. International Journal of Sustainable Development & World
Ecology, 23(6), 541–552. https://doi.org/10.1080/13504509.2016.1168326
Walcher, D., Leube, M., & Blazek, P. (2016). Gender differences in online mass customization: An
empirical consumer study which considers gift-giving. International Journal of Industrial
Engineering and Management, 7(4), 153–158.
White, R. W. (1959). Motivation reconsidered: The concept of competence. Psychological Review,
66(5), 297–333. https://doi.org/10.1037/h0040934
Winters, E., & Ha, S. (2012). Consumer evaluation of customer loyalty programs: The role of
customization in customer loyalty program involvement. Journal of Global Scholars of
Marketing Science, 22(4), 370–385. https://doi.org/10.1080/21639159.2012.720752
Wittink, D. R., Krishnamurthi, L., & Reibstein, D. J. (1990). The effect of differences in the number
of attribute levels on conjoint results. Marketing Letters, 1(2), 113–123. https://doi.org/10.1007/
BF00435295
 JOURNAL OF GLOBAL FASHION MARKETING 175

Wolf, A., & Zhang, L. (2016). The effect of customization and gender on customers’ attitude.
International Journal of Hospitality Management, 56, 28–32. https://doi.org/10.1016/j.ijhm.
2016.04.001
Wu, W.-Y., Lu, H.-Y., Wu, -Y.-Y., & Fu, C.-S. (2012). The effects of product scarcity and
consumers’ need for uniqueness on purchase intention. International Journal of Consumer
Studies, 36(3), 263–274. https://doi.org/10.1111/j.1470-6431.2011.01000.x
YouGov. (2018). YouGov Survey Results. https://campaign.yougov.com/rs/060-QFD-941/images/
Made%20to%20Order%20%E2%80%93%20Personalization%20report.pdf
Zhao, H., McLoughlin, L., Adzhiev, V., & Pasko, A. (2019). Why do we not buy mass customised
products? An investigation of consumer purchase intention of mass customised products.
International Journal of Industrial Engineering and Management, 10(2), 181–190. https://doi.
org/10.24867/IJIEM-2019-2-238
 Innovation
Organization & Management

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/rimp20

Innovation logics in the digital era: a systemic

review of the emerging digital innovation regime

Kalle Lyytinen

To cite this article: Kalle Lyytinen (2021): Innovation logics in the digital era: a systemic review of
the emerging digital innovation regime, Innovation, DOI: 10.1080/14479338.2021.1938579

To link to this article: https://doi.org/10.1080/14479338.2021.1938579

Published online: 01 Jul 2021.

Submit your article to this journal

Article views: 63

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rimp20
INNOVATION: ORGANIZATION & MANAGEMENT
https://doi.org/10.1080/14479338.2021.1938579

ARTICLE

Innovation logics in the digital era: a systemic review of the

emerging digital innovation regime
Kalle Lyytinen
Department of Design & Innovation, Case Western Reserve University, Cleveland, OH, USA

ABSTRACT ARTICLE HISTORY

This essay reviews the ontological status of digital material in industrial Received 16 June 2020
operations and conditions it provides for innovation. It recognises the Accepted 28 May 2021
semiotic quality of digital material and the critical role of von Neumann KEYWORDS
architecture as to show that digital innovation advances through a three Digital innovation; industrial
pronged, largely orthogonal process of embedding- a process of inter­ organisation; scale; scope;
lacing elements of one innovation domain to that of another. Three digitalisation; von Neumann
types of embedding define digital innovation: operational embedding architecture
(code-computer), virtual embedding (real-world phenomena-code) and
contextual embedding (performing code-social setting). Each constitu­
tes a ‘leverage point’ for further expanding digital innovation and they
interact dynamically while industrial innovation is defined by a static vir­
tual and contextual embedding. The processes that underlie innovation
in the two regimes follow differential logics. In industrial regime (1)
discover-synthesise and (2) manufacture-distribute activities combine to
form the leverage points for innovation. In digital regime 1) discover
product or behaviour (ideation); 2) abstract to digital material (virtua­
lise); 3) Implement and replicate digital material (variations in opera­
tional/contextual embedding); and 4) Deliver product and/or behaviour
(perform digital material) activities form the leverage points for innova­
tion enabling wider, open, and faster innovation with a distinct value
logic.

Introduction
When writing this essay in the late spring and summer of 2020, multiple icons of the
20th-century retail industry – J.C. Penney, J Crew, and Neiman Marcus, the early
stalwarts of its business model – filed for bankruptcy, followed by 200-year-old Brooks
Brothers. The pandemic of 2020 was the last straw for these companies and made them
join the earlier failings of Sears, Kmart, and many others. At the same time, the ‘born
digital’ Amazon.com has become one of the most valuable companies globally and an
invincible competitor. Importantly, the failures of these companies cannot be attributed
to the lack of trying and innovating as they have been doing it for decades. Nor can the
failings be attributed to poor management because it is highly unlikely that all these
companies were poorly led and could not hire top managerial talent. Rather, it appears to
be a story of a systemic failure that led the companies to innovate in wrong places, at

CONTACT Kalle Lyytinen [email protected]

© 2021 Informa UK Limited, trading as Taylor & Francis Group
2 K. LYYTINEN

wrong time, and in a wrong way. Essentially, the companies could not ultimately innovate
how they should innovate as dictated by the emerging digital innovation regime.
The challenge for innovation scholars has been how to frame, understand, and explain
the new context of innovation brought about by pervasive digitalisation. The context now
permeates all industries and their business models. It is novel in that it cannot be
explained by older models that allowed many companies of industrial era to succeed
for a century or so. The new context raises a host of so far unanswered questions, such as:
Why does the established, industrial model of innovation invented during the era of mass
manufacturing and distribution – including the concept of retail department stores – and
perfected under the auspices of industrial organisation not work anymore?1 How and
why does the logic of innovation differ, and is it hard to learn and assimilate for
incumbents despite their significant advantages in competencies, skills, resources, and
market visibility? What brings the change so fast and results in such a widespread
devastation? Difficulty in answering these questions raises doubts about the plausibility
of received ideas and explanations of industrial innovation brought forward during the
past century. For example, these explanations cover Chandlerian concepts of how
strategy begets structure and are driven by scale and scope, Porterian ideas of positioning
and value extraction or resource-based explanations of the merits of unique and stable
resource constellations. These assume in one form or other on static notions of industry
and its change and how these are driven by scale and scope effects. Most of the well-tried
ideas to account for innovation in such settings such as product versus administrative
innovation and radical innovation versus incremental innovation, have clearly lost much
of their magic or work only in limited ways.
This raises the question of what explains innovation and makes it successful in an era
of pervasive digitalisation? Are there truly novel systemic ways of engaging and mana­
ging in innovation, which we can call digital innovation regime and are distinct in kind
from that of the previous industrial innovation regime? Much research produced over the
past decade to explain ‘digital’ has offered fresh and reasoned insights into why and how
the digital calls for a new vernacular of innovation. In particular, major advances have
been made to characterise the ‘new’ in digital innovation including:

(1) What the novel features of the digital innovation are such as its layered and loosely
coupled nature (Lee & Berente, 2012; Tilson et al., 2010), its inverse use and value
logics (Baskerville et al., 2020), or how the digital transforms the scope and
features of physical products and their innovation (Baskerville et al., 2020;
Sandberg et al., 2020);
(2) What the ontology of the digital material is and why its properties propel new
forms of innovation (Baskerville et al., 2020; Faulkner & Runde, 2013; Kallinikos
et al., 2013; Tilson et al., 2010; Yoo et al., 2010). These have been commonly
attributed to combinability, reflexivity, programmability, nearly zero variable cost
features of digital material, and so on;
(3) What the features of the digital innovation environment are (Nambisan et al.,
2017) and how they affect innovation processes (Lyytinen et al., 2017; Nambisan
et al., 2017; Yoo et al., 2012);
(4) Why features 1–3 result in powerful tendencies towards network effects, platfor­
mization, and generativity as key organising principles for business (Alaimo et al.,
 INNOVATION: ORGANIZATION & MANAGEMENT 3

2020; Constantinides et al., 2018; Karhu et al., 2018; Parker et al., 2016; De Reuver
et al., 2017).

The primary focus of these works is to identify and compare differences between the old
and the new regimes from a specific angle. These works have shed new and brighter light
on what is different in digital innovation, such as its outcomes (fixed versus malleable),
material (digital versus analogue/physical), processes (close/disciplined versus open/
chaotic), organising principles (hierarchy versus platform), or economic logic (decreas­
ing versus increasing marginal returns). At the same time, these analyses do not connect
such changes to a deeper systemic change, which underlies and builds up the emerging
digital innovation regime. To carry out such systemic analysis calls for a historic,
comparative, relational analysis of how the previous industrial regime in essential,
systemic ways differs from the emerging digital innovation regime. This a similar call
that historians of industrial organisation, such as Chandler (1990) or Hughes (1989),
faced when they asked this regime differs from that of craft-based production. To put it
differently, we need to probe in a more abstract and systemic way the key elements that
inspire innovation during industrial or digital innovation regimes based on how the
elements became organised in separate, unique families of relationships and related
practices consequential for enabling or constraining innovation. Thereby features, activ­
ities, and outcomes as identified in the recent studies become plausible, while the
dominant ones fading away can be shown to become obsolete, infeasible, and thereby
excluded by natural selection. Such form of analysis, which we so far lack, would heed to
systemic assemblages of elements that underlie innovation activity and make it feasible to
organise innovation in specific ways. This invites us to examine how the assemblages
become phased into a specific order of relationships where specific elements are con­
nected or disconnected, enabling alternative forms of action and thinking and related
novelty (DeLanda, 2006). Based on such analysis, we can better understand why the
regimes differ substantially, and there is a systemic break between industrial and digital
(innovation) regimes. Such separation of ‘how things work’ also explains the grand
challenges of why established companies have found it so hard to move from one regime
to another.
To carry out a systemic analysis of this sort, I focus on what defines and distinguishes
the leverage points and related phases of innovation in industrial and digital (innovation)
regimes. I identify spaces where critical elements that enable or constrain innovation
interact as leverage points and define them as critical resource configurations (material
and immaterial) that enable and constrain where and how innovations can proceed and
under which goals. This leads me to locate and identify where and how innovations are
likely to take place in the regimes and the logics of how such innovations get formed and
justified and under what conditions they can proceed and succeed. I offer examples of such
leverage points in both regimes and illustrate that because of these differences, digital
innovation unfolds as a cumulative force that cuts across industries, settings, and envir­
onments as an unstoppable gale of destruction. Generally, the analysis illustrates how
industrial versus digital product properties, processes, and technologies differently shape
the spaces in which innovations can unfold. I hope to narrate why the digital regime
essentially differs from industrial regime in its assemblage of phases and leverage points.
4 K. LYYTINEN

The reminder of the article is organised as follows. I review critical elements and their
connections (leverage points) in two innovation regimes – industrial regime and digital
regime – as assemblages and discuss how the components of each regime are dynamically
constituted by their relations to other parts (DeLanda, 2006). I show that each regime is
consequently defined by distinct goals and constraints and different logics that call for
distinct capabilities to innovate (cognitive, organisational). I show how the industrial
regime has evolved through separate periods and how digitalisation has expanded in the
industrial environment to sow the seeds for the emerging digital innovation regime.
I show how each phase inherits features and constraints from the previous phase while
opening new challenges and capabilities for innovation.

Industrial innovation regime

Classic industrial innovation regime
The industrial innovation regime was established in the context of the emergence of
industrial organisation over a roughly 100-year period between 1850 and 1970. It had its
heyday between 1950 and 1970. The regime is founded on ideas of stable (new) products
assigned to separate markets and market segments. Product innovation mirrors and takes
place in the context of visible markets or anticipated new markets. The regime’s innova­
tion and organising logic are well narrated in Chandler’s detailed discussion of the
evolution of corporate strategies (Chandler, 1977, 1990). Firms had to learn how to
address scale and scope challenges of producing to mass markets and manage, control,
and coordinate related activities to manage demand and supply variations effectively
(Hughes, 1989).
The credo for innovation – introducing novelty in the regime – is how to create new
product categories. This happens primarily through scientific discovery or using scien­
tific principles to address needs in product markets and industries (e.g. car, telephone,
telefax, and record player). These products serve the well-defined needs of identified
customers. Derivative and complementary innovations will combine and follow the sales
and use of such products and address challenges in scaling the production, quality
control, and distribution- or managing-related costs, risks, and benefits. Innovation
takes place sequentially (internally) through a pipe of processes that create/design new
products, scale them up (continuously) for mass markets, and allocate investments to
most profitable and potential market categories to serve latent/identified needs. Classic
examples of companies that did well in this innovation regime are DuPont, GE, and
Daimler.
The innovation regime operates through two separate and tightly connected phases.
The first one focuses in creating scope economies for learning expressed in (1) discovery-
synthesise activities necessary for product innovation. The phase uses discovered proper­
ties of physical matter and competencies in shaping them to serve identified and latent
needs (combustion engine for transportation, radio signals for transmitting sounds, etc.).
The next phase focuses on scale economies of learning expressed in (2) manufacture-
distributed activities, which create and move the products to customer markets and offer
complementary services for how to use them (such as service and gas stations). This
phase relies on pertinent material, engineering, and social system properties and related
 INNOVATION: ORGANIZATION & MANAGEMENT 5

infrastructures (such as communication and transportation systems, standards, and

regulations) that cater to identifying customer needs and balancing supply and demand
in the short and the long run. Overall, the regime assumes, needs, and sustains activity
systems associated with both phases and seeks to establish leverage how they can jointly
promote or sustain innovation. Activities in phase 1 use outputs of activities in phase 2 to
acquire resources to continue innovating and gain information about the markets for
new products. Activities in phase 2 need input from activities in phase 1 to continue to
grow and penetrate new markets and continue to satisfy identified, growing, or segre­
gated customer needs in established markets.2
A critical assumption undergirding innovation in an industrial regime is the concept
of a distinct, discrete product circulating in the market system. Erecting and building
know-how to bundle and operate activity systems for discovery-synthesis and manufac­
ture-distributed phases create significant barriers to entry in terms of knowledge, capital,
and access to markets (market share). This erects boundaries between activity systems
related to different products and promotes the idea of distinct industries with well-
defined learning scope. In each such industrial innovation domain, the innovation by
character can be more or less radical. The more radical it is, the more latent the customer
need, and the more exotic the new market. Consequently, products are innovated
through smaller or bigger change deltas so as to balance the product innovation risks.
One reason for this is that the necessary scale economies kick in only if the innovator can
make the innovation to diffuse to a large enough market to recoup the initial investment
and necessary scaling costs. For many years, there has been a significant interest in
Rogersian type diffusion of innovation models, because they help explain conditions for
successful innovation diffusion (Rogers, 2005). Generally, the innovations around man­
ufacturing and distribution are as critical (such as the idea of a large retail store) as those
related to products as such. These innovations address administrative/process changes
and activity configurations in the manufacture-distributed phase necessary to balance
and coordinate demand-supply challenges in industrial regime (manufacturing, distribu­
tion and operations) (see Hughes, 1989, for detailed history).
Essentially, the regime’s leverage points for innovation are continued expansion of
new (mass market) products through discovery of properties of the matter that allow
them to fulfil new latent customer needs. Initial discovery is followed by constant product
improvement and increased small variation, calling for competencies in industrial design
and market segmentation, and continued innovations in manufacturing, distribution,
and customer services to administer costs. Not surprisingly, digital technologies were
initially mobilised as critical elements for administrative innovation starting in the 1950s
(Yates, 1989). They create unprecedented efficiencies and speed to scale operations and
later offered support for scope economies (such as better integrated product development
and related knowledge management) (Yoo et al., 2010). The industrial regime model
dominated innovation until the early 1980s when the notion of product platforms
emerged in multiple industries (cars, computers, consumer electronics, etc.). This change
was associated with the global reach of the industries, which posed new challenges in
product development, manufacturing, and distribution because of increased variation in
local product needs and the need to use manufacturing and distribution resources
efficiently across the globe. This triggered the second phase of industrial innovation
regime, called modular industrial innovation regime.
6 K. LYYTINEN

Modular industrial innovation regime

This innovation regime extends the industrial innovation regime. Instead of focusing
on product categories as individual innovations, it introduces variability, flexibility, and
speed in responding to emerging customer needs and technological change. It permits
more efficient use of global manufacturing and distribution chains and related capa­
cities. The regime expands the notion of a product innovation towards higher
abstracted class of product platforms and the concept of product variability (Gawer,
2009, 2014). Platform-based innovation needs to be enabled by new administrative
innovations centred on abstracted design rules and capabilities that would leverage
those rules. The design rules specify specific standardised interface principles and rules
that guide decomposing the products into functional components where the relation­
ships between the components are defined by design rules dictating generalised func­
tional relationships between module interfaces (Schilling, 2000). Use of such rules
affords looser coupling between product components and products: one to many
instead of one to one (Baldwin & Clark, 2000). Their use allows significant variation
in products as product components and subsystems treated modules can be varied in
several places of the product hierarchy. Moreover, these hierarchies can be extended
without a need to synthesise a totally new product for each design. Modularity enables
more efficient use of the manufacturing and distribution capabilities in scale and scope
because products and components are now loosely connected during the manufacture-
distributed phase.
Consequently, an extended logic underlies product innovation in modular indus­
trial innovation regime. It now consists of three phases: (1) discover-synthesise; (2)
modularise-combine; and (3) manufacture-distribute. The innovation centred on
modularise-combine phase distributes the product innovation and manufacturing
innovation across the modules of the product architecture as soon as the synthesise-
combine innovation around the platform architecture had been carried out. The
abstracted platform solution enabled by the synthesise-combine phase creates more
flexible and responsive supply chain ecosystems with lower cost and higher product
variability. It also allows inventing new combinations of modules as separate pro­
ducts of new categories (e.g. mobile phone versus game console) creating higher
flexibility to respond to new markets when new product categories emerge.
Leverage points in this state are largely the same as in the previous regime. Discover-
synthesise still dominates, while modularise-combine introduces variation and flexibility
to create faster new products with lower cost. Platform-based product variations open
wider innovation spaces in the supply chain and afford economies of scale and scope and
how to balance trade-offs therein.
One benefit in the modular industrial regime is that innovation can be distributed
down and externalised to modules when the platform architecture has been fixed (Lee &
Berente, 2012). The regime emphasises the critical role of product orchestrators as
enablers and governors of product ecosystems. Truly new product categories emerge
rarely, while variations in products and modules are more frequent because they offer
stepwise means to advance product features and variability across technology generations
when better technologies become available for key components. Examples that deploy
such opportunities are generations of mobile phones, computers, or cars. Iconic
 INNOVATION: ORGANIZATION & MANAGEMENT 7

companies that leveraged effectively modular strategy are Volkswagen/Audi, Toyota,

Nokia, and Airbus. This later form of industrial innovation regime would not have
been possible without innovations in inter-organisational digital capabilities that allowed
sharing standardised product descriptions using CAD/CAM systems (Argyres, 1999; Yoo
et al., 2010) and exchange of standardised trading messages (e.g. Electronic Data
Interchange (EDI) and product identification standards). These allowed more efficient,
faster, and more scalable coordination within the design and supply chains to manage
design, manufacturing, and distribution activities (Malone et al., 1987; Yoo et al., 2010).

Digital innovation regime

Nature of digital material
Until the early 2000s, digital technologies supported mainly design, administration, and
coordination activities across two or three phases of the industrial innovation regime.
Digital technologies carried out narrow functions in specific product components, such
as controlling oxygen and gas levels in a carburettor (Lee & Berente, 2012).3 During the
1990s and early 2000s, because of exponential growth in internet use and capabilities
(Hanseth & Lyytinen, 2010; Tilson et al., 2010) digital technologies turned into infra­
structural functions and capabilities and became everyday objects and experience. This
growth expanded digital capabilities and their use beyond organisational boundaries to
industries, regions, and societies. The technologies could now combine in accessible
forms extensive storage and computation functions with connectivity to varied digitised
data, enabling large-scale use (Lyytinen & Yoo, 2002a). This has resulted in the two
decades that followed a structural shift and reorganisation of the innovation regime to
digital innovation regime. While earlier industrial regime components and capacities
have remained, their use has become increasingly realigned with digital capabilities. At
the same time, new capacities have been invented, creating a new kind of assemblage
where the old and new components have been realigned. As a result, the material
conditions and logics for innovation now fundamentally differ. The infrastructural
nature of the digital material (in that it is accessible at any place, time, device, and to
any user) fundamentally differs from that of physical infrastructure. The underlying
economics and problems of scale, scope, and supply-demand coordination change
accordingly. To understand how this shift has been possible, I briefly discuss the
differentiating features of digital material. Thereafter, I review how the digital material
has now become deeply embedded or connected to a wide range of socio-technical
relationships of most organised activity, offering radically different and novel leverage
points for innovation.
Digital material has been extensively theorised since about 2010 (see Kallinikos et al.,
2013; Tilson et al., 2010; Yoo et al., 2010) for its properties, such as reprogrammability
(Yoo et al., 2010), its use features including nonrivalry (Faulkner & Runde, 2019), and its
generative nature (Henfridsson & Bygstad, 2013; Lyytinen et al., 2017; Tilson et al., 2010;
Yoo et al., 2010; Zittrain, 2006). I build on these reviews with two specific goals in mind.
First, many properties suggested in the past overlap, and we still lack a coherent and
parsimonious set of properties that articulate what distinguishes digital material and its
operations from physical matter. It is also essential that identified properties are separate
8 K. LYYTINEN

from each other and from their higher-level effects (such as generativity). Second, most
analyses ignore that digital material comes in two forms, which are necessary and
complementary for any use: software (code) and data. Consequently, one should state
which properties apply to which type of digital material, that is, which ones are common
to both data and code and which ones apply to only one. I identify four properties of
digital that characterise all digital material. I also observe one property that applies to
software (code) only. These five properties offer a minimal foundation for articulating the
effects of the properties of digital material that are essential for building up digital
innovation regime.
All digital material is semiotic in nature in that it consist of strings of 0’s and 1’s drawn
from an alphabet of digital binaries to form digital representations.4 The property of
using bit strings allows exactness in expressing any digital object and makes such objects
addressable, that is, each can be retrieved based on their unique signature of 0’s and 1’s
(Yoo, 2010). A universal alphabet leads to homogenisation in that all digital materials, no
matter what they ultimately stand for – be it code, sensor data, audio, text, video – are
subject to the same representation scheme (Yoo et al., 2010). The semiotic and conven­
tion-based nature of representation opens up an abstract and unbounded space for
expression. Anything can be expressed and dressed in digital material as long as the
rules, conventions, and operations have been established whereby the strings of 0 and 1
are connected with one another and their counterparts outside the digital object. Digital
objects are not restricted by the limitations of the physical world; they are virtual
outcomes of mutual agreements of what each digital object stands for.
Von Neumann architecture offers a physical principle and generic solution to
manipulate and compute over digital objects. Use of von Neumann architecture
adds two important properties to digital material. First is the possibility to realise
and store bits in an electronic format. Electrification (and its ephemeral nature), in
contrast to using physical wiring-based computational solutions, makes digital
material editable and therefore fluid and ephemeral. Editability enables easy and
low-cost modification (Kallinikos et al., 2013), making digital objects malleable in
ways not possible for physical products (Yoo, 2010). Second, digital material is
executable and performative in contrast to conceiving the computation as just
a mathematical representation of a Turing machine abstraction. Due to electrifica­
tion, bit strings stand for executable operations on real machines that manipulate
manifestations of digital objects fast and with low cost. Commonly, the code guiding
such processes is referred to as software. It is important to understand full sym­
metry, which any programmer who has programmed in machine code knows:
software is also digital material and can be easily edited or made to refer to other
data (code). This makes software executable but also brittle.5 Furthermore, the
semiotic quality of code – the separation of code from physical realisation (perfor­
mance) – makes software reprogrammable. It allows creating an infinite variety of
software functions and related performances that are only constrained by the
imagination and skills of the code originators (Yoo et al., 2010). The expansive
and executable properties have led to an increased performative role of code in
business settings due to its ‘its apparent ability to “make things happen”’ (Mackenzie
& Vurdubakis, 2011, p. 6).
 INNOVATION: ORGANIZATION & MANAGEMENT 9

Modular layered architecture and embedding

Digital material is inherently modular. This helps express the code’s functionality at the
interfaces and hide how the functionality is implemented (Baldwin & Clark, 2000;
Parnas, 1972). Yet these interfaces are semantic agreements of what the bit strings
expressing the interface mean. In consequence, digital objects can be connected to
multiple types of topological arrangements as long as the semantics of the connections
at the interfaces have been sorted out. Because of performativity of material, the mod­
ularisation can be realised in run time, as currently implemented in cloud solutions and
web services. However, modularity ‘runs much wider and deeper in digital material’
(Kallinikos et al., 2013, p. 360). It is not constrained by physical features but dictated by
the semantics either in the form of code functions or data being represented and
transferred at the interface. By using gateways, there is also the possibility of creating
bridges between incompatible interfaces (Hanseth & Lyytinen, 2010). Combinations of
modularity and reprogrammability and use of gateways produce excess recombinability
of digital material (Faulkner & Runde, 2013). For example, software modules can be
freely reused (due to nonrivalry), and they do not suffer from wear and tear. They can be
connected through a large variety of interfaces or replaced with other modules. The same
applies to data as long as there is a way to combine meaningfully data to larger chunks
operated by the code (Kallinikos et al., 2013). This allows digital material to be organised
in multiple alternative ways across several modular hierarchies, which each manifests
specific functions or needs (Lyytinen, Yoo, & Boland, 2016; Yoo et al., 2010). This
property of digital material undergirds the concept of loose coupling available in digital
product architectures. Loose coupling guides organising code functions and computing
resources into multiple abstraction layers connected through shared interfaces, called
modular layered architecture (Tilson et al., 2010; Yoo et al., 2010) (Figure 1).
In the modular layered architecture (Figure 1), distinct functions and data are
organised into separate, loosely coupled layers that jointly create a service stack, offering
a digitally enabled service. This is the idea of the loosely coupled stack of N-layers
expressed in the middle of Figure 1. This generic product architecture underlies now
most products that share extensive digital components, from smartphones to cars and
automation systems (Lyytinen et al., 2016; Sandberg et al., 2020; Yoo et al., 2010). The
architecture follows von Neumann architecture in that the lowest layer represents the
‘executing’ (though replaceable) hardware that computes, stores, and plumbs the data.
The second layer represents the abstracted digital material as a semiotic entity – the
executable code (the stored programme) and data. The material can be organised further
into specific functions that deal with separate elements composing the computing task as
is normally done in organising the code: content layer (data/storage of data and code)
and service layer (computable code for users with a business function), connected
through an abstracted networking capacity (connectivity between devices). These func­
tions are shown in Figure 1 by the N digital layers. At the top, the computing process sits
an interpretation of what the computing is functionally/organisationally about, that is,
what the bit strings represent in terms of real-world actors, assets, exchanges, or services
(such as a ride share) in a
To innovate in any meaningful way, the architecture orchestrates sets of activities that
will tie the three layers together and make them coevolve over time. In this way, the
10 K. LYYTINEN

Real-world
phenomena

Layered
modular
architecture

Figure 1. Modular layered architecture and digital-embedding processes use context with its guiding
laws, constraints, norms, and habits.

digital and the two physical layers make the computing with the digital material tangible,
reliable, and scalable and the bits being manipulated meaningful for the setting. Most of
the processes that make this possible need to be worked out prior to the execution of the
service stack in any setting. They form necessary conditions for any successful digital
innovation, which involve the initiating element of creating an abstract digital material
(to-be-performed digital twin) and connecting this material back with the physical and
social worlds for novel outcomes. I posit that each process connecting the digital to the
material (or between digital layers) establishes a critical leverage point for digital innova­
tion to proceed.
The three sets of activities connecting the layers – what I call operational embedding,
virtual embedding, and contextual embedding – are shown by arrows 1–3 in Figure 1.6
They are further detailed in Table 1. They can be analytically inferred by probing the
properties of the digital material and imagining their consequences for processes where
digital material participates. ‘Operational embedding’ (arrow 1) refers to innovations that
embed sets of specific digital objects with a set of hardware to be executed and operated as
physical assets. In the opposite direction (upward), virtual embedding (arrow 2) refers to
how real-world phenomena, such as an organisation’s assets, actors, entities in physical
environment, immaterial objects (money, equity), and all sorts of activities (buying),
become virtualised through agreed-on representations in digital material. The goal in
virtualisation is eloquently and fully expressed in the idea of a digital twin: anything that
takes place or is represented in the physical/social world can have a mirror representation
in the digital material (Grieves & Vickers, 2017). Finally, contextual embedding (arrow 3)
refers to how the digital material, when performed, becomes contextualised into an
(organisational) setting and performed in ways that give a meaning, direction and new
functions (agency) to that setting (Mackenzie & Vurdubakis, 2011). This happens when
Table 1. Leverage points of digital innovation regime.
Three innova­
tion leverage Constraining
points Definition Outcomes Value conditions
Operational The digital material coded and stored physically and run on Operational embedding consists of sensing of digital material Exponential Constraints of physics
embedding hardware allowing sensing, recalling, connecting, computing, from input devices, storing and computing on them, and speed and Engineering
and display. actuating the results on output device embedded in scale challenges for
physical and social environments. Efficiency hardware and
Digital material needs to be embedded physically and run to (cost software systems
have performative effects. performance) Security
Digital innovation relies on exponential scale and speed Loose
advances and cost reductions from operational embedding coupling
so that real-world phenomena can be virtualised and through
performed. virtualisation
Virtual Process of virtualising real-world phenomena such as assets, Virtualisation makes focal phenomena (e.g. states, actors, Utility and What can be
embedding actors, activities, locations into a shared (semiotic) digital activities, places) representable in digital material such that value potentially
representation so that the code can perform on representations the code can perform and operate on the material in Expansion virtualised and
(a.k.a. creation of a digital twin). a meaningful way. Adaptability expressed
Created representations can be processed and performed only Virtualisation advances available social, institutional, and Modularity Challenges in
when operational embedding is present. organisational categorisations and enables integration of achieving
new types of properties into the evolving digital image of agreements on
the world (e.g. by adding a new sensor). representations
Digital innovation requires expansive forms of virtual
embedding in that the capture of digital material expands
permits new innovation searches that enable growth of
novel or more efficient services, products, and exchanges.
Contextual Computing contextualised into an organisational setting where Involves human actors interacting with hardware in ways that Utility and Social acceptance
embedding the digital material performs and gives a novel meaning and has significance to the unfolding of the activity. hedonistic Habits
function for the setting. Digital material contextualised as performances in the here value Regulations
and now in ways that change the actor’s proceeding and the Real-world Institutional
setting. material constraints
effects
Efficiency
INNOVATION: ORGANIZATION & MANAGEMENT

and
effectiveness
New activity
11
12 K. LYYTINEN

a trade is carried out in an electronic stock exchange as it is performed by a high-

frequency trading algorithm running on the cloud or when a mechanical engineer reads
results of a crash simulation and changes the design of a car.7
Generally, Figure 1 depicts digital material in grey and nondigital physical/social
matters in white. It expresses a bifurcated sandwich architecture that characterises
the digital innovation regime. This also clarifies why it radically differs from an
industrial regime. The latter operates primarily on material properties and their
representations (car design → car), whereas digital representations address coordi­
nation and control problems of associated activities. In this regime, innovation
consists sequences such as: material → (discovery) representation (synthesis) →
material (manufacturing) → representation (product information). Presentations
precede manufacturing (product designs), and representations follow manufacturing
and distribution of raw materials and products (e.g. accounting, production plan­
ning, operations data). In the industrial regime, the physical (atom) forms the
starting point of the regime, and the representation (bit) resides ‘on top of it’ in
select points to render design, manufacturing, and distribution possible by addres­
sing control and coordination problems.
In the digital innovation regime, digital material precedes the physical material. It is
rendered digital by virtual embedding, which forms its outcome – a new type of digital
material. This needs to be aligned with two temporally and organisationally distinct
materialisation processes called operational and contextual embedding. These processes
set up specific conditions and have specific effects on how digital innovation will unfold.
Ultimately, the virtualised digital material results in innovative outcomes in the real
world when all processes and properties are present. Consequently, all three processes
must be leveraged during digital innovation. The most common way is to rely on
multiple combinations of them over time. This promotes continued, fluid and expansive
forms of innovation (Lyytinen et al., 2016; Nambisan et al., 2017). A typical innovation
sequence in such digital innovation regime would consist of virtual (digital material) →
operational (performed material) → contextual (performative effects) → virtual (new
digital material) → operational (new performed material) → contextual (new performa­
tive effects) → contextual (use of the material in other context), and so on. I review each
process and related activities separately.

Operational embedding
From a computing perspective, digital material need to be embedded physically and made to
run on concrete hardware (what Dyson, 2012 calls ‘mud’). Without such embedding, the
digital material represents only a mathematical idea of computation. I call this embedding
operational embedding because it translates digital material to real operations on an instance
of concrete physical hardware (arrow 1 in Figure 1). The abstraction between the hardware
and digital material mirroring the hardware behaviours offers a design rule (and boundary)
between the physical hardware and its evolution and separates it from the evolution of the
digital material. The separation makes virtual embedding possible and separates digital
innovation from the industrial innovation regime, which does not have such embedding.
Operational embedding involves design and engineering principles to deploy some material
properties in ways that run the digital material faster and more efficiently or makes it easier,
 INNOVATION: ORGANIZATION & MANAGEMENT 13

more portable, more reliable, more fault tolerant, and so on. Embedding due to increases in
computing speed and scale (Moore’s law) enables further virtualisation (positive feedback
loop) (Kurzweil, 2004). As a result, organisations continually seek to carry out growing set of
their growing set of tasks by digitising related representations.
Generally, the hardware enabling operational embedding covers all elements of the
computing environment. These elements change over time and therefore how the code
can be run evolves to faster and less expensive embedding, changing the economies of
what can be computed. The computing environment consists of auxiliary code and
physical artefacts, which make the storage and representation of digital material as inputs
and outputs possible, more user-friendly, and easier (Yoo et al., 2010). Operational
embedding involves always a real physical system and takes place in a specific social
setting. Therefore, it comes with its sociotechnical design challenges (and innovations) in
that human actors need to be able to operate the computing environment by manipulat­
ing input and output devices, sensors, keyboards, and the like.
Over the past half century, engineers and computer scientists have advanced enor­
mously in making the bit turn faster, more efficiently, and easier. This has been witnessed
in exponential advances in electronics and miniaturisation expressed in Moore’s law
(Kurzweil, 2004). On one hand, the advances have moved computing from isolated, air-
conditioned halls into pockets, skin, and eyeglasses (Lyytinen & Yoo, 2002b). On the
other hand, it has enabled the use of massive server farms that compute in scale and
hyper-speed in the cloud.8 Ultimately, the exponential growth in the speed, scale, and
scope of operational embedding has enabled constant innovation in where, how, and for
what digital material can be used, pushing the use of digital material into a pervasive,
mass scale taking place anywhere, anytime (Lyytinen & Yoo, 2002a). This has conse­
quently promoted new forms of virtual embedding.

Virtual embedding
To be useful, digital materials needs to be tied up with some outside phenomena. This
entity, which the digital material stands for, can be anything in physical, social, real, or
other digital material (due its semiotic quality). For example, ride-sharing services like
Lyft and Uber, although being fully mediated by digital material and algorithmic pro­
cesses, need to be connected at some point to real cars, driven in real settings, with real
customers, transferring real money. For such a service to work, a lot of things need to be
virtualised prior the service: the driver has to have a digital identity tied to a real person;
a mobile phone needs to run the app representing and recording the activity and
executing the service performance; the route needs to be represented digitally as GPS
points; and the money needs to be transferred digitally by connecting to credit card
accounts of users and bank accounts of drivers. I called the process of abstracting focal
real-world phenomena to a set of sharable digital representations of virtual embeddings.
This embedding involves social processes and linguistic processes (and related innova­
tions) through which a set of focal salient real-world phenomena are mapped into digital
material in the form of specifications.9
Virtual embedding often begins with established categories in the existing social order,
such as categories of a taxi driver, his or her ID, and credit card number. To make
innovation fluid and expansive, the virtual embedding must expand the agreements in
14 K. LYYTINEN

digital material to new kinds of agreements and representations, such as use of GPS
coordinates, rating of drivers and customers, and so on. Generally, virtualisation as part
of digital innovation encompasses continued expansion, integration, combination, and
redeployment of digital material-related agreements as they are increasingly perfected to
create a digital mirror of the world. New types of virtual embedding arise from technical
advances, such as new sensors (e.g. kinetic sensors that can result in agreements of your
sleep and heartbeat patterns), new software functions (e.g. h-index in citation analyses),
and innovative categories enabled by digital representations (e.g. a shared playlist). All
such expansions open opportunities for innovating with digital services and products.
Finally, all performances on digital material can be recorded and represented in the
digital material (due to its semiotic quality). This creates perfect traces of how the
material has been performed, creating another wave of opportunities to invent new
categories using data mining (Alaimo et al., 2020). Recent developments in process
automation and robotics, for example, use intelligent assistants that digitally emulate
social and affective behaviours of human workers by showing affection and courteous
behaviour, things that previously were the domain of humans. A majority of the novelty
in use that accrues in the digital innovation regime originates from the increased value of
using digital material in multiple ways for different purposes. This is possible because of
the modularity of the material and its continued expansion and adaptation.

Contextual embedding
Whenever a real-world product or service becomes mediated by digital material, the
material and its performances need to be ultimately connected to some concrete setting
(such as trading activity, a ride share, or an autonomous car driving on a road). This
setting is where the digital material gets performed as part of a meaningful activity in the
social or physical world, such as conducting a trade or an driving for a ride-share service.
This is where the digital innovation meets the industrial regime and related physical/
social changes. The activity will have effects on the involved actor’s proceeding and the
setting (such as how the customer gets his or her ride). Performance on a circumscribed
digital material in that setting – that is, where both operational and virtual embedding are
present – moves in a reverse direction. The performance connects behaviourally to
a concrete social and physical setting that forms and is an instance of the upper material
tier of the sandwich. The move in reverse becomes the contextual embedding of the
performance of the digital material.
Ultimately, the escalating performative effects of digital material in our lives are the
outcome of the widening and intensifying virtualisation, which increases the number and
the intensity of touchpoints where digital performances interact with and within real-
world settings. In the early 2000s, the growth in such touchpoints was eloquently
expressed in the 5A slogan: ‘any time, any place, any service, any network, any
device.’10 Contextual embedding and its deepening and intensification offer leverage
points for expanding the digital innovation regime. The growth in the touchpoints gained
through continuously intensified contextual embedding offers, on one hand, new ways to
virtualise further, and on the other hand, ways to increase the effects of performances by
introducing new performances (due to the access to the setting where digital material is
performed). The digitisation levels of settings – covering both virtualisation and
 INNOVATION: ORGANIZATION & MANAGEMENT 15

operational embedding of computing resources – still vary significantly across different

use contexts. These embeddings and their proceedings are path-dependent processes
expressed in technology standards, regulations, cultural and institutional scripts, and
habits that all regulate the use context and contextual embedding. The ways the con­
textual embedding takes place therefore differ significantly from one context to another.
This applies even to the same digital innovation and related product and service that
relies on the same virtual and operational embedding.11 This variation can also explain
the intense shake-up of the music industry when musical content and it representations
were virtualised and contextualised into new settings mostly by outside innovators, such
as Napster, Apple, and Spotify. They leveraged continuous innovations in the operational
embedding (e.g. the internet, peer-to-peer networks, servers, MP3 players, smartphones,
the cloud) and contextual embedding (a setting without a strong enforcement of existing
regulations). The greater resilience of the newspapers can be explained in how the
contextual embedding in this industry followed alternative routes, because of subscrip­
tion models, advertisement money, and differences in experiences offered by having
access to physical paper. No mapping into digital representations is perfect – it’s just
a momentary agreement – but there are limits as to how far virtualisations can be
performed across contexts (e.g. challenges how to represent music for varying groups
such as classical music listeners).

Digital innovation regime: towards a research agenda

This essay has sought to conduct a broad systemic, comparative analysis that has tried to
clarify the essential differences of where the innovations originate, the constraints on
such innovation, and the primary logics for creating novelty and value in the industrial
and digital innovation regimes. The analysis expands past work of the differences in the
product architecture (Gawer, 2014; Yoo et al., 2010) and principles of organising (Gawer,
2014; Parker et al., 2016) to a broader comparative analysis of innovation systems. This
allowed me to identify three leverage points, which make digital innovation possible: (1)
operational embedding, (2) virtual embedding, and (3) contextual embedding. These
embedding processes and their properties are summarised in Table 1. By perusing Table
1, we can understand why the three embedding processes form the epicentre of digita­
lisation. The intellectual merit of such analytical separation is that it invites us to focus on
the dynamic, evolving interplay between digital, the social, and the institutional and how
that interplay evolves and transforms when new virtualizations are invented and their
operational materialisation evolves. The processes are highly distinct from the leverage
points of the industrial innovation regime, which relied on the discovery of properties of
the matter and organisational capacities to synthesise them into usable functions and
forms needed for fulfiling human needs. Another critical leverage point was how to
manufacture and distribute these products in ways that reached scale and scope
economies.
The three separate leverage points entail that the rift of digital innovation regime from
the industrial regime runs deeper than the idea of separating digitising and digital
information from analogous information processing (Tilson et al., 2010). Each embed­
ding offers a lens to understand a unique condition for furthering digital innovation,
where each process extends beyond the technical concept of digitising by illuminating
16 K. LYYTINEN

how digitalisation as a concrete, contextual, sociotechnical process becomes deeply

transformative in separate spheres of organised activity. Each embedding connects the
digital material to and with varied technical manifestations that are performed across
multiple social worlds.
The three processes offer some germane insights that explain the unprecedented
generativity associated with digital technologies and material (Lyytinen et al., 2017;
Tilson et al., 2010) (see Table 1). First, the three processes operate independently in
separate contexts and under different logics; they assume different expertise and knowl­
edge and pursue different goals and ways of evaluating solution quality for each embed­
ding. Second, these separations are only analytically meaningful. Each enable and is
conducive for activities in other embedding. For example, the more prominently digital
material presents and performs the social and the material world (i.e. the more real-world
phenomena become digitised and are performed as part of the world), the more sig­
nificant these embedding processes and their leverage become in shaping behaviours and
bringing novelty. Likewise, the more activities are performed with the digital material,
the more of the phenomena, not yet digitised but with the potential for virtualisation,
need to be virtualised for furthering digital innovation. Third, as the range and diversity
of digital material grows, it needs to be more contextualised. This calls for more
improvements and innovations in operational and contextual embedding, creating
a strong positive feedback cycle.
The research in digitising has always started by observing the critical importance of
advances in operational embedding. In different forms, this ultimately drives all digital
innovation (Kurzweil, 2004). Indeed, major improvements in this embedding have been
necessary to advance digital innovation from its humble beginnings as electronic data
processing in the 1960s to its current transformative power. Hardware is absolutely
necessary to render digital material from a mathematic abstraction into meaningful
social activity, and its cost needs to decrease at exponential rate to perform a growing
bulk of digital material. This has been enabled by and called for improvements in
understanding the physics of computing and architecting solutions that reliably compute
in scale and speed (e.g. recent virtualisation to cloud). These advances have changed the
overall organising of organisational computing functions, such as how to make available
processing power, storage, and connectivity for organisational tasks and members. As
a result, computing is now becoming a market-based utility organised in loosely coupled
forms to deliver varying services using market-based mechanisms. This has created the
need to innovate around new manufacturing processes to change the scale and scope of
the operational embedding along the principles that characterise the industrial innova­
tion regime. In operational embedding, the critical challenges currently related to the
economics of scale and scope for varying needs for operational embedding and the
increased quest for reliability and security, which are necessary for the continued
expansion of operational embedding.
The analysis highlights the critical nature of virtual embedding. This forms the
industrial, organisational, and social foundation of digitalisation and lays a necessary
foundation for furthering digital innovation. Through this embedding, real-world phe­
nomena become virtualised and subjected to common semantic agreements and orga­
nisational processes enabled by such agreements. This process relies on social and
linguistic conventions and standard-based agreements and is therefore unbounded by
 INNOVATION: ORGANIZATION & MANAGEMENT 17

physical constraints. It cuts across domains and context whenever there is a possibility to
create a way to semantically connect any two elements of digital material (Kallinikos
et al., 2013). Because of its creative, unbounded nature, the embedding forms the key
driver in furthering digital innovation and generativity. Every novel service, product
innovation, and digital feature calls for original semantic agreements, related perfor­
mances, and their sharing in ways that offers value and utility. The embedding explains
why new data and discovery of their connections through algorithms now lay the
foundation for continued digital innovation. It explains the heightened interest in the
value of data as the new oil of economy (Economist, 2020). The embedding also explains
the recent growth of the platforms and their role in economic activity and organising
(Gawer, 2014; De Reuver et al., 2017; Yoo, 2010). All platforms rely on virtualisation (1)
to represent any salient product, object, or feature in the real world to be shared or
exchanged (such as music, news, games, friends); and (2) to perform exchanges enabled
by such representations and collect data of such performances for future prediction.
These processes rest initially on cognitive, social inventions of sense-giving and sense-
making associated with attaching alternative or expanded meanings to semiotic, digital
material. The continued growth of virtual embedding and its combinatorial nature
(Alaimo et al., 2020) promotes the expansion of digital innovation and forms the key
leverage point for value creation and utility of the innovation. Although a fair amount of
research has been conducted on contexts and processes of virtualisation such as those
related social media (Alaimo et al., 2020), travel exchanges (Orlikowski & Scott, 2015),
stock exchanges and trading (Mattli, 2019), or physical products (Grieves & Vickers,
2017) we lack more systematic accounts how such processes unfold and why they succeed
or fail, and how they connect to and depend on activities involve operational embedding
and contextual embedding of this virtualised digital material.
Contextual embedding is the most visible and obvious part of digital innovation. On
one hand, the embedding resembles the adoption of innovations under the industrial
innovation regime. Ultimately, some actor needs to find value in using digital material
performances in some setting – such as its economic value, information benefit, useful­
ness, pleasure, or gratification. Such experience will ultimately drive the decision on
adopting such innovation. Consequently, contextual embedding and its explanations
resemble accounts of traditional product diffusion processes, such as how and why
telephone use diffuses in a user populations. Like with telephone, the diffusion is
conditioned by how performances of the digital material are regulated in each use
context, such as recent regulations of tracking software during the COVID-19 pandemic.
The diffusion is also is driven by the cost of use and level of effort required to assimilate
the innovation. The latter is largely determined by the complexity and versatility of the
innovation and the actor’s absorptive capacity. On the other hand, the contextual
embedding process differs significantly from diffusing industrial products. First, it allows
and draws on expansive virtualisation of digital innovation to further innovation.
Because of the conventional semantics of the digital material, the diffusion allows
constant repurposing of the meaning and uses of the digital material often to unexpected
purposes and settings. Scholars call this high interpretive flexibility of digital material
(Arvidsson et al., 2014). The success of contextual embedding depends highly on the state
of the operational embedding and to what extent and how the digital material perfor­
mances have been properly contextualised with specific hardware. This may change
18 K. LYYTINEN

during the use and consequently change the value of the innovation. For example, ride
sharing as a digital innovation became possible only when the whole service could be
contextually embedded and performed using powerful enough and common smart­
phones owned by drivers that could run both the ride-sharing app and a navigation
service (complementarity) as part of the service. The former offered a means to make any
car a taxi at the moment the driver and the car were connected to the service, and the
latter replaced the driver’s knowledge of how to drive from point A to point B with
a performed digital material (mapping service). This allowed any car owner to connect to
a ride-sharing service (Tilson et al., 2021).
Based on the foregoing analysis, digital Innovation regime is generally organised into
four phases: (1) discover product or behaviour (ideation); (2) abstract to digital material
(virtualise); (3) implement and replicate digital material as performance (variations in
operational/contextual embedding); and (4) deliver product and/or behaviour though
a performance of digital material. In this sequence, only the first phase is similar to the
industrial innovation regime. The innovation starts with a novel idea that takes advan­
tage of some novel features of the physical world or introduces novelty into the social
world or individual behaviours. In the next phase, the idea must be virtualised into
a model of a digital material specification (data, code) which is then implemented,
replicated, and contextualised through ongoing operational and contextual embedding.
The implementation and replication phase is deferred or ephemeral in the sense that the
actual performance of the product is ultimately determined by the concrete digital
material/hardware combination during runtime when the innovation is deployed as
performance at the delivery phase of innovation.
Generally, in the digital regime, innovation moves from the idea first to the middle of
the sandwich, that is, into addressing the correct digital material representation, and then
to forms of carrying out the operational and contextual embedding through implemen­
tation and replication. The logic of physical discovery still applies with physical products
(hardware) instrumental for implementing and replicating digital material through
operational and contextual embedding. But this logic is not a central pathway towards
innovation. It is replaced by another (inner) semiotic layer that follows a different logic of
discovery of semantic categories and behaviours. The logic of physical discovery, synth­
esis, and manufacturing and related costs represents a boundary condition of how to
replicate and effectively deliver digital material performances. Because of this, the
innovation is incomplete – there is always more to virtualise and contextualise – and it
can change goals and functions due to high interpretive flexibility and the iterative nature
of the innovation. One reason for this is that the three processes largely run on
orthogonal paths. The initial operational embedding nearly always differs from how
the operational embedding is carried out during the actual delivery. Similarly, the
contextual embedding can be varied significantly over time depending on the chosen
materialisation and delivery strategy (due to the loosely organised service stack of
N-digital layers in Figure 1). The phases and related embedding processes are never
sequential, but always interleave and run parallel. Moreover, the feedback loops between
the phases are stronger and often immediate. This results in the high fluidity in the
innovation scope, activities, participants, and diffusion outcomes (Nambisan et al., 2017).
Another unique characteristic is the increased speed of innovation and scaling: it is easier
to shape and modify digital representations and change related agreements than to shape
 INNOVATION: ORGANIZATION & MANAGEMENT 19

physical material and change how to shape it and distribute it. It is also easier to scale
(Nambisan et al., 2017). As a result, digital innovation processes advance and change at
different rates and speed compared with innovation processes in the industrial innova­
tion regime.
The point of this essay is to identify essential differences in what enables and is
conducive for digital versus industrial innovation and how they organise as systems.
While writing this piece, I was motivated by the need to increase clarity of the new logic
of digital innovation – a topic I initiated with my colleagues in Yoo et al. (2010) that
clarified the impact of the N-layer architecture of the digital service stack on innovation.
I expand this view to the sandwich model of how the stack relates to innovation
environment, and I show that the differences in the logic are centred on differences in
leverage points in the system where actors can produce novelty. In the industrial regime,
this is circumscribed broadly into the discovery and synthesis and the manufacture and
distributed phases. In the digital innovation regime, the points are circumscribed into
expansive changes in the operational, virtual, and contextual embedding and their
mutually interdependent dynamics, which over time ushers the digital innovation for­
ward. The next step in the analysis will be to expand and map how the material world
increasingly melds with digital material. This will enable new types of interactions and
performance in material and social environments (such as robotics and autonomous
cars). This shift will involve a new round of discovery about how the three processes need
to be carried out and can be leveraged to expand digital innovation. The next round of
change will cut into pieces the industrial regime from which the new digital innovation
regime was borne.

Notes
1. There are several excellent historic analyses of how such models emerged and were perfected
during a period from 1850 to 1970 (see Chandler, 1990; Hughes, 1989).
2. This argument is simplified in purpose somewhat. Industrial innovation in can be also
nonlinear in that manufacturing capabilities may already be available from which new
innovations emerge, for example, by repurposing some existing capabilities for COVID-
19 equipment. Or modular platforms can be in place before ideas come into fruition, as
exemplified by investments in traditional technology platforms for automobiles, or in
knowledge platforms such as the ones in 3M (Karnøe & Garud, 2012).
3. How such digital technologies were organised in car industry at the height of this period is
reported in King and Lyytinen (2004).
4. We call digital representation any string of 0 and 1 that has a shared agreed meaning in any
given context. A digital object is a digital representation with clear definition and use
purpose in a setting like a payroll record or a CAD/CAM model.
5. Just a change of one bit’s value can fundamentally change the meaning of the code.
6. Note that in Figure 1 the direction of the embedding does not match with the direction of
the arrows, that is, up or down, but, instead it conveys whether it is the digital (grey) being
embedded with physical (white), or vice versa.
7. Both examples have a different and evolving operational and virtual embedding. In the first
phase, those cover agreements and digital representations of stocks, prices, money, trades,
and so on. In the latter they involve representations of car geometry, hit strength or point,
weight, speed, and so on.
8. Recently, this embedding has become a subject of increased virtualisation (called cloud
computing), whereby the connection between the digital and hardware has been rendered
20 K. LYYTINEN

dynamic and non-transparent by adding a separate digital layer that coordinates and
controls how the digital material hits the hardware in a distributed way during a dynamic
embedding of the operations.
9. These presentations follow strictly Ludwig von Wittgenstein’s famous last proposition in
Tractactus: ‘Whereof one cannot speak in ‘0ʹ and ‘1ʹ, thereof one must be silent.’
(Wittgenstein, 1922). In this sense, virtual embedding has a fixity and comes with social
inertia related to agreements. But the form and structure of representation can be changed
relatively easily as well as what the agreement means (a common trick in repurposing data
fields for other uses in many applications).
10. Recently, the device is rather any product because cars, refrigerators, and faucets may share
computing capacity.
11. Consider rules and conventions around how we are expected to use mobile phones and
related services in different settings in work and family life and at different times of the day.

Acknowledgments
I am grateful to Robin Gustafsson and Kimmo Karhu for lively discussions, where many of the
ideas originated and were debated and shared. Special thanks to Janis Kallinikos for moral support
and shared interest and sharp comments on the topic. I thank the guest editors of this special issue,
Llewellyn D.W. Thomas, Youngjin Yoo, and Raghu Garud, for their insightful comments and
suggestions for improvement. All mistakes, errors, and misunderstandings remain solely mine.

Disclosure statement
No potential conflict of interest was reported by the author(s).

References
Alaimo, C., Kallinikos, J., & Valderrama, E. (2020). Platforms as service ecosystems: Lessons from
social media. Journal of Information Technology, 35(1), 25–48. https://doi.org/10.1177/
0268396219881462
Argyres, C. (1999). The impact of information technology on coordination: Evidence from the B-2
“Stealth” bomber. Organization Science, 10(2), 162–180. https://doi.org/10.1287/orsc.10.2.162
Arvidsson, V., Holmström, J., & Lyytinen, K. (2014). Information systems use as strategy practice:
A multi-dimensional view of strategic information system implementation and use. Journal of
Strategic Information Systems, 23(1), 45–61. https://doi.org/10.1016/j.jsis.2014.01.004
Baldwin, C. Y., & Clark, K. B. (2000). Design rules: The power of modularity. MIT Press.
Baskerville, R., Myers, M., & Yoo, Y. (2020). Digital first: The ontological reversal and new
challenges for information systems research. MIS Quarterly, 44(2), 509–523. https://doi.org/
10.25300/MISQ/2020/14418
Chandler, A., Jr. (1977). The visible hand: The managerial revolution in American business. Belknap
Press of Harvard University Press.
Chandler, A., Jr. (1990). Scale and scope: The dynamics of industrial capitalism. Belknap Press of
Harvard University Press.
Constantinides, P., Henfridsson, O., & Parker, G. G. (2018). Introduction—platforms and infra­
structures in the digital age. Information Systems Research, 29(2), 381–400. https://doi.org/10.
1287/isre.2018.0794
De Reuver, M., Sørensen, C., & Basole, R. C. (2017). The digital platform: A research agenda.
Journal of Information Technology, 33(2), 1–12. https://doi.org/10.1057/s41265-016-0033-3
DeLanda, M. (2006). A new philosophy of society: Assemblage theory and social complexity. A&C
Black.
Dyson, G. (2012). Turing’s cathedral: The origins of the digital universe. Pantheon Books.
 INNOVATION: ORGANIZATION & MANAGEMENT 21

Economist. (2020). Are data more like oil or sunlight? https://www.economist.com/special-report

/2020/02/20/are-data-more-like-oil-or-sunlight
Faulkner, P., & Runde, J. (2013). Technological objects, social positions, and the transformational
model of social activity. MIS Quarterly, 37(3) 803–818.
Faulkner, P., & Runde, J. (2019). Theorizing the digital object. MIS Quarterly, 43(4), 1279–1302.
https://doi.org/10.17863/CAM.40371
Gawer, A. (2009). Platforms, markets and innovation. Edward Elgar.
Gawer, A. (2014). Bridging differing perspectives on technological platforms: Toward an inte­
grative framework. Research Policy, 43(7), 1239–1249. https://doi.org/10.1016/j.respol.2014.
03.006
Grieves, M., & Vickers, J. (2017). Digital twin: Mitigating unpredictable, undesirable emergent
behavior in complex systems. In F.-J. Kahlen, S. Flumerfelt, & A. Alves (Eds.), Transdisciplinary
perspectives on complex systems (pp. 85–113). Springer. https://doi.org/10.1007/978-3-319-
38756-7_4
Hanseth, O., & Lyytinen, K. (2010). Design theory for adaptive complexity in information
infrastructures. Journal of Information Technology, 25(1), 1–19. https://doi.org/10.1057/jit.
2009.19
Henfridsson, O., & Bygstad, B. (2013). The generative mechanisms of digital infrastructure
evolution. MIS Quarterly, 37(3), 907–931. https://doi.org/10.25300/MISQ/2013/37.3.11
Hughes, T. (1989). American genesis: A century of invention and technological enthusiasm, 1870–
1970. University of Chicago Press.
Kallinikos, J., Aaltonen, A., & Marton, A. (2013). The ambivalent ontology of digital artifacts. MIS
Quarterly, 37(2), 357–370. https://doi.org/10.25300/MISQ/2013/37.2.02
Karhu, K., Gustafsson, R., & Lyytinen, K. (2018). Exploiting and defending open digital platforms
with boundary resources: Android’s five platform forks. Information Systems Research, 29(2),
479–497. https://doi.org/10.1287/isre.2018.0786
Karnøe, P., & Garud, R. (2012). Path creation: Co-creation of heterogeneous resources in the
emergence of the Danish wind turbine cluster. European Planning Studies, 20(5), 733–752.
https://doi.org/10.1080/09654313.2012.667923
King, J., & Lyytinen, K. (2004). Automotive informatics: Information technology and enterprise
transformation in the automobile industry. In W. Dutton, B. Kahin, R. O’Callaghan, &
A. Wyckoff (Eds.), Transforming enterprise: The economic and social implications of information
technology (pp. 283–331). MIT Press.
Kurzweil, R. (2004). The law of accelerating returns. In C. Teuscher (Ed.), Alan Turing: Life and
legacy of a great thinker (pp. 381–416). Springer.
Lee, J., & Berente, N. (2012). Digital innovation and the division of innovative labor: Digital
controls in the automotive industry. Organization Science, 23(5), 1428–1447. https://doi.org/10.
1287/orsc.1110.0707
Lyytinen, K., Sorensen, C., & Tilson, D. (2017). Generativity in digital infrastructures: A research
note. In D. Galliers & K. Stein (Eds.), The Routledge companion to management information
systems (pp. 253–275). Routledge.
Lyytinen, K., & Yoo, Y. (2002a). The next wave of nomadic computing: A research agenda for
information systems research. Information Systems Research, 14(4), 377–388. https://doi.org/10.
1287/isre.13.4.377.75
Lyytinen, K., & Yoo, Y. (2002b). Issues and challenges in ubiquitous computing—introduction.
Communications of the ACM, 45(12), 62–65. https://doi.org/10.1145/585597.585616
Lyytinen, K., Yoo, Y., & Boland, R. (2016). Digital product innovation within four classes of
innovation networks. Information System Journal, 26(1), 47–75.
Mackenzie, A., & Vurdubakis, T. (2011). Codes and codings in crisis: Signification, performativity
and excess. Theory, Culture & Society, 28(6), 3–23. https://doi.org/10.1177/0263276411424761
Malone, T., Yates, J., & Benjamin, R. (1987). Electronic markets and electronic hierarchies.
Communications of the ACM, 30(6), 484–497. https://doi.org/10.1145/214762.214766
Mattli, W. (2019). Darkness by design. Princeton University Press.
22 K. LYYTINEN

Nambisan, S., Lyytinen, K., Majrzchak, A., & Song, M. (2017). Digital innovation management:
Reinventing innovation management research in a digital world. MIS Quarterly, 41(1), 223–238.
https://doi.org/10.25300/MISQ/2017/41:1.03
Orlikowski, W., & Scott, S. V. (2015). The algorithm and the crowd: Considering the materiality of
service innovation. MIS Quarterly, 39(1), 201–216. https://doi.org/10.25300/MISQ/2015/39.1.09
Parker, G. G., Van Alstyne, M. W., & Choudary, S. P. (2016). Platform revolution: How networked
markets are transforming the economy and how to make them work for you. Norton.
Parnas, D. L. (1972). On the criteria to be used in decomposing systems into modules.
Communications of the ACM, 15(12), 1053–1058. https://doi.org/10.1145/361598.361623
Rogers, E. (2005). Diffusion of innovations (5th ed.). Free Press.
Sandberg, J., Holmström, J., & Lyytinen, K. (2020). Digitization and phase transitions in platform
organizing logics: Evidence from the process automation industry. MIS Quarterly, 44(1),
129–153. https://doi.org/10.25300/MISQ/2020/14520
Schilling, M. A. (2000). Toward a general modular systems theory and its application to interfirm
product modularity. Academy of Management Review, 25(2), 312–334. https://doi.org/10.5465/
amr.2000.3312918
Tilson, D., Lyytinen, K., & Sørensen, C. (2010). Digital infrastructures: The missing IS research
agenda. Information Systems Research, 21(4), 748–758. https://doi.org/10.1287/isre.1100.0318
Tilson, D., Sørensen, C., & Lyytinen, K. (2021). Digitally induced industry paradoxes: Disruptive
innovations of taxiwork and music streaming beyond organizational boundaries. Research in
the Sociology of Organizations, Part 73A, 175–196.
Wittgenstein, L. (1922). Tractatus logico-philosophicus. Routledge & Kegan Paul. Chicago (2015,
15th ed.).
Yates, J. (1989). Control through communication: The rise of system in American management.
Johns Hopkins University Press.
Yoo, Y. (2010). Computing in everyday life: A call for research on experiential computing. MIS
Quarterly, 34(2), 213–231. https://doi.org/10.2307/20721425
Yoo, Y., Boland, R. J., Jr., Lyytinen, K., & Majchrzak, A. (2012). Organizing for innovation in the
digitized world. Organization Science, 23(5), 1398–1408. https://doi.org/10.1287/orsc.1120.0771
Yoo, Y., Henfridsson, O., & Lyytinen, K. (2010). Research commentary—The new organizing logic
of digital innovation: An agenda for information systems research. Information Systems
Research, 21(4), 724–735. https://doi.org/10.1287/isre.1100.0322
Zittrain, J. L. (2006). The generative internet. Harvard Law Review, 119(7), 1975–2040. https://doi.
org/10.1145/1435417.1435426
 Technological Forecasting & Social Change 102 (2016) 225–239

Contents lists available at ScienceDirect

Technological Forecasting & Social Change

Additive manufacturing for consumer-centric business models:

Implications for supply chains in consumer goods manufacturing
Marcel Bogers a,⁎, Ronen Hadar b, Arne Bilberg b
a
University of Copenhagen, Denmark
b
University of Southern Denmark, Denmark

a r t i c l e i n f o a b s t r a c t

Article history: Digital fabrication—including additive manufacturing (AM), rapid prototyping and 3D printing—has the potential
Received 13 January 2014 to revolutionize the way in which products are produced and delivered to the customer. Therefore, it challenges
Received in revised form 22 May 2015 companies to reinvent their business model—describing the logic of creating and capturing value. In this paper,
Accepted 28 July 2015
we explore the implications that AM technologies have for manufacturing systems in the new business models
Available online 26 September 2015
that they enable. In particular, we consider how a consumer goods manufacturer can organize the operations
Keywords:
of a more open business model when moving from a manufacturer-centric to a consumer-centric value logic. A
3D printing major shift includes a move from centralized to decentralized supply chains, where consumer goods manufac-
Additive manufacturing turers can implement a “hybrid” approach with a focus on localization and accessibility or develop a fully person-
Business models alized model where the consumer effectively takes over the productive activities of the manufacturer. We discuss
Digital fabrication some of the main implications for research and practice of consumer-centric business models and the changing
Glocalized production decoupling point in consumer goods' manufacturing supply chains.
Rapid manufacturing © 2015 Elsevier Inc. All rights reserved.
Rapid prototyping
Supply chains

1. Introduction Manufacturers of customized products in domains as dental, bio-

medical, fashion and apparel have so far successfully adopted AM tech-
Additive manufacturing (AM), which is also known as rapid nologies. The hyper flexibility of AM technologies allows for customized
manufacturing or 3D printing, has emerged as a new and disruptive shapes, digital interaction with consumers and direct manufacturing,
manufacturing technology that has major implications for companies which gives benefits in terms of lower costs, reduced supply chain
and industries at large (Phaal et al., 2011). Given that the AM industry complexity and lead times, etc. However, despite the potential, many
is currently assessed at more than $3 billion, with an expected rise to questions pertain, for example related to the justification for mass man-
$13 billion by 2018 and $21 billion by 2020 (Wohlers, 2014), AM tech- ufacturers of commodity products to use AM technologies, the types of
nologies have an enormous potential, although they also imply impor- business models that they would have to employ to capitalize on the
tant and necessary changes to companies' business models—their logic flexibility that AM offers, and how these changes would affect their op-
of creating and capturing value (Afuah, 2014; Zott et al., 2011). As a erations and supply chain structures. Accordingly, our research question
hyper-flexible technology that can provide highly customized and is: How do emerging AM technologies impact business model develop-
personalized products and production, AM provides a specific set of op- ment and operations in consumer goods manufacturing?
portunities and challenges for developing new business models (Piller In this paper, we explore the new possibilities and challenges that
et al., 2015; Ponfoort et al., 2014). In this paper, we analyze the recent AM presents to consumer goods manufacturers' business models with
advances in AM technologies and we explore their implications for busi- a particular focus on the potential to open up to a higher degree of con-
ness models in the consumer goods manufacturing industry, where sumer involvement and on the associated implications for the organiza-
they have a big potential to revolutionize the way products are tion of production activities. In particular, shifting productive activities
produced (Berman, 2012; Gibson et al., 2010; Huang et al., 2013; Tuck from manufacturers to consumers challenges the centralized nature of
et al., 2007). production systems and thus calls for a decentralization of supply
chains. We will present an inductive study that is based on the general
developments within AM technologies in the context of the consumer
goods manufacturing industry. Specifically, we will propose that AM
⁎ Corresponding author at: University of Copenhagen, Department of Food and
Resource Economics, Section for Production, Markets and Policy, Rolighedsvej 25, 1958
technologies fundamentally change the role of the consumer in
Frederiksberg C, Denmark. consumer goods manufacturers' business models with a particular
E-mail address: [email protected] (M. Bogers). implication being that supply chains are becoming more distributed

http://dx.doi.org/10.1016/j.techfore.2015.07.024
0040-1625/© 2015 Elsevier Inc. All rights reserved.
226 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239

and decentralized to enable more personalized production of consumer to derive business models that could leverage the latent value of these
goods. Effectively, productive activities shift from the manufacturer to technologies. In this iterative process, we identified particular charac-
the consumer, which leads to a need to decentralize and decouple the teristics of the business model that could be derived from considering
organization of the manufacturer's supply chain to embrace the central both the general developments in the industry and the particular devel-
role of the individual consumer in the value creation-capture process. opments within the focal consumer goods manufacturer. Ultimately,
this led to a specification of key business model design parameters
2. Background and related implications for supply chains.

There is a diversity of perspectives on the business model concept in

the literature, without a wide consensus regarding its precise conceptu- 4. Additive manufacturing: from production technologies to
alization (Zott et al., 2011). In general, a company's business model business models
describes its logic of creating and capturing value (Afuah, 2014;
Osterwalder and Pigneur, 2010; Zott et al., 2011). The concept arose Here, we describe the state of the art of AM technologies, starting
from the emergence of the Internet, the growth of emerging markets, with AM as a manufacturing concept, and then leading up to a descrip-
and the appearance of postindustrial technologies (Zott et al., 2011). tion of how AM fundamentally changes the logic of how companies can
Consequently, the growth of e-businesses promoted the need for a create and capture value (i.e. business models).
value-capturing model as a reaction to the value potential that was cre-
ated (Afuah and Tucci, 2001). On a more detailed level, a business model
refers to a system of interdependent activities within and across the or- 4.1. An overview of additive manufacturing technologies
ganizational boundaries that enables the organization and its partners
to create value and capture part of that value (Zott and Amit, 2010). In this paper, we refer to AM as the utilization of additive technolo-
Amit and Zott (2001) moreover present a specific framework that com- gies for the production of customer-specific consumer goods. In
prises efficiency, complementarities, lock-in, and novelty to determine contrast to rapid prototyping—the use of additive technologies for the
the value creation logic. Given that activities within a business model manufacturing of single or multiple prototypes—AM is in principle
can also take place across organizational boundaries, the business repeatable and scalable as a production process. AM technologies
model determines the logic of purposively managed knowledge flows have existed since the beginning of the 1980s—initially mostly as a
in open innovation (Chesbrough and Bogers, 2014). Open business prototyping tool—and they have recently emerged as a viable
models accordingly use the “division of innovation labor” to create manufacturing technology due to significant improvements in part
greater value by leveraging more ideas, resources and other assets quality, price and manufacturing process time. Principles such as
that are available outside of the companies' boundaries (Chesbrough, “lean” and “just in time” can also be considered here in the context of
2006; Frankenberger et al., 2013; Vanhaverbeke and Chesbrough, full-scale small batch production, with a focus on the customer and cre-
2014). ating value, with more or less waste (“muda”) (cf. Tuck et al., 2007).
Recently, AM, a hyper-flexible technology that can provide highly In order to understand the different AM platforms and technologies,
customized and personalized products and production, provides a new we conducted a detailed state-of-the-art1 analysis of AM technologies.2
set of opportunities for developing a new logic for creating and capturing We explored the different AM technologies' characteristics, advantages
value from such products and processes (cf. Piller et al., 2015; Ponfoort and disadvantages, and their feasibility for consumer goods production.
et al., 2014; Wohlers, 2013, 2014). These changes imply enormous Table 1 presents the six leading technologies that we identified, while
challenges—not the least for incumbent manufacturers—addressing var- Appendix A provides a more detailed overview of each technology, its
ious aspects of traditional business models, such as the value proposition, method of operation, and its current main use.
cost structure and value chain (e.g. Afuah, 2014; Chesbrough and The basic characteristics of these emerging AM technologies have
Rosenbloom, 2002; Osterwalder and Pigneur, 2010). Given that there important implications for consumer goods production systems. Based
are different AM technologies available in the market and that traditional on our analysis, we identified a number of dimensions that we consider
manufacturing technologies are still widely used, manufacturing compa- essential to assess the feasibility of each of the AM technologies (see
nies need to explore or experiment with new business models based on Appendix B for a more detailed assessment).3 These dimensions are
the emerging technologies (Brunswicker et al., 2013; McGrath, 2010). the components of production that must be satisfied for AM to be recog-
These exploratory processes imply an important interaction between nized as a feasible production concept. Table 2 shows a description of
technology and business model innovation (cf. Baden-Fuller and the dimensions as well as the results of the comparison of the different
Haefliger, 2013), while furthermore making the link to the organization technologies based on these dimensions. In the table, a score of 5
of production, including supply chains (cf. Bogers et al., 2015; Johnson represents high (i.e. optimal) results, while a score of 1 represents low
and Whang, 2002; Koren, 2010). (i.e. critical) results. Results are benchmarked to existing injection
molding manufacturing capabilities and have been evaluated with the
3. A note on research design use of experts in AM production processes. While Appendix B provides
a more detailed assessment of the AM technologies in consumer goods
The empirical base consists of the general developments in AM tech- production systems, we will focus (below) on the overall assessment
nologies, business models, and supply chains, although we also rely on and subsequently present the implications for business models.
the recent experience of a large internationally-oriented manufacturer
1
within the plastic component industry. The company has been utilizing The analysis was concluded in 2014, which is therefore the reference point for this
digital fabrication for the purpose of prototyping for more than 20 years, analysis.
2
We note that the technologies explored in this paper are polymer AM production
and it has recently been working toward the adaptation of AM as con-
technologies alone. Several technologies use metal as well for rapid tool manufacturing
sumer goods manufacturing concept. While some of the observations and general metal production, such as Selective Laser Melting (SLM) or Metal Selective La-
and analyses in this paper are based on the company's experience, our ser Sintering (MSLS). Other technologies use wax and ceramics as base materials. These
ultimate objective is to present the general case of the consumer technologies will not be explored in this paper as the focus (in the consumer goods man-
goods manufacturing industry. ufacturer that serves as the empirical base) is plastic parts production for consumer goods.
3
Our analysis is primarily based on the assessment of the potential of AM technologies
Based on the original research that we conducted, with the above- for the large consumer goods manufacturer that serves as the empirical base in this paper,
mentioned empirical base, we engaged in an inductive study in which while it thereby, at least indirectly, also extends to a part of the consumer goods industry
we built on our investigation of the state-of-the-art AM technologies at large.
 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239 227

Table 1 volume—without having to invest in expensive tools, such as molds

Leading additive manufacturing technologies for consumer goods manufacturing. (cf. the long tail). In the particular example of the focal consumer
AM technology Description goods manufacturer, mold costs can amount to more than 500,000
Fused Filament A process in which a filament (a string of plastic) is fed
Euros, depending on the complexity of the part and the tools. Ac-
Fabrication (FFF) to a heating element (the nozzle) and is then cordingly, implementing AM can reduce the sunk costs and general
semi-molten in order to be able to dispense the plastic investments in equipment (Hopkinson et al., 2006). Besides, the re-
as desired. duction and elimination of traditional tooling, such as molds, AM
Selective Laser Sintering A process during which thin layers of polymer are
will potentially increase a company's product introduction rate. Since
(SLS) being distributed in the building chamber and are
bound with the use of a laser deflection system mold and tool development and testing time can be reduced (due to
(mirrors deflect the laser beam to a desired location). free forming in AM machines), companies will be able to generate
Stereolithography A process of solidifying liquid resin (usually either more rapid test concepts and release them to the market (cf. Bogers
Apparatus (SLA) acrylic or epoxy) by photo-polymerization using a and Horst, 2014; Mortara et al., 2009; Wong and Hernandez, 2012). Fur-
(also: SL) controlled laser beam.
Direct Light Processing A process, very similar to SLA, that is also based on a
thermore, adjustments to geometries, concepts, new parts for existing
(DLP) photo-polymerization process; however instead of products, etc. will be possible after a product has been launched and re-
using a laser beam, a light projector is projecting leased to the market. This will enable faster and better reactions to
images onto the polymer surface. market and trend changes (Hopkinson et al., 2006).
Polyjet Matrix (PJM) A process that uses the familiar concept of inkjet
(also: Inkjet 3D printing, as used by 2D printing companies worldwide.
printing) Instead of using ink, the print head ejects a 4.3. A business model perspective on additive manufacturing: from a
photo-polymer material that is then hardened by a UV manufacturer-centric to a consumer-centric value logic
light that is connected to the print head on both its
sides. The above characteristics of AM directly affect how a manufacturer
Inkjet ZCorporation A powder-binding process based on a liquid binder
(the focal company in this case) in the consumer goods industry can
technology (Zcorp) (often colored) that is dispensed on to a bed of plaster
powder, layer by layer, using an inkjet multi-nozzle “do business” by offering value to its customer while also capturing
printing mechanism. part of that value (Afuah, 2014; Massa and Tucci, 2014; Zott et al.,
Note: Assessment based on focal manufacturer, active in the plastic component industry. 2011). In this paper, we refer to a business model as a system of interde-
pendent activities within and across the organizational boundaries that
enables the organization and its partners to create value and capture
4.2. Additive manufacturing technologies' potential sources of value part of that value (Zott and Amit, 2010). For example, the fact that
free forming and personalization are now possible to a larger extent im-
As seen from Table 2 and Appendix B, none of the technologies ob- plies a possible shift of value-adding activities from the manufacturer to
tain a maximum score (i.e. optimal performance for the focal consumer the consumer. Besides, advances in Internet capability (the so-called
goods manufacturer). However, several technologies show more prom- “Web 2.0”) make it even more possible to establish a direct contact
ise than others due to the inherent characteristics of their technologies. between manufacturers and consumers (Wirtz et al., 2010). A major
In this analysis, it is clear that FFF is the leading technology with its po- implication of AM technologies is therefore the potential to shift
tential to be implemented as a manufacturing concept. However, ad- productive activities from the manufacturer to the consumer.
vances in other technologies, for example biocompatible materials in Generally, considering business models for AM is important given
PJM and color technology for SLS, could change this assessment in that technology itself has no real value to a company if not commercial-
the future and, as such, alter the relative ranking of any of these ized in a profitable manner (Chesbrough and Rosenbloom, 2002). While
technologies.4 Although they show exceptional part accuracy, the char- new technologies can sometimes be implemented with existing busi-
acteristics of SLA and DLP's materials make these technologies as less fa- ness models, AM technologies are disruptive to the extent that they
vorable for consumer goods manufacturing. Zcorp was also deemed may require some reshaping or reinvention of the business model in
unfit for manufacturing and is considered a prototyping technology, at order to capture its value (Chesbrough, 2010; Johnson et al., 2008). In
least based on the current assessment. particular, a major shift involves the transition from a manufacturer-
Based on our assessment, the more general advantage of AM as a centric to a consumer-centric value logic in the business model, which
manufacturing concept lies in a number of factors. First, AM is a mass- implies that consumers can be more directly involved in productive
customization and personalization-enabling technology (Hague et al., and value-adding activities. In the context of consumer goods
2003; Gibson et al., 2010). It is a free forming process that enables the manufacturing, this has a number of important implications.
manufacturing of almost unlimited geometries at once without special Here, we build on Amit and Zott's (2001) framework of efficiency,
tools and equipment. Furthermore, AM makes it possible to manufac- complementarities, lock-in, and novelty to determine the value creation
ture various parts in the same batch (Gibson et al., 2010). This presents logic. Table 3 presents a comparison of a traditional manufacturer-
some implications for the potential of “mass customization” (Reeves centric logic with a new consumer-centric logic within this business
et al., 2011; Tseng and Piller, 2003), while also supporting the decentral- model framework. In the table, efficiency refers to cost reductions of-
ization and localization of production (Walter et al., 2004; Hadar and fered by the manufacturing platform. Ease of accessibility between
Bilberg, 2012). Due to their high flexibility, AM platforms can free sellers and buyers, transparency throughout the supply chain, leverag-
form shapes without the use of conventional tooling and machining. ing opportunistic behavior, and accelerating the purchasing process
Since the same machine can produce almost any given part in any via direct connectivity, all contribute to a potential high efficiency.
given location, it allows companies to implement it in any of their loca- Moreover, complementarities refer to the ability to offer products and
tions, whether a production facility or even an inventory or a distribu- bundles in such a matter that offering them together increases their
tion center (Geraedts et al., 2012; Hopkinson et al., 2006; Walter et al., perceived value for the consumer. This can be achieved by offering
2004). after-sale services or complementing products, offline services, offering
Moreover, AM will allow companies to produce “low scale parts”—the products that are not directly related to the main product, and products
parts in the company's product portfolio that are manufactured with low from other providers. Furthermore, lock-in refers to the ability to create
loyalty by consumers in order to establish repeating transactions. It
4
On this basis, the focal consumer goods manufacturer has decided to not prefer a par-
aims at rewarding customers for reoccurring purchases, for example
ticular technology but rather develop on leading technologies (FFF, SLS, and PJM) through loyalty programs, a strong and favorable design, and customi-
simultaneously. zation. Finally, novelty refers to the ability to renew product offerings,
228 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239

Table 2
Benchmarking additive manufacturing technologies for consumer good manufacturer.

FFF SLS SLA DLP PJM Zcorp

Mechanical properties The strength, dimensional accuracy, elongation, fatigue, etc. of the printed part 3 3 2 2 3 1
Chemical properties Toxicity, viscosity, etc., of the printed part 5 5 1 1 2 1
Visual finish The surface quality of the printed part 2 3 5 5 4 2
Cost The manufacturing cost of the part, including material, production, post-processes, manual processes 5 2 1 1 1 1
Time The time it takes to manufacture a part, including preparation, production, post-processes 1 1 1 1 1 1
Volume The possible volume a machine can output per printing job considering build speed, chamber size, etc. 2 2 2 2 2 2
Multicolor The ability of the platform to manufacture a part in a single color without post-processing 4 2 1 1 4 4
Decoration The ability of the platform to create color patterns on the part while manufacturing 1 1 1 1 4 4
Sum of scores 23 19 14 14 21 16

Note: 5 represents high (optimal) results, 1 represents low (critical) results.

which goes beyond the mere development of products, technologies, and the consumer's activities—much in line with Koren's (2010) discus-
production methods, etc. In this case, a first mover strategy can be ben- sion of the changing role of the customer when moving from mass pro-
eficial as it can contribute furthermore to consumer lock-in. duction to mass customization to personalized production (Fig. 1). With
According to the analysis, a central element in customer-centric respect to the supporting supply chain, this creates the potential and
business models for AM involves a direct co-creation with users, even necessitates a move toward a more decentralized supply chain
which creates value externally for the consumer. In an extreme case, structure, even though AM could in general be integrated with central-
the creative part can even be done purely by the consumer, or con- ized supply chains as well.
sumers may form strong communities around the design and service When considering supply chain structures for AM, Walter et al.
based on the AM technology. Lock-in will mainly revolve around the (2004) describe centralized and decentralized as two possible models.
specific experience, which can still only be offered by the manufacturer They describe the effect utilizing AM technologies would have on the
that owns the platform. Another source of value is the potential differ- supply chain, especially considering the long tail in relation to the pro-
entiation that the manufacturer can offer to create value for the con- duction of low scale parts (based on Christopher, 1992). Using AM will
sumer. AM will also enable the production of specialized parts for reduce the inventory levels necessary in distribution centers while
special edition series and customized products. There is still potential keeping a centralized AM production facility. This will allow companies
for the manufacturer to offer complementing services, for example by to increase the delayed differentiation and reduce general inventory
using partnerships with 3D generating software providers, AM platform levels while allowing companies to better fit supply to real demand.
manufacturers, other accessory providers, or by directly selling services Centralized inventory with centralized AM production for low scale
related to the AM process, design rules for the products, material knowl- parts will decrease inventory while keeping investments low. This
edge, etc. means that the company may have the more standardized products
on the inventory shelves with high turnover.
This is then supplemented with specialized customer-specific prod-
5. Operationalizing a consumer-centric business model: implications
for supply chains ucts. In this way, the inventoried components can be produced based on
lean principles and kanban signals to replenish the more standardized
Moving from a manufacturer-centric to a consumer-centric value components and the customized products are manufactured just in
time based on the AM technologies. This setup is reducing complexity
logic—in the sense that the business model implies more productive
and thereby value-adding activities by the consumer—does not only in the supply chain (balancing capacity and inventory), decreases the
have strategic relevance but it also holds important implications for response time, and changes the business model as such, as in line with
the operational activities of the manufacturer. the mass customization approach (Da Silveira et al., 2001; Pine, 1999;
Tseng and Piller, 2003). One of the more interesting debates in recent
years concerning supply chain strategy is based on “lean” and “agile”
5.1. From centralized to decentralized supply chains philosophies (cf. Tuck et al., 2007). The focus of lean thinking has been
on the reduction or elimination of waste (muda). It has been suggested
The characteristics of a consumer-centric business model based on
AM technologies affect the decoupling point between the manufacturer's

Table 3
Comparison of manufacturer-centric and consumer-centric business models.

Manufacturer-centric Consumer-centric

Efficiency • Process transparency • Low inventory cost

• Economies of scale • Print on demand
• Quality monitoring • Low operating cost
• Model reuse
Complementarities • Portfolio-centric • Indirect linkage to portfolio
product development and product designers
• Designer creativity • Multi-partner platforms
Lock-in • Direct relation to • Support in creation and
product portfolio printing
• Company-centric • Availability of platforms
community and sharing • Community-driven sharing
Novelty • Freedom for designers • Co-creation central to design
• Unique design for • Personalized designs
special editions • Localized markets
• Co-creation optional • Optional subscription
Fig. 1. From a manufacturer-centric to a consumer-centric value logic.
Note: Analysis primarily based on assessment at focal consumer goods manufacturer. Note: Adapted and adjusted from Koren (2010).
 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239 229

that lean concepts work well where demand is relatively stable and and the local to achieve optimization of companies' business activities
hence predictable and where variety is low (Christopher et al., 2006). (Svensson, 2001).
Conversely, in those contexts where demand is volatile and the cus- The local aspect of the glocalized concept calls for a self-sufficient
tomer requirement for variety is high, a different approach is called for. supply chain in such a way that the factory is providing materials,
This is the concept of “agility”, which is concerned with responsiveness. parts, pre-assembled elements, etc. from local suppliers and supply
It is about the ability to match supply and demand in turbulent and un- only local customers. The global aspect comes into effect when this
predictable markets. In essence, it is about being demand-driven rather structure is implemented in several regions around the globe. Suppliers
than forecast-driven, and some of these approaches may fit the AM and partners must also operate on a glocalized structure in order to fully
business model. In some cases, the two ideas of lean and agile can be complement it and make the entire supply chain responsive and flexi-
brought together as a hybrid “leagile” solution (cf. Tuck and Hague, ble. Separating the regions of the supply chain and creating alliances
2006). One such hybrid solution is to utilize lean principles when de- with local partners and customers may be part of the new business
signing supply chains for predictable standard products and agile prin- model. This global production model, with production bases around
ciples for unpredictable or “special” products. Or again it may be that the world and more diverse supply chains and distribution, is increas-
total demand for a product can be separated as “base” and “surge” ingly developed and is becoming particularly relevant for AM business
demand. Base demand is more predictable and less risky so that lean models. Companies need to become more responsive, and reduce trans-
principles can be applied, while using agile approaches to cope with portation and inventory. This creates challenges in transferring tacit
surge demand. knowledge between production sites, securing inventories and quality
At the same time, the free forming nature of AM and the lack of need management, and the growing risk of technology being leaked (Kang,
for tooling enables decentralized production close to the market. Com- 2010). Accordingly, complex strategic decisions have to be made that
panies can produce decentralized at production facilities close to their integrate various relations of different levels (Jaehne et al., 2009).
customers with a relatively low investment due to the reduction in ded-
icated tools and the flexibility of AM platforms. This will further reduce 5.3. Implementing a fully consumer-centric value logic? Toward a
inventory and decrease dependency on forecasting. Furthermore, it will personalized model
decrease lead times in a company's production while increasing respon-
siveness for market changes. To exemplify this, Walter et al. (2004) The characteristics of AM technologies, as presented earlier, give rise
describe a case from the aerospace industry, where original equipment to the opportunity to implement a consumer-centric business model
manufacturers (OEMs) in the industry produce high volume parts through a decentralized supply chain. While a decentralized production
in their facilities in a centralized manner—whether using AM or tradi- system allows for localization and accessibility of a manufacturer's offer,
tional manufacturing methods—while producing low scale parts at e.g. through local brand stores, a next step implies an even further
decentralized facilities close to their customers. By that, the OEMs movement of the manufacturer–consumer decoupling point, where
were able to reduce inventory costs, increase their responsiveness, the consumer takes over productive activities of the manufacturer.
lower their investments, and increase their ability to customize The notion of a consumer acting as a producer goes back to at least
products to customers' needs (Walter et al., 2004). Toffler's (1980) suggestions that so-called “prosumers” become more
In the decentralized production system, the business plan will re- important in the face of the digital and information age, the rise of envi-
volve around localization and accessibility. Localization refers to the ronmentalism, interactive interfaces, etc. Recently, this prosumerism
fact that a manufacturer will be able to utilize local brand stores (or often refers to the use of media, where consumers are also producers
other facilities) for the production of low volume parts and by that re- of media in which very personalized content replaces the centralized
duce the inventory needed and its dependency on forecasting. By that, structure—supported by the rise of social media and accessibility of
it will be able to increase its responsiveness to market needs and chang- the Internet (Rennie, 2007).
es. This mainly creates internal value as its benefits are targeted toward AM technologies have particular potential for customer-centric and
its operations. Additionally, localization can also be implemented in a personalized production systems as they allow consumer to produce
localized product focus, as the manufacturer will be able to tailor prod- parts, products, and machines, as users of dedicated AM technologies.
ucts to their markets. Accessibility refers to the manufacturer offering In such a model, end users of AM can use the technology to produce a
consumers the platforms to print their parts if they do not own one, in variety of parts for themselves and others, which then implies a funda-
addition to offering the knowledge to create a model if they do not pos- mental change to the global structure of manufacturing, amplifying the
sess the knowledge to do so. Using this business model, with a shifting change from centralized to decentralized supply chains. The evidence
decoupling point that puts more emphasis on the consumer, the manu- indeed shows an exponential growth in the development of the person-
facturer can create a lock-in as many of its consumers would like to use al desktop printers between 2007 and 2012 (Wohlers, 2013). Such
the service and rely on the manufacturer to do so. Besides internal value, printers are no longer in the possession of industrial companies and
this creates value externally for users and consumers. individual innovators but rather infiltrating early adopters as well.
Given that the prices of a “personal” 3D printer will continue to drop
significantly in the coming years, there is an enormous potential that
5.2. Decentralized supply chains for additive manufacturing: a “glocalized” an increasing amount of consumers will own a personal 3D printer as
approach the technology percolates into the mass market (Wohlers, 2014).
Fully customer-centric production systems emphasize the personal-
Decentralized supply chains have been introduced and researched ization of manufacturing. Manufacturers can increase the individualiza-
as a competitive model since the early 1990s (Hadar and Bilberg, tion of a specific consumer by offering consumers who own a 3D printer
2012). While decentralized supply chains are in general more flexible the possibility to print their own parts. An online platform using a ser-
and more likely to withstand turbulences, a supply chain model that vice design could satisfy the supporting business model. While home
presents a further step by stressing the local aspect of supply chains is printing would imply a full implementation of a decoupled supply
the “glocalized” approach. This model advocates a local, independent chain, manufacturers could also provide printing solutions for con-
supply chain structure on a global scale, i.e. several locally-minded, sumers who create their own designs, which may be shared with others,
self-sufficient supply chains around the globe. The term “glocalization” creating the potential to create a vibrant community (cf. Jeppesen and
combines globalization and localization to create a hybrid that will Frederiksen, 2006; Lakhani and von Hippel, 2003). Payment methods
provide the best of both worlds (Robertson, 1994). The glocalization ap- could be charged per print, increasing the profit of a CAD file since it
proach emphasizes the coexistence and interpenetration of the global can be used by many users, or as a subscription, which will create a
230 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239

high lock-in. This can also be used to boost manufacturers' current prod- how companies can create and capture value from these technologies
ucts as designs offered by the manufacturer can be used in the portfolio. (Chesbrough and Rosenbloom, 2002; Zott et al., 2011). For incumbents
In this process, the manufacturer can in principle still keep the original in the industry, adopting AM as a manufacturing concept is particularly
designs (or files) proprietary given it can communicate with the equip- challenging given the inherent barriers to business model innovation,
ment of the consumer who will not need the actual file but rather only such as conflicts with the existing business model and the cognitive
the G-Code. frames that prevent managers from understanding these barriers
Consolidating the above shows value creation—both externally and (Chesbrough, 2010). Given that there are different AM technologies
internally—for the consumer and for a manufacturer's operations. Inter- available in the market and that traditional manufacturing technologies
nal value creation manifests itself in supply chain and operations as AM are still widely used, the case moreover reveals key considerations
provides benefits to solve challenges and create opportunities for mass when exploring or experimenting with new business models
customization, in particular in consumer-centric models with an em- (cf. Bogers et al., 2015; Brunswicker et al., 2013; McGrath, 2010).
phasis on personalization (Tseng and Piller, 2003). It supports the local- As a part of the digital revolution, AM presents opportunities for new
ized production concept and decentralization of supply chain while still and innovative business models as well as the utilization of more tradi-
providing value to the manufacturer. External value creation manifests tional business models. For example, digital manufacturing enables di-
itself in differentiation and personalization for the manufacturer's prod- rect input from consumers and hence enables customization and
ucts and parts. This enables direct manufacturing, taking advantage of personalization of products. Advances in Internet capability (Web 2.0)
the Web 2.0 and online interfaces to enable co-creation with users, make more direct contact between companies and consumers possible,
which is also an important building block of mass customization and which in turn will also affect the manufacturer–consumer decoupling
personalization. point—moving more upstream than in traditional models (cf. Wirtz
et al., 2010). Accordingly, there is a shift from manufacturer-centric to
6. Concluding discussion consumer-centric business models, which are characterized by a direct
co-creation with users (O'Hern and Rindfleisch, 2009; Prahalad and
This paper presented the state of the art of AM technologies, and it Ramaswamy, 2003; Sawhney et al., 2005). In an extreme case, the crea-
explored how these technologies may influence the viable business tive part can even be done purely by the consumer, or consumers may
models within the consumer goods manufacturing industry—all with form strong communities around the design and service based on the
the aim to answer the question how emerging AM technologies impact AM technology. Lock-in can be established through the specific experi-
business model development and operations in consumer goods ence, which can still only be offered by the manufacturer owning the
manufacturing. The empirical base of this research included the general platform. Another source of value is the potential differentiation and
developments in AM technologies, business models, and supply chains, the production of specialized parts for special edition series and custom-
while we also specifically relied on the recent explorations at a leading ized products.
consumer goods manufacturer within the plastic component industry.
In this context, the manufacturer has been particularly interested in 6.2. Designing and implementing old versus new business models: a
the utilization of additive technologies for the production of customer- complementarity perspective
specific consumer goods on a large scale. This development is fueled
by the fact that, in recent years, AM technologies have emerged as a vi- From a product portfolio point of view, the new AM business models
able manufacturing technology due to significant improvements in part will not replace traditional business models, but should be seen as com-
quality, price and manufacturing speed. Below, we will summarize our plementary and thus possibly co-existing (Benson-Rea et al., 2013). The
main findings and further discuss these findings to highlight possible mainstream products may still follow the traditional business models,
implications and future research directions. but the personalized consumer-designed products may be on top of
that. The new business model may offer a competitive advantage, and
6.1. Emerging additive manufacturing technologies and new business it could complement and thereby even augment the value that is cap-
models: toward a consumer-centric value logic tured through the traditional business model. Nevertheless, the poten-
tial to disrupt or cannibalize the traditional technologies and business
In this paper, we identified six leading AM technologies based on models will also need to be considered (Christensen, 1997; Tushman
the technologies' capabilities, materials, processes, etc. The six and Anderson, 1986). This is particularly salient for incumbent manu-
technologies—FFF, SLS, SLA, DLP, PJM, and Zcorp—were analyzed along facturers, given that their capabilities are rendered obsolete in the face
with a number of relevant parameters to assess their potential (with a of competence-destroying innovations (Tushman and Anderson,
focus on the focal manufacturer's requirements and specifications). In 1986; Henderson and Clark, 1990; Hill and Rothaermel, 2003). In fact,
our assessment, based on the existing and known technologies for con- the same capabilities that can be critical to a company's old business
sumer goods, FFF is the leading technology considering its potential as a model may become rigidities that handicap performance in the face
manufacturing concept—mainly due to the strict nature of the particular of the new business model (Chesbrough, 2010; Leonard-Barton,
industry's chemical standards and the technology's ability to print in 1992).
color, which is a prerequisite in the consumer goods industry. However, Moreover, AM technologies do not only support a move to a more
advances in other technologies, for example biocompatible materials in consumer-centric business model but they can also support the existing
PJM and color technology for SLS, could potentially change this position business models as they may replace some modules within the entire
and make any of them the preferred technology over time. While this manufacturing system or architecture. While this modularity as such
assessment is expected to show the general direction of any assessment can hold important implications for future developments of technolo-
in consumer goods manufacturing (where the same characteristics are gies and business models (cf. Henderson and Clark, 1990; Kodama,
deemed important), future studies should determine to what ex- 2004), a concrete consideration is how different AM technologies fit to-
tent these findings are generalizable. Practically, this framework gether, as more or less connected modules, in the overall strategic
could be used as an exercise for manufacturers that are interested framework of the company (cf. Teece, 2010). For example, AM can be
in adapting AM as a manufacturing capability, bearing in mind the used for rapid prototyping and for production or replacement of tools,
possible subjectivity of the analyses given the context in which molds, or other equipment. Fig. 2 proposes a possible framework that
they are analyzed. highlights the possible solutions or shifts within manufacturer-centric
These developments in AM technology are expected to change the and consumer-centric business models as supported by the emerging
future consumer goods business models, which describe the logic of AM technologies.
 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239 231

From a supply chain point of view, decentralized production can

manifest itself as manufacturing with AM technologies in several
brand stores around the world. In this structure, production of low
scale parts can take place close to the consumers to enable production
to real demand. In doing so, the manufacturer will be able to reduce
its inventory levels in the distribution centers close to the markets and
increase its responsiveness simultaneously. Furthermore, the manufac-
turer can offer a user experience around AM for its consumers, especial-
ly for the ones who are uneducated in 3D modeling and printing. This
would also enable a “glocalized” supply chain in which a company local-
izes its global production system, effectively separating global supply
chain modules into disconnected parts, which each rely on local input
and output. The AM technology in combination with scarcity of re-
sources and volatility in commodities and raw materials are some of
the reasons for consumer good manufacturers to re-examine their busi-
ness models and global footprints (Hadar and Bilberg, 2012). Further re-
search is needed to better understand and respond to the challenges of
globalization locally (Hesse and Rodrigue, 2006), and especially in di-
Fig. 2. A framework for AM-based solutions in business models and supply chains. rect relation to the supply chain and production—the old norms of
“local for local” manufacturing in favor for the new “global village”
(Christopher et al., 2006).
How the business models will be implemented, in detail, will have to Taking a further step in the decentralized consumer-centric produc-
be further researched. While there are many unknown challenges that tion system, the manufacturer can provide a platform for home printing
will emerge in the future, a particular issue that has not been dealt (sometimes referred to as the “prosumer” approach). In this case, the
with so far is the legal liability in the context of increased consumer in- manufacturer can offer truly personalized parts for home printing, in
volvement and responsibility for home manufacturing (cf. Bradshaw addition to low volume parts, which will not be financially justified
et al., 2010). More generally, however, the literature of innovation with traditional production methods. The manufacturer will have
toolkits has shown a great potential benefit when involving users in to create an online platform for users to create, share, or download
such innovation and customization processes (Bogers et al., 2010; geometrical models and files for printing. This will naturally open
Franke and Piller, 2003; von Hippel, 2005), even though the profitability new possibilities for the manufacturer to create interactive plat-
and potential remains to be determined (cf. Piller et al., 2004; Prahalad forms in addition to “outsourcing” the production of specialized
and Ramaswamy, 2003). Nevertheless, developing the appropriate parts to the consumers.
toolkits and platforms will be required to effectively leverage the latent
potential of integrating various stakeholders in the innovation process 6.4. Business model and supply chain challenges of additive manufacturing
(Eisenmann et al., 2006; Frankenberger et al., 2013; Gawer and
Cusumano, 2014; West and Bogers, 2014). While one of the advantages of AM is the ability to co-create with
consumers in order to increase the customization of unique parts, the
6.3. From manufacturer-centric to consumer-centric business models: question here is how the manufacturer can manage and control that
implications for supply chains process. Giving consumers the ability to contribute to the design will in-
crease individualization but at the same time demand the company to
Moving toward more consumer-centric business models can be op- approve every unique design for safety, mechanical behavior, etc.
erationalized through a decentralized production system in which the Additionally, if the company would like to keep control of the designs
decoupling point within the supply chain is changing due to a larger in- that the consumers contributed to, it will tremendously increase the
volvement of the end consumer. Fig. 2 includes the supply chain dimen- number of parts and thereby the supply chain complexity. In a purely
sion in a proposed framework for AM-based manufacturing solutions. decentralized setting, it is virtually impossible to handle such large var-
While centralized production refers to printing at the different manu- iance. Moreover, if consumers print at their own home—based on the
facturer facilities around the globe, a decentralized system breaks this many platforms for 3D printing that are or will be available—how will
system up into more localized facilities. In a centralized model, the man- companies make sure the printing parameters fit the design specifica-
ufacturer can still utilize AM platforms for production of specialized tions? Consumers will be able to print with many materials and with
parts for low production series and special editions. It can also increase platforms that have different specifications. That would mean that
the freedom of design for designers and enable them to incorporate the final product will be different depending on the platform,
more unique parts to the products. Moreover, the manufacturer can which will make it very hard to guarantee a part behavior or proper-
offer online solutions to enable co-creation with consumers—effectively ties. A question of legal liability thus also arises when parts are not
moving toward a customer-centric model. In this model, the manufac- being centrally produced and approved in the official manufacturer's
turer can create the platform for users to customize or personalize processes.
their own parts and ship them on demand. The changes in the production decoupling point between the manu-
This model has the additional benefit that the manufacturer can also facturer and the consumer moreover have legal implications depending
use the models created by users as inspiration for novel product designs on whether or not the manufacturer keeps control of the original de-
and as a source of market data gathering. More generally, if the signs (or files) or the underlying source code. At the same time, a possi-
decentralized supply chain connects the manufacturer and consumer ble consideration is the parallel industry development in 3D scanning,
through an accessible platform, this direct contact can also be used to which refers to the ability to transform physical objects to digital files.
collect valuable consumer data—thus linking back to the consumer- This could further enable copying and sharing of digital designs and
centric business model. This could also possibly be used to monitor files over the Internet. Drawing a parallel to the music and movie indus-
trends in the industry, ranging from design to purchasing behavior, tries, the digitization of music and movie files by consumers involved
and it can also be used to generate and establish a co-creation platform copying and (freely) sharing these files, while it took the industry a
(Sawhney et al., 2005; von Hippel, 2005). long time to develop viable business models in this domain (e.g. Spotify
232 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239

and Netflix). All-in-all, future developments within a decentralized future research, while it also holds important implications for practice.
supply chain that supports AM-based business models will need to On the one hand, it needs to be better understood how the general de-
consider the intellectual property (IP) implications (cf. Bradshaw velopments in technology affect existing and future business models,
et al., 2010). Possible implications are a change in IP and licensing and vice versa, while we identify the specific role of the supply chain
strategies (Bogers et al., 2012) or in IP sharing norms more general- and organization in general (e.g. structure and capabilities) as being a
ly (Fauchart and von Hippel, 2008; von Hippel and von Krogh, particularly important factors in this process. Besides, the specific devel-
2006). opments in the consumer goods industry will need to be further scruti-
Yet another parallel process that may challenge the opportunities nized in order to better understand the particular contingencies in this
that a manufacturer has is the design and production of 3D printers. industry. Particular attention will need to be paid to the implementation
This development will affect the AM-based business models and supply of new technologies, business models and organizations, given that this
chains given that the availability and offering of 3D printers will become will provide a specific set of challenges that go beyond the development
part of the value creation process and related cost structures within a phase as considered in this paper.
consumer-centric value logic. Besides the actual functionality and deliv- On the other hand, this paper gives some insights into the state of
ery, this for example refers to the design of 3D printers, which is cur- the art of the developments within AM technologies and promising
rently still inspired by their rapid prototyping functionality (and look). business models with supporting supply chain, which may be of rele-
Machines are generally built for a lot of manual handling and are not op- vance to other practitioners within consumer goods and beyond. As
timized for a production layout. The printing process relies on many such, the paper implies certain attention points for managers who
post-processes, which adds complexity to the supply chain and inaccu- need to deal with new developments in technologies, business models
racy to the process. and supply chains—within AM and beyond—such as a set of key evalu-
ation criteria, key considerations for business model and related supply
6.5. Complementarities between technology, business models and chain development based on new technologies, and the importance of
organization considering operational and organizational issues in the face of techno-
logical and business model development.
Our investigation of the implication of AM technologies on business Finally, some key practical takeaways include the presentation and
models and supply chains in the consumer goods industry has provided benchmarking of the leading AM technologies (Tables 1 and 2), which
the state of the art of available technologies in this particular domain. give a concise overview of the state of the art in terms of production
This study showed how such a technology concept is fundamentally technologies for the consumer goods industry. Managers could use
intertwined with the business model concept. As in line with Baden- these insights as reference points for their own explorations within
Fuller and Haefliger (2013), this co-evolution implies that the techno- AM, while also offering the basis for considering more consumer-
logical developments directly relate to certain business model decisions. centric business models in which they will need to manage the chang-
As they also highlight, these decisions involve openness and user en- ing decoupling point in the consumer goods' manufacturing supply
gagement, which in this case implies a consideration of how the busi- chain. Accordingly, business model development in the face of emerging
ness model needs to align with the relevant knowledge sources within AM technologies implies managing organization change and more
or outside the organization to create and capture the most possible openness toward external sources. A balanced view will be needed
value (cf. Chesbrough, 2003; Chesbrough and Rosenbloom, 2002). This though as an increasing consumer role can have major implications
directly contributes to our understanding of open innovation, which is for technologies, business models, supply chains, intellectual property,
a distributed innovation process based on purposefully managed and so on, while also the threats, costs and downsides will need to be
knowledge flows across organizational boundaries in line with the managed. Moreover, given the uncertainties surrounding AM technolo-
organization's business model (Chesbrough, 2006; Chesbrough and gies and business models, companies will need to explore or experi-
Bogers, 2014; Vanhaverbeke and Chesbrough, 2014). Specifically, we ment with new developments, including organizations and supply
provide a framework for understanding how the co-evolution of tech- chains (Baden-Fuller and Haefliger, 2013; Bogers et al., 2015;
nology and business models links to the organization of productive ac- Brunswicker et al., 2013; Koren, 2010; McGrath, 2010). Managing all
tivities, thus directly extending this general understanding to the these processes in parallel could create certain paradoxes, which will
context of (innovative) supply chains. What become clear in this con- need to be balanced with the right leadership and strategies
text is that there are strong complementarities in this innovation pro- (Benson-Rea et al., 2013; Bouncken et al., 2015; Chesbrough, 2010;
cess, not only across these concepts but also within, such as the Laursen and Salter, 2014; Smith et al., 2010). And given the holistic na-
interdependence between product and process development (cf. Kraft, ture of the framework and solutions that can or may be derived from
1990; Reichstein et al., 2008). AM technologies and business models, it will advantageous to consider
This interdependent development of technology and business the possible benefits for or effects on other domains and functional
models further highlights the role of supply chain structures to create areas, beyond R&D and manufacturing, such as marketing or entrepre-
and capture value. More generally, the case reveals that, whichever neurship (Chesbrough and Bogers, 2014; Coombes and Nicholson,
combination of AM technology and business model emerges, there are 2013; Ehret et al., 2013; George and Bock, 2011).
important implications for how the company organizes its activities.
The choice between a manufacturer-centric and consumer-centric busi-
Acknowledgment
ness model has a significant impact on the organizational structures and
capabilities that would support the operationalization of the (new)
We would like to thank the reviewers and editors for their construc-
business models. Therefore, considering a business model as a system
tive comments and guidance during the review process.
of interdependent activities (Zott and Amit, 2010), there is an important
connotation to finding new ways to “organize” these activities
(Casadesus-Masanell and Zhu, 2013; Foss and Saebi, 2014). It is there- Appendix A. State of the art of additive manufacturing technologies
fore important to consider the co-evolution of technology and business
models in relation to the organizational structures that enable the In order to understand the different additive manufacturing (AM)
organization to enact the commercial opportunity they identity platforms and technologies, we present a detailed state of the art anal-
(George and Bock, 2011). ysis of AM technologies. The analysis explores the different AM technol-
This co-evolution between technology, business model and supply ogies' advantages and disadvantages, and their feasibility for consumer
chains (or organization more generally) is an important avenue for goods production with a particular focus on the focal manufacturer that
 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239 233

serves as an example for the more general developments within the materials to be developed. An example for that is the Ultem 9085 mate-
consumer goods industry. rial, which is used mainly for the aerospace industry (Stratasys, 2013a).
The main supplier of industrial FFF machines is Stratasys LTD. which
A.1. Exploring additive manufacturing technologies is based in Rehovot, Israel and Eden Prairie, Minnesota, USA. The indus-
trial machine series is referred to as the Fortus series and consists of 4
Before presenting an overview of the state of the art, we note that machines: the Fortus 250mc, 360mc, 400mc, and 900mc, and they differ
the technologies explored in this paper are polymer AM production by the size of the chamber, tip sizes for production, production speed,
technologies alone. Several technologies use metal as well for rapid and other variants. The smallest layer thickness available in FFF is
tool manufacturing and general metal production, such as Selective 0.127 mm (0.005 in.) and is available in the Fortus 250mc, and 360mc.
Laser Melting (SLM) or Metal Selective Laser Sintering (MSLS). Other The tip size chosen also has a great influence on the accuracy of the
technologies use wax and ceramics as base materials. These technolo- part. There are several tip sizes available, such as T16 (0.016 in. in diam-
gies will not be explored in this paper as the focus (in the consumer eter), T12 and T10, correspondingly. Though a smaller tip size and layer
goods manufacturer that serves as the empirical base) is plastic parts thickness will yield in higher accuracy and better surface finish, it will
production for consumer goods. also compromise the mechanical property of the part as less material
In cooperation with the focal manufacturer, leading technologies applied will translate to less binding surface and consequently worse
have been identified considering the technologies' capabilities, mate- mechanical integrity. Changes in tolerances will of course influence
rials, processes, etc. and in light of the manufacturer's requirements the dimensional accuracy of the part. Different systems have different
and specifications. Six leading potential technologies used for rapid dimensional accuracies, depending on their mechanics and input
prototyping and AM have been identified, namely: Fused Filament parameters (Wohlers, 2013, Ahn et al., 2002, Gibson et al., 2010,
Fabrication (FFF), Selective Laser Sintering (SLS), Stereolithography Stratasys, 2013b).
Apparatus (SLA), Direct Light Processing (DLP), Polyjet Matrix (PJM),
and Inkjet ZCorporation technology (Zcorp). The following section will
A.3. Selective Laser Sintering (SLS)
give a brief overview of each technology, its method of operation, and
its main use today.
SLS is a process during which thin layers of polymer are being dis-
tributed in the building chamber and are bound with the use of a laser
A.2. Fused Filament Fabrication (FFF)5 deflection system (mirrors deflect the laser beam to a desired location).
The laser will create a path in the polymer's layer that will consolidate
FFF is a process in which a filament (a string of plastic) is fed to a the layers. After the binding of a layer takes place, a new layer will be
heating element (the nozzle) and is then semi-molten in order to be dispensed and the process will repeat itself until a shape is created
able to dispense the plastic as desired. Since the material dispensed (Kruth et al., 2003; Kruth et al., 2005).
from the nozzle is in a semi-molten state, it binds with the layer beneath SLS machines use different variance of technologies and methods for
it to create a solid object (Fig. A.1). The printing head often moves in the the process. Different platforms use, for example, different power
x and y directions while the platform itself moves in the z, although in dispensing mechanisms (arcing vs. linear), different atmospheres (dif-
some machines the print head is capable of moving in all 3 directions ferent gases used in the chamber while producing have different affects
(Ahn et al., 2002, Gibson et al., 2010). of the processes), and different laser techniques (different mirrors, dif-
In addition to the building material, high-end platforms use a sup- ferent sizes, intensity, etc.). Fig. A.2 shows a visual representation of
port material disposal system as well. This support material is used to the principal concept (Kruth et al., 2005). SLS machines do not use sup-
create geometries that would otherwise collapse in the process of build- port materials as the un-bound powder left in the chamber is used as a
support structure for the coming layers. This restricts to a certain degree
the dimensions and geometries possible with SLS technology (Gibson
et al., 2010).
One of the main suppliers of SLS machines is EOS GmbH with its
headquarters in Munich, Germany. EOS's industrial AM machines are
mainly of the EOSINT series, which, in relation to production, comprise
of the P395, P760, and the P800. Other machines from EOS include the
Formiga series, which is designed for rapid prototyping more than for

Fig. A.1. The FFF (FDM) process (Ahn et al., 2002).

ing. The support material is then extracted (usually dissolved), leaving

only the building material behind. Common materials used in FFF
are mostly ABS and PLA, though some applications require specific

5
Fused Filament Fabrication (FFF) is also referred to as Fused Deposition Modeling
(FDM), which is a trademark of Stratasys. Fig. A.2. The SLS process (Deckers et al., 2012).
234 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239

production. Most SLS machines use different PA powders, though some characteristic of its resin, can achieve down to a micron level accuracy.
applications required reinforcing additives, such as glass, aluminum, The iPro machines operate with a dual laser spot size capability for
and carbon fibers (EOS, 2013). The majority of materials used in SLS faster building rate, with a beam size of 0.13 mm in diameter for borders
are white although some materials are offered in black as well. and 0.76 mm for filling (Melchels et al., 2010; 3D systems, 2013b; Pham
Coloring is only possible with the use of post-processes. and Gault, 1998).
The main difference between the different platforms is the building
chamber's size. Layer thickness in the machines can reach down to be- A.5. Direct Light Processing (DLP)
tween 0.06 and 0.18 mm, depending on the size of the laser and produc-
tion parameters, and the building speed can be as fast as 30 mm/h. DLP is a very similar process to SLA. DLP is also based on a photo-
Going down in layer and beam size will increase the details and the ac- polymerization process; however instead of using a laser beam, a light
curacy of the part but will reduce the speed of the process. The higher projector is projecting images onto the polymer surface. Contrary to
models (P760 and P800) use a twin laser system in order to increase SLA, in which the platforms moved downwards as patterns are
the building speed (EOS, 2013) projected on it, DLP's light projector is located under the platform and
Other than by EOS, the SLS systems is offered by 3D Systems Inc. patterns are projected onto a clear tray that contains the resin, curing
with its headquarters in Rock Hill, SC, USA. 3D Systems' SLS AM series the resin from beneath. The platform holding the part will then move
is the sPro series with a variety of machines from the sPro 60 SD to up and a new layer of resin will be applied on the tray, as can be seen
the sPro 230 HS, which differ mostly by their chamber size. The platform in Fig. A.4. Due to its light projecting technology that project a 2D
can reach a layer thickness between 0.08 and 0.15 mm and a building image and not a single point laser, building time can be significantly re-
speed between 1 and 5 L/h (3D systems, 2013a). duced and numerous parts can be built simultaneously (Melchels et al.,
2010; Gibson et al., 2010).
A.4. Stereolithography Apparatus (SLA) One of the main suppliers of DLP technologies is EnvisionTEC with
its headquarters in Marl, Germany. With its Perfactory series,
SLA (occasionally referred to as SL) is the process of solidifying liquid EnvisionTEC has introduced a wide range of printers used for both
resin (usually either acrylic or epoxy) by photo-polymerization using a rapid prototyping and AM. Interesting platforms for manufacturing
controlled laser beam. The laser creates patterns on the surface of the are the Perfactory Standard and Xtreme product ranges. Print resolution
resin, and to a degree of depth inside the resin, forcing it to solidify in these machines can go down to 0.05 mm with a 25 mm/h building
when exposed to the UV light. The building platform is then pulled rate. Layer thickness can vary between 0.025 to 0.15 mm depending
down, the part is covered with a new layer of liquid resin, and the pro- on the material and the platform used. Materials used in the DLP tech-
cess repeats for the full formation of the part. Post processing in UV cur- nology are essentially the same as are used in SLA — epoxy based pho-
ing oven is necessary after the part is removed from the machine due to topolymers. The materials are mostly transparent or have a yellow/red
the high viscosity of the resin in order to ensure that no liquid or partly color to them (EnvisionTEC, 2013). Color printing is not possible with-
liquefied resin remain on the part. Fig. A.3 shows a representation of the out the use of post-processing.
process (Gibson et al., 2010; Melchels et al., 2010). Materials used in SLA
are either transparent or white and cannot be colored during the A.6. Polyjet Matrix (PJM)
process but with use of a post-process (gray colors are possible on
some platforms). PJM (often referred to as Inkjet 3D printing) is one of the youngest
One of the main suppliers of SLA machines is 3D Systems. 3D Sys- AM technologies currently available on the market. It was commercial-
tems has a wide range of SLA machines used for production from the ized by the former Israeli company Objet, which merged with Stratasys
iPro series, namely: 8000, 8000MP, 9000, 9000XL, which differ mostly in 2013, and is now offered by Stratasys (Stratasys, 2013c). The technol-
on print size capabilities. 3D Systems machines can achieve a layer ogy uses a familiar concept of inkjet printing, as used by 2D printing
thickness of between 0.05 and 0.15 mm and due to the liquid companies worldwide. Instead of using ink, the print head ejects a

Fig. A.3. The SLA process (http://www.custompartnet.com/wu/stereolithography).

 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239 235

to the finished products. Other platforms can print with the full CMYK
color range to a high resolution (Stratasys, 2013d).
In recent years, a growing interest in the biomedical and dental in-
dustry gave rise to interesting developments in materials in SLA, DLP,
and PJM technologies. Biocompatible materials targeting these applica-
tions have been developed in the dental, hearing aid, prosthetics, etc. in-
dustries. These epoxy or acrylic based materials were developed to
withstand harsh health regulations in these industries (Stratasys,
2013d; 3D systems, 2013c).

A.7. Inkjet ZCorporation technology (Zcorp)

The last technology that has been one of the main ones as explored
by the focal manufacturer is the Zcorp powder-binding technology
made commercial by Z Corporation, which was bought by 3D Systems
in 2012. The powder-binding 3D printing technology (hereafter:
Zcorp) was initially developed at Massachusetts Institute of Technology
(Sachs et al., 1993) and was licensed to Z Corporation. The Zcorp tech-
nology is based on a liquid binder (often colored) that is dispensed on
to a bed of plaster powder, layer by layer, using an inkjet multi-nozzle
printing mechanism. The drops applied by the mechanism are colored
as they are applied (by combining basic colors, similar to a regular print-
er) hence the printer is capable of printing the entire CMYK color range.
Fig. A.4. The DLP process (http://3dprinterdlp.com/definition-and-technology/).
Zcorp does not use support materials as it is built very similarly to SLS
while unbound material is used for support, as can be seen in Fig. A.6.
photo-polymer material which is then hardened by a UV light which is The remaining unbound powder is removed manually using a vacuum
connected to the print head on both its sides, as can be seen in Fig. A.5. In system and is mostly reused again (Gibson et al., 2010; 3D systems,
addition to the building material applied by the print head, a wax based 2013d).
support material is applied in order to expand the range of geometries Some of the larger machines used for higher volume production are
achievable by the machine. The support material can be removed the ZPrinter 450, 650, and 850. Layer thickness varies between 0.09 and
using pressurized water (Vaupotič et al., 2006; Brajlih et al., 2006). 0.2 mm. Materials used in the technology are mostly ceramic composite
Although the technology is advertised as specialized for rapid powders and elastomers, although other materials, such as sandstone
prototyping and not for AM, it is one of the fastest and more accurate can be used as well (Gibson et al., 2010; 3D systems, 2013d; Walters
free forming 3D printing technologies in the market. The speed of the et al., 2009). Though it is clear that Zcorp machines are not meant for
process is attributed to two characteristics of the technology. The first
is the multiple ejection of material from the print head and the second
is the dual UV light system in both sides of the print head. Both attri-
butes enable the production of multiple parts simultaneously and a
faster curing process (Melchels et al., 2010).

Fig. A.6. The Zcorp process (Sachs et al., 1993).

production, they nonetheless have some interesting characteristics for

the development of AM.

A.8. Other manufacturers

Although the companies mentioned above are clearly the market

leaders (Wohlers, 2013; Gibson et al., 2010), there are several other
Fig. A.5. The PJM process (Vaupotič et al., 2006).
companies with manufacturing platforms that hold some potential.
For example, the Germany-based Voxeljet has developed a technique
Stratasys has several product ranges it is commercializing: the Eden that is similar to the Zcorp technology. In this process PMMA plastic is
and the Connex. Suitable printers for AM are potentially the Objet Eden bound with a layer thickness of 0.08–0.2 mm and dimensional accuracy
350V and 500V, and the Objet350 and Objet500 Connex. Layer thickness of 0.55 mm. The innovative concept is that the system prints in a 45°
is specified as 0.016 mm with accuracy of about 0.085 mm, depending angle and an input and output on the sides of the machine, which
on geometries and materials (Stratasys, 2013c). Materials vary from enables a potential infinite part construction (Voxeljet, 2013).
ABS and PP like materials to rubber materials. Several platforms are Another potential technology that is still in its infancy is called Selec-
able to combine soft and hard materials and create different hardness tive Heat Sintering, which is commercialized by Denmark-based
236 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239

Blueprinter. The technology is very similar to SLS and uses the same ma- manufacturer and are in fact used in the injection molding process.
terials. The main difference is the use of a heating element instead of a That means that the general material properties and behavior are famil-
laser to bind layer of the powdered material. The achievable layer thick- iar to the manufacturer and are approved for use in consumer goods.
ness is currently 0.1 mm with a building speed of 2–3 mm/h The second reason is that these thermoplastic (ABS and PA) have both
(Blueprinter, 2013). The technology is still under development but already withstood the harsh requirement for chemical approval speci-
could be interesting in the near future. fied by the manufacturer. That will ease the integration and acceptance
Yet another interesting development in the AM industry is when in- of the materials in the organization.
jection molding giant Arburg revealed its Freeformer 3D printer. The All resin-based material scored low in the evaluation due to their
printer utilizes the newly developed technology Arburg Plastic high toxicity. Moreover, the UV curing processes (or light exposure)
Freeforming (AKF, after the German abbreviation). The novelty of the used in SLA, DLP and PJM are not yet stable and cannot guarantee the
technology is the use of the exact same screwing mechanism used in in- required migration standard. In general, UV curing processes are not
jection molding, which melts granulate to its moldable state. The melted
granulate is then applied drop by drop onto a moving platform. This de-
velopment is groundbreaking because it enables manufacturers to use
the same materials they have been using to date in injection molding.
This is not only beneficial for material and mechanical properties of
parts but also for the reduction of part cost as normal granulate is
much cheaper than currently offered AM materials. Moreover, Arburg
combined a 5-axis robot to its moving platform that enables construc-
tion on the part from its side (Arburg, 2013).

Appendix B. Assessment of additive manufacturing technologies in

consumer goods production systems

B.1. Mechanical properties

Table B.1 shows part of the mechanical testing conducted in order to Fig. B.1. Different printing orientations for AM.
assess the mechanical properties of the different technologies and
benchmark those to current traditional injection molding. Tests con-
ducted during these experiments are tensile tests and modulus, impact accepted at the manufacturer due to that reason. That said, there is an
tests, fall, and break tests. FFF parts were tested from two manufac- exception to be made due to recent developments in biocompatible ma-
turers: Stratasys as a high-end industrial printing manufacturer, and terials that have been developed for the biomedical industry, as men-
Makerbot as a low cost desktop printer manufacturer. The purpose tioned in the state of the art section. Chemical tests on these materials
was the compare high-end parts with low-cost parts. Additionally, have not been conducted with the manufacturer's industry standard
two PJM materials were tested to examine the difference of their me- and hence were not included in this examination. It is possible that
chanical properties, and SLS and SLA parts as well. All these results the new materials and processes originally designed for the biomedical
were compared to injection molded parts, also represented in the table. industry can uphold the harsh requirements set by the manufacturer.
The best results were obtained by FFF parts printed on an industrial
machine and SLS parts. A surprising factor was the mechanical tests ob-
tained from PJM parts that presented impressive mechanical properties. B.3. Visual finish
Two noticeable differences are between FFF parts manufactured in
high-end machine and the direction of their production. As can be Visual finish holds two criteria. The first is the roughness of the sur-
seen by the tensile modulus, FFF parts manufactured on the Fortus plat- face in the printed element. The second is the dimensional accuracy of
form presented high results which are comparable to molded parts the part, i.e. the accuracy of the finished part compared with its CAD
while the low cost printer yielded poor results. Also, parts manufactured file and the repeatability of the process. In order to evaluate the visual
on the Y direction (on their side) present significantly better results finish of printed parts in the different technologies, a test using high
than those manufactured on the Z direction (up right), as can be seen precision scanning was conducted. This scan has just several microns
in Fig. B.1. in resolution and can plot a visual representation of the parts accuracy.
The results for FFF (y and x printing directions), SLS and SLA were
B.2. Chemical properties presented and compared with an injection molded original element.
Both FFF parts in both orientations presented good results, with some
FFF and SLS both scored high in the chemical properties evaluation. lacking material. SLS also showed good results, though some surplus
The first reason is that they are both familiar materials to the focal of material was noticeable. SLA showed very good results on shafts

Table B.1
Mechanical testing for AM and molded parts.

Material Tech Layer thickness Color Type/variant Fall test Impact test w. Impact test w.o. Tensile modulus Tensile test/UTS Break Direction
(mm) (cm) notch (kJ/m2) notch (kJ/m2) (N/mm2) (N/mm2) (N/mm2)

2C Epoxy PJM 0.016 VeroGray FullCure850 Objet 9 2.22 28.32 2007.69 63.42 46.27 Y
2C Epoxy PJM 0.016 VeroClear FullCure810 Objet 12 1.48 23.7 2712.79 58.38 50.81 Y
ABS FFF 0.3 Green MakerBot 11 11.78 11.67 1764.75 26.23 25.66 Y
ABS FFF 0.127 Nature Stratasys 14 22.52 2401.2 36.02 57.4 Y
ABS FFF 0.127 Nature Stratasys 14 3.01 3.3 1424.7 8.22 8.22 Z
2C Epoxy SLA 0.05 Accura Xtreme Gray 3DS 16 1.93 18.74 1687.3 45.34 41.45 Y
PA2200 SLS 0.06 White EOS 25 13.98 35.8 1665.67 48.34 42.57 Y
ABS Molding 24.2 2301.32 49.3 32.31
 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239 237

and round parts but presented poor results on plates with a surplus of B.5. Multi-color and decoration
material.
Though SLA's dimensional accuracy might be low in several areas of Multi-color is defined as the possibility of making a part with a single
the element, it presented the lowest surface roughness. SLA and DLP color while decoration is defined as the possibility to create color pat-
showed the lowest surface roughness and measurements were as low terns on the part at the time of production. Table B.3 shows the different
as 10 μm. FFF showed poor results, originating from the clear layer struc- platforms' ability to multi-color and decorate. As can be seen, PJM and
ture on its surface. PJM shows very poor results, though in reality its Zcorp are the only technologies that are capable of producing decorated
accuracy is very high. The reason for these poor results may be the re- parts with patterns. This is due to the fact that both technologies are
maining of support material on the parts. Due to the manual nature of inkjet based and color is applied to individual drops in the processes—to
the process, the repeatability of the process is low and different tests the glue in Zcorp and to the photopolymer in PJM. FFF is capable of
showed different results. multi-color since filaments can be colored in a desired color and fed to
FFF's mechanical and chemical properties attained promising results the system. Color changing is possible in the process but not to a
due to the fact that these technologies use familiar materials to the man- decoration level.
ufacturer and widely used polymers while its visual finish is lacking due It is important to note that SLS, SLA and DLP, offer some variance of
to the relative high layer thickness. SLS shows good results in terms of colors in their materials. Black SLS powder is available, and so are differ-
chemical properties due to the use of the technology in polyamide ma- ent gradients of gray in SLA and yellow/red in DLP. However, these
terials, which are considered quite safe and “consumer-friendly” by the colors cannot be changed without post-processing and hence do not
manufacturer. However, both mechanical properties and visual finish qualify as multi-color or decoration.
show less promising results due to the grainy surface of the printed
parts that resulted from the powder materials the technology is based Table B.3
The technologies ability to produce with color.
on.
All resin based technologies (SLA, DLP, PJM) scored low both on FFF SLS SLA DLP PJM
mechanical and chemical properties due to the brittleness of their mate- Multi-color Yes No No No Yes
rials and its high toxicity, but scored high on visual finish, mainly due to Decoration No No No No Yes
their thin layer thickness and the use of liquid-based resin which allows
the smooth surfaces. Among all technologies, Zcorp has scored the
lowest as a manufacturing technology due to the brittle nature of its References
material, the grainy surface that is a result of its powder based material,
3D systems, 2013a. Retrieved December 2013, from 3D Systems SLS Platforms: http://
and the chemical composite of the binder used in the process.
www.3dsystems.com/sites/www.3dsystems.com/files/sPro-Family-USEN.pdf.
3D systems, 2013b. Retrieved December 2013, from 3D Systems SLA Platform: http://
B.4. Time, cost, volume www.3dsystems.com/sites/www.3dsystems.com/files/ipro-family-a4-uk.pdf.
3D systems, 2013c. Retrieved December 2013, from 3D Systems SLA Healthcare Mate-
rials: http://www.3dsystems.com/solutions/healthcare.
When benchmarking to injection molding machines with just sec- 3D systems, 2013d. Retrieved December 2013, from ZCorp Platform: http://www.zcorp.
onds as a cycle time and significant economies of scale, it is clear that com/en/Products/3D-Printers/spage.aspx.
all AM technologies scored notably low. It is for that reason that their Afuah, A., 2014. Business Model Innovation: Concepts, Analysis and Cases. Routledge,
New York, NY.
score in the evaluation remained low. Nevertheless, there was value in Afuah, A., Tucci, C.L., 2001. Internet Business Models and Strategies: Text and Cases.
benchmarking the technologies to each other. McGraw-Hill, New York, NY.
Evaluating time of production is difficult across all technologies. Ahn, S.-H., Montero, M., Dan, O., Roundy, S., Wright, P.K., 2002. Anisotropic material
properties of fused deposition modeling ABS. Rapid Prototyp. J. 8 (4), 248–257.
Many factors have to be taken into consideration such as geometries, Amit, R., Zott, C., 2001. Value creation in e-business. Strateg. Manag. J. 22 (6–7), 493–520.
materials, platforms, and post-processing. The evaluation presented Arburg, 2013. Arburg Freeformer Retrieved December 2013 http://www.arburg.com/en/
here is rough determination of the time all platforms will take to man- solutions/freeformer/.
Baden-Fuller, C., Haefliger, S., 2013. Business models and technological innovation. Long
ufacture a specified part. The evaluation shows that all technologies
Range Plan. 46 (6), 419–426.
will take more than an hour for the process, while an injection molded Benson-Rea, M., Brodie, R.J., Sima, H., 2013. The plurality of co-existing business models:
part is manufactured in less than 20 s. The results of the analysis are investigating the complexity of value drivers. Ind. Mark. Manag. 42 (5), 717–729.
Berman, B., 2012. 3-D printing: the new industrial revolution. Bus. Horiz. 55 (2), 155–162.
presented in Table B.2.
Blueprinter, 2013. Retrieved December 2013, from Blueprinter SHS Technology: http://
The cost evaluation is made taking into consideration the price of www.blueprinter.dk/shs.html.
manufacturing a part while manufacturing 10,000 pieces. This is done Bogers, M., Horst, W., 2014. Collaborative prototyping: cross-fertilization of knowledge in
in order to take into account the volume of the machines and to better prototype-driven problem solving. J. Prod. Innov. Manag. 31 (4), 744–764.
Bogers, M., Afuah, A., Bastian, B., 2010. Users as innovators: a review, critique, and future
benchmark it to the scale of an injection molding machine. Clearly, research directions. J. Manag. 36 (4), 857–875.
AM machines are much slower than an injection molding machine. Bogers, M., Bekkers, R., Granstrand, O., 2012. Intellectual property and licensing strategies
However FFF prices can go as down as 0.2 euro per part, which makes in open collaborative innovation. In: de Pablos Heredero, C., López, D. (Eds.), Open In-
novation at Firms and Public Administrations: Technologies for Value Creation. IGI
it an attractive technology. SLA and PJM are significantly more expen- Global, Hershey, PA, pp. 37–58.
sive than FFF, while SLS is somewhat cheaper. Bogers, M., Sund, K., Villarroel, J.A., 2015. The organizational dimension of business model
exploration: evidence from the European postal industry. In: Foss, N.J., Saebi, T.
(Eds.), Business Model Innovation: The Organizational Dimension. Oxford University
Press, Oxford, pp. 269–287.
Bouncken, R.B., Gast, J., Kraus, S., Bogers, M., 2015. Coopetition: a systematic review, syn-
thesis, and future research directions. Rev. Manag. Sci. http://dx.doi.org/10.1007/
Table B.2 s11846-11015-10168-11846 Published online ahead of print.
Bradshaw, S., Bowyer, A., Haufe, P., 2010. The intellectual property implications of low-
Time, cost and volume comparison for the production of parts.
cost 3D printing. SCRIPTed 7 (1), 5–31.
FFF SLS SLA PJM Brajlih, T., Drstvensek, I., Kovacic, M., Balic, J., 2006. Optimizing scale factors of the
PolyJet™ rapid prototyping procedure by genetic programming. J. Achiev. Mater.
Time (h) 1.5 2 2 1.2 Manuf. Eng. 16 (1–2), 101–106.
Cost (Euro) 0.20 2.15 3.7 3.6 Brunswicker, S., Wrigley, C., Bucolo, S., 2013. Business model experimentation: what is
Volume (mm) 914 × 610 × 700 × 380 × 550 × 550 × 1000 × 800 × the role of design-led prototyping in developing novel business models? In: Curley,
14 580 750 500 M., Formica, P. (Eds.), The Experimental Nature of New Venture Creation. Springer In-
ternational Publishing, pp. 139–151
Note: Zcorp technology was not evaluated in this case as its platforms are generally rather Casadesus-Masanell, R., Zhu, F., 2013. Business model innovation and competitive imita-
small and are not suitable for large quantities. tion: the case of sponsor-based business models. Strateg. Manag. J. 34 (4), 464–482.
238 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239

Chesbrough, H., 2003. Open Innovation: The New Imperative for Creating and Profiting Kruth, J.P., Mercelis, P., Van Vaerenbergh, J., Froyen, L., Rombouts, M., 2005. Binding mech-
from Technology. Harvard Business School Press, Boston, MA. anisms in selective laser sintering and selective laser melting. Rapid Prototyp. J. 11
Chesbrough, H., 2006. Open Business Models: How to Thrive in the New Innovation Land- (1), 26–36.
scape. Harvard Business School Press, Boston, MA. Lakhani, K.R., von Hippel, E., 2003. How open source software works: ‘free’ user-to-user
Chesbrough, H., 2010. Business model innovation: opportunities and barriers. Long Range assistance. Res. Policy 32 (6), 923–943.
Plan. 43 (2–3), 354–363. Laursen, K., Salter, A., 2014. The paradox of openness: appropriability, external search and
Chesbrough, H., Bogers, M., 2014. Explicating open innovation: clarifying an emerging collaboration. Res. Policy 43 (5), 867–878.
paradigm for understanding innovation. In: Chesbrough, H., Vanhaverbeke, W., Leonard-Barton, D., 1992. Core capabilities and core rigidities: a paradox in managing
West, J. (Eds.), New Frontiers in Open Innovation. Oxford University Press, Oxford, new product development. Strateg. Manag. J. 13 (Summer Special Issue), 111–125.
pp. 3–28. Massa, L., Tucci, C.L., 2014. Business model innovation. In: Dodgson, M., Gann, D.M.,
Chesbrough, H., Rosenbloom, R.S., 2002. The role of the business model in capturing value Phillips, N. (Eds.), The Oxford Handbook of Innovation Management. Oxford Univer-
from innovation: evidence from Xerox Corporation's technology spin-off companies. sity Press, Oxford, pp. 420–441.
Ind. Corp. Chang. 11 (3), 529–555. McGrath, R.G., 2010. Business models: a discovery driven approach. Long Range Plan. 43
Christensen, C.M., 1997. The Innovator's Dilemma: When New Technologies Cause Great (2–3), 247–261.
Firms to Fail. Harvard Business School Press, Boston, MA. Melchels, F.P., Feijen, J., Grijpma, D.W., 2010. A review on stereolithography and its
Christopher, M., 1992. Logistics and Supply Chain Management. Pitman Publishing, applications in biomedical engineering. Biomaterials 31 (24), 6121–6130.
London. Mortara, L., Hughes, J., Ramsundar, P.S., Livesey, F., Probert, D.R., 2009. Proposed classifica-
Christopher, M., Peck, H., Towill, D., 2006. A taxonomy for selecting global supply chain tion scheme for direct writing technologies. Rapid Prototyping J. 15 (4), 299–309.
strategies. Int. J. Logist. Manag. 17 (2), 277–287. O'Hern, M.S., Rindfleisch, A., 2009. Customer co-creation: a typology and research agenda.
Coombes, P.H., Nicholson, J.D., 2013. Business models and their relationship with market- In: Malholtra, N.K. (Ed.)Review of Marketing Research Vol. 6. M.E. Sharpe, Armonk,
ing: a systematic literature review. Ind. Mark. Manag. 42 (5), 656–664. NY, pp. 84–106.
Da Silveira, G., Borenstein, D., Fogliatto, F.S., 2001. Mass customization: literature review Osterwalder, A., Pigneur, Y., 2010. Business Model Generation: A Handbook for Vision-
and research directions. Int. J. Prod. Econ. 72 (1), 1–13. aries, Game Changers, and Challengers. Wiley, Hoboken, NJ.
Deckers, J., Shahzad, K., Vleugels, J., Kruth, J., 2012. Isostatic pressing assisted indirect se- Phaal, R., O'Sullivan, E., Routley, M., Ford, S., Probert, D., 2011. A framework for mapping
lective laser sintering of alumina components. Rapid Prototyp. J. 18 (5), 409–419. industrial emergence. Technol. Forecast. Soc. Chang. 78 (2), 217–230.
Ehret, M., Kashyap, V., Wirtz, J., 2013. Business models: impact on business markets and Pham, D., Gault, R., 1998. A comparison of rapid prototyping technologies. Int. J. Mach.
opportunities for marketing research. Ind. Mark. Manag. 42 (5), 649–655. Tools Manuf. 38 (10–11), 1257–1287.
Eisenmann, T., Parker, G., van Alstyne, M.W., 2006. Strategies for two-sided markets. Harv. Piller, F.T., Moeslein, K., Stotko, C.M., 2004. Does mass customization pay?An economic
Bus. Rev. 84 (10), 92–101. approach to evaluate customer integration. Prod. Plan. Control 15 (4), 435–444.
EnvisionTEC, 2013. Retrieved December 2013, from EnvisionTEC DLP Technology: http:// Piller, F., Weller, C., Kleer, R., 2015. Business models with additive manufacturing: oppor-
www.arburg.com/en/solutions/freeformer/. tunities and challenges from the perspective of economics and management. In:
EOS, 2013. Retrieved December 2013, from EOS Plastic Technology: http://www.eos.info/ Brecher, C. (Ed.), Advances in Production Technology. Springer International Publish-
systems_solutions/plastic/systems_equipment. ing, pp. 39–48.
Fauchart, E., von Hippel, E., 2008. Norms-based intellectual property systems: the case of Pine II, B.J., 1999. Mass Customization: The New Frontier in Business Competition.
French chefs. Organ. Sci. 19 (2), 187–201. Harvard Business School Press, Boston, MA.
Foss, N.J., Saebi, T. (Eds.), 2014. Business Model Innovation: The Organizational Dimen- Ponfoort, O., Ambrosius, W., Barten, L., Duivenvoorde, G., van den Hurk, L., Sabirovic, A.,
sion. Oxford University Press, Oxford. Teunissen, E., 2014. Successful Business Models for 3D Printing: Seizing Opportunities
Franke, N., Piller, F.T., 2003. Key research issues in user interaction with user toolkits in a with a Game Changing Technology. Berenschot, Utrecht.
mass customisation system. Int. J. Technol. Manag. 26 (5), 578–599. Prahalad, C.K., Ramaswamy, V., 2003. The Future of Competition: Co-Creating Unique
Frankenberger, K., Weiblen, T., Gassmann, O., 2013. Network configuration, customer cen- Value with Customers. Harvard Business School Press, Boston, MA.
tricity, and performance of open business models: a solution provider perspective. Reeves, P., Tuck, C., Richard, H., 2011. Additive manufacturing for mass customization. In:
Ind. Mark. Manag. 42 (5), 671–682. Fogliatto, F., da Silveira, G.J. (Eds.), Mass Customization — Engineering and Managing
Gawer, A., Cusumano, M.A., 2014. Industry platforms and ecosystem innovation. J. Prod. Global Operations (pp. 275–290). Springer.
Innov. Manag. 31 (3), 417–433. Reichstein, T., Salter, A.J., Gann, D.M., 2008. Break on through: sources and determinants
George, G., Bock, A.J., 2011. The business model in practice and its implications for entre- of product and process innovation among UK construction firms. Ind. Innov. 15 (6),
preneurship research. Enterp. Theory Pract. 35 (1), 83–111. 601–625.
Geraedts, J.M., Doubrovski, E.L., Verlinden, J., Stellingwerff, M., 2012. Three views on addi- Rennie, E., 2007. Community media in the prosumer era. J. Community Citiz. Third Sector
tive manufacturing: business, research, and education. Proceedings of the TMCE. Media Commun. 3, 25–32.
Karlsruhe, Germany, pp. 1–15. Robertson, R., 1994. Globalisation or glocalisation? J. Int. Commun. 1 (1), 33–52.
Gibson, I., Rosen, D.W., Stucker, B., 2010. Additive Manufacturing Technologies: Rapid Sachs, E. M., Haggerty, J. S., Cima, M. J., & Williams, P. (1993). Patent No. 5204055. USA.
Prototyping to Direct Digital Manufacturing. Springer, New York, NY. Sawhney, M., Verona, G., Prandelli, E., 2005. Collaborating to create: the Internet as a
Hadar, R., Bilberg, A., 2012. Glocalized manufacturing: local supply chains on a global platform for customer engagement in product innovation. J. Interact. Mark. 19
scale and changeable technologies. Proceedings of the 22nd International Confer- (4), 4–17.
ence on Flexible Automation and Intelligent Manufacturing (FAIM), Helsinki, Smith, W.K., Binns, A., Tushman, M.L., 2010. Complex business models: managing strate-
Finland. gic paradoxes simultaneously. Long Range Plan. 43 (2–3), 448–461.
Hague, R., Campbell, R.I., Dickens, P.M., 2003. Implications on design of rapid manufactur- Stratasys, 2013a. Retrieved December 2013, from Stratasys — Ultem: http://www.
ing. Proc. Inst. Mech. Eng. C J. Mech. Eng. Sci. 217 (1), 25–30. stratasys.com/materials/fdm/ultem-9085.
Henderson, R.M., Clark, K.B., 1990. Architectural innovation: the reconfiguration of Stratasys, 2013b. Retrieved December 2013, from Stratasys Production Systems: http://
existing product technologies and the failure of established firms. Adm. Sci. Q. 35 www.stratasys.com/3d-printers/production-series.
(1), 9–30. Stratasys, 2013c. Retrieved December 2013, from Stratasys PJM platform: http://www.
Hesse, M., Rodrigue, J.-P., 2006. Global production networks and the role of logistics and stratasys.com/3D-Printers/technology/polyjet-technology.
transportation. Growth Change 37 (4), 499–509. Stratasys, 2013d. Retrieved December 2013, from Stratasys PLJ materials: http://www.
Hill, C.W.L., Rothaermel, F.T., 2003. The performance of incumbent firms in the face of stratasys.com/materials/polyjet.
radical technological innovation. Acad. Manag. Rev. 28 (2), 257–274. Svensson, G., 2001. “Glocalization” of business activities: a “glocal strategy” approach.
Hopkinson, N., Hague, R., Dickens, P., 2006. Rapid Manufacturing: An Industrial Revolu- Manag. Decis. 39 (1), 6–18.
tion for the Digital Age. John Wiley & Sons, Chichester. Teece, D.J., 2010. Business models, business strategy and innovation. Long Range Plan. 43
Huang, S.H., Liu, P., Mokasdar, A., Hou, L., 2013. Additive manufacturing and its societal (2–3), 172–194.
impact: a literature review. Int. J. Adv. Manuf. Technol. 67 (5–8), 1191–1203. Toffler, A., 1980. The Third Wave. Bantam Books.
Jaehne, D.M., Li, M., Riedel, R., Mueller, E., 2009. Configuring and operating global produc- Tseng, M.M., Piller, F.T. (Eds.), 2003. The Customer Centric Enterprise: Advances in Mass
tion networks. Int. J. Prod. Res. 47 (8), 2013–2030. Customization and Personalization. Springer, Berlin.
Jeppesen, L.B., Frederiksen, L., 2006. Why do users contribute to firm-hosted user commu- Tuck, C., Hague, R., 2006. The pivotal role of rapid manufacturing in the production of
nities? The case of computer-controlled music instruments. Organ. Sci. 17 (1), 45–63. cost-effective customised products. Int. J. Mass Customisation 1 (2), 360–373.
Johnson, M.E., Whang, S., 2002. E-business and supply chain management: an overview Tuck, C., Hague, R., Burns, N., 2007. Rapid manufacturing: impact on supply chain meth-
and framework. Prod. Oper. Manag. 11 (4), 413–423. odologies and practice. Int. J. Serv. Oper. Manag. 3 (1), 1–22.
Johnson, M.W., Christensen, C.M., Kagermann, H., 2008. Reinventing your business model. Tushman, M.L., Anderson, P., 1986. Technological discontinuities and organizational envi-
Harv. Bus. Rev. 68 (12), 57–68. ronments. Adm. Sci. Q. 31 (3), 439–465.
Kang, M.H., 2010. Risks of global production systems: lessons from Toyota's mass recalls. Vanhaverbeke, W., Chesbrough, H., 2014. A classification of open innovation and open
SERI Q. 65–71. business models. In: Chesbrough, H., Vanhaverbeke, W., West, J. (Eds.), New Frontiers
Kodama, F., 2004. Measuring emerging categories of innovation: modularity and business in Open Innovation. Oxford University Press, Oxford, pp. 50–68.
model. Technol. Forecast. Soc. Chang. 71 (6), 623–633. Vaupotič, B., Brezočnik, M., Balič, J., 2006. Use of PolyJet technology in manufacture of new
Koren, Y., 2010. The Global Manufacturing Revolution: Product–Process–Business product. J. Achiev. Mater. Manuf Eng. 18 (1–2), 319–322.
Integration and Reconfigurable Systems. Wiley, Hoboken, NJ. von Hippel, E., 2005. Democratizing Innovation. MIT Press, Cambridge, MA.
Kraft, K., 1990. Are product and process innovations independent of each other? Appl. von Hippel, E., von Krogh, G., 2006. Free revealing and the private-collective model of
Econ. 22 (8), 1029–1038. innovation incentives. R&D Manag. 36 (3), 295–306.
Kruth, J., Wang, X., Laoui, T., Froyen, L., 2003. Lasers and materials in selective laser Voxeljet, 2013. Retrieved December 2013, from Voxeljet Systems: http://www.voxeljet.
sintering. Assem. Autom. 23 (4), 357–371. de/en/systems/.
 M. Bogers et al. / Technological Forecasting & Social Change 102 (2016) 225–239 239

Walter, M., Holmström, J., Yrjölä, H., 2004. Rapid manufacturing and its impact on supply Ronen Hadar is a senior manager specialist at the frontend technology development cen-
chain management. Logistics Research Network Annual Conference, September 9–10. ter of a large manufacturer in Denmark. His role is within scouting and evaluating new
Dublin, Ireland. technologies, developing them, and providing a proof of concept that will demonstrate
Walters, P., Huson, D., Parraman, C., Stanić, M., 2009. 3D printing in colour: technical eval- the feasibility and advantages of selected technologies. During his industrial career, Ronen
uation and creative applications. Impact Multidisciplinary Printmaking Conference, worked with many manufacturing and product technologies and with an emphasis on
pp. 16–19 Bristol. digital technologies and manufacturing. He obtained a B.Sc. and M.Sc. in Engineering in In-
West, J., Bogers, M., 2014. Leveraging external sources of innovation: a review of research novation & Business and an industrial Ph.D. in Product Design and Innovation from the
on open innovation. J. Prod. Innov. Manag. 31 (7), 814–831. University of Southern Denmark. In the past, he owned several technology-based start-
Wirtz, B.W., Schilke, O., Ullrich, S., 2010. Strategic development of business models: impli- ups. His main interests are additive manufacturing, product and business development,
cations of the Web 2.0 for creating value on the Internet. Long Range Plan. 43 (2–3), upcoming production technologies and platforms, and supply chain structures.
272–290.
Wohlers, T., 2013. Additive manufacturing and 3D printing state of the industry. Annual
Worldwide Progress Report. Wohlers Associates. Arne Bilberg is an associate professor in Operations Management and Technology Innova-
Wohlers, T., 2014. 3D printing and additive manufacturing state of the industry. Annual tion. He obtained a M.Sc. in Mechanical Engineering in 1985, and a Ph.D. in Computer
Worldwide Progress Report. Wohlers Associates. Integrated Manufacturing in 1989 from the Technical University of Denmark. From 1990
Wong, K.V., Hernandez, A., 2012. A review of additive manufacturing. ISRN Mech. Eng. to 1991, he was a visiting professor at State University of New York, responsible for tech-
2012, 1–10. nology transfer and building up a laboratory within Computer Integrated Manufacturing.
Zott, C., Amit, R., 2010. Business model design: an activity system perspective. Long Range In 1995, he was appointed associate professor at the Institute of Process and Production
Plan. 43 (2–3), 216–226. Engineering at the Technical University of Denmark. In 1998, he was employed as Tech-
Zott, C., Amit, R., Massa, L., 2011. The business model: recent developments and future nology Architect at the Danish international company Linak A/S, with responsibility for
research. J. Manag. 37 (4), 1019–1042. technology innovation. In 2004, he was appointed associate professor at the University
of Southern Denmark, in the research area Mechatronics Products and Manufacturing In-
novation. In 2005, he was the research manager for the Technology Group with research in
Marcel Bogers is an associate professor of innovation and entrepreneurship at the Depart-
mechatronics product innovation. In 2012, he became the research manager in Innovation
ment of Food and Resource Economics (Section for Production, Markets and Policy) at the
and Business with a research focus at the Smart Factory of the Future, based on collabora-
University of Copenhagen. He was formerly at the Mads Clausen Institute at the University
tion among SMEs and automation solutions with a human touch, the so-called Lean
of Southern Denmark where this research was carried out. He obtained a combined B.Sc.
Automation approach.
and M.Sc. in Technology and Society (Innovation Sciences) from Eindhoven University
of Technology and a Ph.D. in Management of Technology from Ecole Polytechnique
Fédérale de Lausanne (Swiss Federal Institute of Technology). He has held visiting scholar
positions at the University of Trento and Chalmers University of Technology. His main in-
terests center on the design, organization and management of technology, innovation and
entrepreneurship. More specifically, his research explores openness and participation in
innovation and entrepreneurial processes within, outside and between organizations. In
this context, he has studied issues such as open innovation, business models, family
businesses, users as innovators, collaborative prototyping, improvisation, and learning
by doing.
This article was downloaded by: [129.93.16.3] On: 02 April 2015, At: 21:28
Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA

Organization Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org

Digital Innovation and the Division of Innovative Labor:

Digital Controls in the Automotive Industry
Jaegul Lee, Nicholas Berente,

To cite this article:

Jaegul Lee, Nicholas Berente, (2012) Digital Innovation and the Division of Innovative Labor: Digital Controls in the
Automotive Industry. Organization Science 23(5):1428-1447. http://dx.doi.org/10.1287/orsc.1110.0707

Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact [email protected].

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.

Copyright © 2012, INFORMS

Please scroll down for article—it is on subsequent pages

INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
 Organization Science
Vol. 23, No. 5, September–October 2012, pp. 1428–1447
ISSN 1047-7039 (print) — ISSN 1526-5455 (online) http://dx.doi.org/10.1287/orsc.1110.0707
© 2012 INFORMS

Digital Innovation and the Division of Innovative Labor:

Digital Controls in the Automotive Industry
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

Jaegul Lee
School of Business Administration, Wayne State University, Detroit, Michigan 48202,
[email protected]

Nicholas Berente
Terry College of Business, University of Georgia, Athens, Georgia 30602,
[email protected]

n this study of the U.S. automobile industry, we highlight the way the division of innovative labor across firms in the
I supply chain can be influenced by a particular form of digital innovation known as “digital control systems.” Digital
control systems are becoming ubiquitous in complex products, and these digital innovations integrate other components
across a product structure and introduce a level of indeterminacy and unpredictability in the organization of the interfirm
division of innovative labor. Much of organizational scholarship holds that accompanying a shift toward increasingly mod-
ular product structures, component suppliers are engaging in relatively more design and invention around the components
that they supply. We find that the evolution of digital controls may reverse this pattern, because in the wake of a major
shift in the digital controls technology, suppliers actually engage in relatively less component innovation in comparison
with their large manufacturing customers. To explain this shift, we characterize complex product structures in terms of
two distinct product hierarchies: the inclusionary and the digital control hierarchy. In using this distinction to analyze the
evolution of automotive emission control systems from 1970 to 1998, we reconcile two competing views about the interfirm
division of innovative labor.
Key words: digital innovation; digital controls; mirroring hypothesis; division of innovative labor; systems integration;
dual-product hierarchy; automotive industry; inclusionary hierarchy; digital control hierarchy
History: Published online in Articles in Advance December 2, 2011.

Introduction work to make their complex products less integrated

Contemporary complex product development takes place and more modular (Brusoni 2005, Christensen 2006,
throughout supply chains, where diverse, specialized Argyres and Bigelow 2010). Modular product struc-
knowledge sources collaborate across firm boundaries tures allow firms to specialize and divide product devel-
(Von Hippel 1988, Powell et al. 1996, Inkpen and Dinur opment labor efficiently across multiple organizations
1998, Helper et al. 2000, Van de Ven 2005). This “divi- and also outsource many design activities to suppli-
sion of innovative labor” (Arora and Gambardella 1994) ers (Langlois 2003, Christensen 2006). According to
is becoming increasingly important to organizations the mirroring hypothesis, suppliers responsible for man-
because rapid advancements in information technolo- ufacturing components are taking a greater hand in
gies and trends toward product modularization enable innovating around those components (i.e., “component
unprecedented levels of interorganizational design coor- innovation”), whereas their large industrial customers
dination (Brusoni 2005). Thus, there is growing atten- focus more on innovating at a broader, productwide level
tion to the division of innovative labor across a vari- that is more concerned with the interactions between
ety of traditions of organizational scholarship (e.g., components (Anderson and Tushman 1990, Baldwin
Takeishi 2001, Brusoni and Prencipe 2006, Lichten- 2008). This productwide innovation is referred to as
thaler 2008, Argyres and Bigelow 2010, Hoang and “architectural innovation” (Henderson and Clark 1990),
Rothaermel 2010). in that such activity changes elements of the schema
The prevailing explanation for the shape of the divi- through which product components are organized (i.e.,
sion of innovative labor is known as the “mirroring product architectures; see Ulrich 1995).
hypothesis,” which states that innovative efforts take Digital technologies, however, may pose a problem
place across organizations in a way that is gener- to the mirroring hypothesis. A certain class of digi-
ally isomorphic, or “mirrors,” the structure of prod- tal technologies, known as “digital control systems,” is
ucts that these firms produce (Baldwin 2008). Much becoming ubiquitous in contemporary complex products,
of this literature focuses on the activities of large spanning widely different components across those prod-
industrial manufacturers and their suppliers as they ucts (Franklin et al. 1998). These digital systems monitor
1428
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1429

and control components with respect to other indirectly they are perhaps the first widespread automotive inno-
connected components, often acting as the core integra- vation that would not have been possible without digital
tive systems of those complex products (Hobday 1998, technology (Jurgen 1999). Emission control technologies
p. 695). The transformational potential of digital con- have been explored both in the literature on organiza-
trols blurs the stabilized boundaries between compo- tional innovation (Lee and Veloso 2008) and in research
nents in what can be fairly modular product architectures into the effects of digital innovation on the automotive
because these digital technologies integrate across previ- industry (King and Lyytinen 2005).
ously unconnected product components. Digital control By studying the technological evolution of emission
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

systems are separate from the traditional product struc- control systems, we look to understand the relationship
ture and integrate, monitor, and control the components between technological changes and the resulting divi-
that form that structure. This cross-component integra- sion of innovative labor. In particular, we investigate two
tion undermines the ability of component interfaces to innovations in emission controls: the first involves tech-
substitute for coordination and component knowledge nological change in the inclusionary hierarchy (catalyst
and thus leads us to the question of whether the division architecture in 1981) and the second involves change in
of innovative labor suggested by the mirroring hypoth- digital control systems (on-board computing in 1994).
esis holds in contexts abundant in component-spanning Through an analysis of patenting activity between 1970
digital control systems. An emerging view, referred to and 1998, we find that automotive original equipment
as the “systems integration perspective” (Prencipe et al. manufacturers (OEMs) were inclined to increase their
2003), offers an alternative lens for understanding the relative focus on architectural innovation after a tech-
division of innovative labor when digital controls span nological shift in the inclusionary product hierarchy—a
complex product architectures. finding that is consistent with the mirroring hypothesis.
The systems integration perspective holds that man- However, after the shift in the digital control hierarchy,
ufacturers of complex products need to integrate and OEMs refocus on component innovation—a finding con-
coordinate both internally and externally designed com- sistent with the systems integration perspective. Thus,
ponents within a given product architecture, thus main- we find support for the key argument that by distin-
taining capabilities wider than those needed to support guishing between two product hierarchies, two distinct
their production (Brusoni et al. 2001; Fine and Whitney theoretical positions (i.e., mirroring hypothesis and sys-
tems integration perspective) can be integrated. Further-
1998; Grandstrand et al. 1997; Prencipe 2000; Takeishi
more, we offer an explanation for why the emerging
2001, 2002). Hence, in the wake of a major shift in
systems integration perspective is becoming increasingly
a product’s architecture, systems integrators need to
relevant, because it helps to explain elements of digital
develop internal capabilities in components to gain better
innovation.
understanding of the inner workings of the new com-
The remainder of this paper is organized as follows.
ponents. They cannot rely solely on external suppli-
We briefly introduce the concept of dual product hierar-
ers to gain knowledge about new developments (Fine
chies in complex system architectures, followed by our
and Whitney 1999, Takeishi 2002). Furthermore, these
hypotheses on the division of innovative labor in the
suppliers should, in turn, gain broader, cross-subsystem wake of technological change across each form of hier-
expertise to better provide bundled systems to their cus- archy. Then we present our research and findings. We
tomers (Davies 2003). conclude with a discussion of the implications of this
The mirroring hypothesis and the systems integra- research.
tion perspective often present contradictory assump-
tions about the division of innovative labor after major
shifts in product structures. However, both perspectives Digital Controls, Control Hierarchies, and
characterize this product structure in terms of a sin- Complex Products
gle product hierarchy—the product hierarchy that Mur- Whether systems are biological, mechanical, or social
mann and Frenken (2006) refer to as an “inclusionary in nature, any given system can be characterized in
hierarchy.” In this study, we draw on systems control a variety of hierarchical ordering schemes (Bertalanffy
theory (Mesarovic et al. 1970) to integrate these perspec- 1968). Although there are multiple hierarchical schemes
tives by considering digital control systems as a second through which we can make sense of systems, it is cus-
product hierarchy in complex products. In doing so, we tomary to focus on a single hierarchy at a time or to take
look to the automotive industry—an industry that has two in opposition. Two ordering schemes commonly
been dramatically affected by digital innovation (Yoo held in contrast are structural and functional hierarchies.
2010), where digital components have grown exponen- The former refers to components or parts, whereas the
tially in recent decades (King and Lyytinen 2005, Leen latter describes processes and actions (Bertalanffy 1968).
and Heffernan 2002, Henfridsson et al. 2009). Emission The former emphasizes the physical, whereas the lat-
controls are particularly relevant to this study because ter emphasizes information. This duality is present in
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1430 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

Figure 1 Illustration of the Dual-Product Hierarchy View of of the system (Mesarovic et al. 1970). Using an auto-
Complex Systems mobile as an example, cruise control can be considered
a decision unit that is local to the components that pro-
Subsystem duce acceleration, which addresses the goal of keeping
Local the speed within a range, whereas a human driver can be
decision regarded as the coordinator that sets the cruise control
unit
Subsystem in relation to the external environment. In the context of
Coordinator contemporary complex products, we expect to find both
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

a hierarchically decomposed inclusionary (“controlled”)

Local
hierarchy and a multilevel control hierarchy (see Fig-
Subsystem Subsystem decision ure 1). The control hierarchy is viewed as a hierarchy (as
unit
opposed to a network, for example) because of the sub-
ordinate arrangement of the local decision units to the
coordinator, and the coordinators to other coordinators,
including human coordinators.
Inclusionary hierarchy Control hierarchy Traditionally, organizational theorists did not need
(“controlled” system) (control system/network) to distinguish between inclusionary and control hierar-
chies because local decision units (whether digital or
discussions across the entire range of systems types— analog) were generally considered part of the subsys-
from biological systems to linguistic systems. Biological tem they control, largely because previous technologies
systems contain both hierarchies of particles and hierar- had limited capacity for spanning multiple subsystems.
chies of genetic code, whereas linguistic systems contain Local controllers are tightly coupled with the subsys-
both hierarchies of physiology and hierarchies of syn- tems that they control and thus fit nicely within any
tax (Bertalanffy 1968, Wilson 1969). In each of these characterization of a complex system in terms of its
examples, the latter, information-related hierarchy, com- inclusionary product hierarchy. However, this is chang-
prises the rules or decision elements associated with the ing in recent decades with the emergence of more pow-
proper function of the system. This second hierarchy erful digital controls. As a result, digital controllers are
controls and coordinates the system, whereas the former being increasingly decoupled from the subassemblies
describes the parts, components, and subsystems that they control.
comprise the system to be controlled and coordinated. Digital controls involve the “control of physical
The seminal work on “multilevel control hierarchies” systems with a digital computer or microcontroller”
by Mesarovic et al. (1970) applied this dual-hierarchy (Franklin et al. 1998, p. 1). Digital controllers offer
view to decision making associated with complex the ability to span local decision units increasingly
sociotechnical systems. Such systems are composed of without human intervention. With the exponential gains
their functional components, on the one hand, and the in computing power, microprocessors are capable of
multilevel “control hierarchy,” which is often separate handling progressively more complex decisions, and
from the functional hierarchy of the system itself, on decision logic can be changed even after the controllers
the other (see Figure 1; Mesarovic et al. 1970).1 Simple are implemented, lending such software-based controls
decisions can be programmed into controls within each an emergent, unpredictable dimension. As cross-module,
element or module, but complex decisions must span digitally integrated controllers become more and more
these elements in a form of multilevel control that can prevalent, microprocessors offer an alternative avenue
include automated controllers and humans. This concept for fast, accurate, and reliable decision making through
of a multilevel control hierarchy distinguishes between programmed software (Findeisen et al. 1980, Drouin
the controlled system hierarchy, which is sometimes et al. 1991). Furthermore, digital controls have a genera-
referred to as the “inclusionary” hierarchy (e.g., see tive aspect to them. Their inclusion in a complex system
Murmann and Frenken 2006), and the control hierarchy. tends to stimulate additional, often unforeseen, digital
The term “control hierarchy” refers to the organiza- applications (Franklin et al. 1998). Thus, digital con-
tion of decision-making units controlling a system and trol systems do not merely replace functionally equiv-
involves humans, digital controls, analog controls, or alent inclusionary controls but also dramatically extend
some combination thereof (Findeisen et al. 1980). Con- the range and scope of automated control, enabling
trol units in a multilevel hierarchy are distinguished by a variety of new applications and thus compound-
goals. Local decision units nearer to the component level ing the uncertainty and unpredictability associated with
may address local, operational goals related to specific potential applications. These novel applications, in turn,
components and subsystems, whereas coordinator units increase the complexity associated with contemporary
span multiple distributed units and accommodate higher- products. Tightly coupled, inclusionary controls, in con-
level goals associated with increasingly broader portions trast, inevitably work to constrain further innovation
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1431

of the components that they control at a certain point means that components are largely decoupled from each
in the product’s evolution (Prencipe 2000). Advanced other except through standard and well-defined inter-
microprocessors can be programmed to handle more faces (Ulrich 1995). Integrated architectures have a far
complexity to relax some of these constraints. more complex mapping of functions to components, and
In contemporary automobiles, for example, embed- their components have multiple connections and interac-
ded digital control systems now integrate components tions (Ulrich 1995).
across increasingly diverse domains (King and Lyytinen The firms in a supply chain of a given complex prod-
2005). Digital controls do not merely take the place uct are also characterized in terms of two broad cate-
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

of analog forms of control found within inclusionary gories: OEMs and suppliers (Brusoni 2005, Christensen
automotive subsystems but also enable new possibilities 2006, Argyres and Bigelow 2010). The OEMs are large
(Koopman 2002, Leen and Hefferman 2002). Through industrial firms that act as assemblers, integrators, and
computational (decision-making) performance, accuracy, marketers that bridge customers and the supply chain
flexibility, scalability, and networkability (Drouin et al. and generally exercise a great deal of control across the
1991, Jurgen 1999, Pop et al. 2004), digital controls supply chain. Suppliers provide components to OEMS,
enable engineers to equip automobiles with the ability to and, according to much of the literature, OEMs are
monitor and regulate subsystems in a variety of ways, increasingly relying on suppliers to produce increasingly
to process data in real time quickly and reliably, and complex products (Langlois 2003, Argyres and Bigelow
to link otherwise decoupled subsystems so that these 2010). When specialized suppliers merely manufacture
subsystems become responsive to each other (Dorf and components for OEMs, this can be referred to as the
Bishop 2008). Through technological advancements in interfirm division of labor. When those suppliers engage
recent decades, digital controls have enabled the con- in the design and innovation around the components that
trol of unprecedented levels of product performance and they produce, this can be characterized as the interfirm
complexity. division of innovative labor (Arora and Gambardella
Thus, the fundamental premise of this research is that 1994), where a more broadly held set of capabilities and
digital controls are reshaping the way firms organize for knowledge specializations apply. There are a variety of
complex product innovation. Specifically, as a result of explanations for the relationship between complex prod-
digital innovation, control systems are decoupling from uct architectures and the distribution of innovative activ-
the inclusionary hierarchy (Pop et al. 2004, Steinmueller ity across OEMs and suppliers (Brusoni 2005, Chris-
2003), and this digital control hierarchy may bring with tensen 2006).
it the requirements for a new set of integrative capabil- These explanations can be categorized in terms of
ities (Curtis et al. 1988, Hobday 1998). To investigate two broad perspectives about the relationship between
the impact of a view of complex products that considers changes to technological architecture, what we refer to
dual product hierarchies, next we address two competing as “technological shifts,” and the interfirm division of
explanations of the relationship between technological innovative labor. Both of these explanations attend to
change and the division of innovative labor across a sup- issues associated with the knowledge resources neces-
ply chain. sary to innovate with respect to either components or
architectures in complex product development. The first
Hypotheses: Technological Change and explanation is the mirroring hypothesis (Baldwin 2008),
the Division of Innovative Labor which finds its roots in Simon’s (1996) concept of nearly
Organizational research has a history of theorizing about decomposable systems and modularity, and in the strate-
the relationship between technological changes to com- gic theories of a firm (i.e., transaction cost economics
plex product structures and the activities between firms and resource-based view). Consistent with this view, the
in a supply chain (Abernathy and Utterback 1978, relation between changing product architectures and the
Anderson and Tushman 1990, Henderson and Clark division of innovative labor is captured in the notion
1990, Christensen 2006). In this literature, technological of a dominant design (Anderson and Tushman 1990).
changes are typically characterized as either component The second explanation is the systems integration per-
innovations or architectural innovations (Henderson and spective (Prencipe et al. 2003), rooted in the knowledge-
Clark 1990, Ulrich 1995), where the former involves based view of the firm (Kogut and Zander 1992), which
changes to technologies within individual components or indicates that complex product manufacturers need to
subsystems and the latter involves changes to schemes maintain competencies beyond their immediate needs to
for organizing among individual components or subsys- innovate effectively.
tems in the context of broader systems (Whitney 1990, In the previous section, we distinguished between
Ulrich 1995). Architectures are typically conceived of as inclusionary and digital control hierarchies. Although
either modular or integrated (Ulrich 1995, Sanchez and both product hierarchies comprise increasingly modular
Mahoney 1996). Modular architectures involve one-to- architectures, the effect of the digital control hierarchy
one mapping between functions and components, which is that of integration across the inclusionary hierarchy.
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1432 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

Based on this distinction, we expect the dynamics asso- or subassembly. Because of this specialized knowledge,
ciated with changes to inclusionary hierarchies to remain they offer their customers advantages in quality, cost,
consistent with the mirroring hypothesis and changes to and time to market the components and subassemblies
digital control hierarchies to be more consistent with they design and produce (Clark 1989, Clark and Fuji-
the systems integration perspective. Next, we detail this moto 1991, Ragatz et al. 1997, Handfield et al. 1999).
logic and present our hypotheses. The use of trusted suppliers reduces an OEM’s need
to maintain knowledge in every component or sub-
Inclusionary Hierarchies and assembly, allowing them to focus more on architectural
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

the Mirroring Hypothesis and systemwide innovation. Thus, after a technological

According to the mirroring hypothesis, periods of shift in a given complex product architecture, accord-
dramatic technological change often lead to the emer- ing to the mirroring hypothesis, there is a decreased
gence of a dominant design, after which the compet- share in overall component-level innovation on the part
itive dynamics in an industry become more focused of the assembling manufacturers (Henderson and Clark
on incremental innovation (Abernathy and Utterback 1990, Murmann and Frenken 2006). Furthermore, there
1978, Anderson and Tushman 1990). A dominant design is less need for suppliers to attend to architectural issues
implies a moment of stability in the evolution of a prod- because the architecture is settled, and the interfaces
uct’s architecture, indicating an agreement among vari- are clearly defined and become standard. However, even
ous technical, political, and social forces (Anderson and in the context of highly modular product architectures,
Tushman 1990). This point of agreement is followed by OEMs will not relinquish component innovation entirely,
a period of autonomous technological progress, where particularly with respect to strategically important com-
limited communities of specialists focus on incremental ponents. The knowledge of key components remains
enhancement within the components associated with the important to an organization’s ability to innovate on the
dominant design (Tushman and Murmann 1998). product’s architecture and to coordinate product devel-
Whether focusing on the architecture of the entire opment across the supply chain (Sanchez and Mahoney
complex products or on particular subsystems, the liter- 1996, Christensen 2006). Therefore, finding situations
ature on dominant design finds its roots in the observa- wherein OEMs completely abandon component innova-
tion that patterns of technological innovation associated tion, for example, is unlikely. What can be found instead
with a complex system design are mirrored by the rel- is the decrease in relative proportion or overall share in
evant structures within and across organizations in an the ever-changing level of component innovation com-
industry (Abernathy and Utterback 1978). Thus, orga- pared with the share of the suppliers.
nizations responsible for higher levels of product hier- As technological change in the inclusionary hierarchy
archy, such as OEMs, are more directly involved in of complex products is relatively modular and repre-
component-level innovation before a dominant design is sents the nested components, subsystems, and systems
established, but they leave component innovation more most often characterized in organizational literature, we
to their suppliers after the dominant design emerges expect changes in the inclusionary hierarchy to reflect
and clear performance metrics and product interfaces the assumptions of the mirroring hypothesis. Therefore,
are established (Murmann and Frenken 2006). Similarly, we hypothesize the following.
competing suppliers engage in architectural-level inno-
Hypothesis 1A (H1A). After a technological shift in
vation before a dominant design emerges in the hopes of
the inclusionary product hierarchy of a complex prod-
influencing the architecture in their favor. However, this
uct, the suppliers’ overall share in component innovation
diminishes after a standard architecture is established.
increases.
This situation results in relative stability and incremental
change until the next technological shift to a subsequent Hypothesis 1B (H1B). After a technological shift in
dominant design (Anderson and Tushman 1990). the inclusionary product hierarchy of a complex prod-
When a dominant design emerges for a system or sub- uct, the OEMs’ overall share in architectural innovation
system, the focus of overall innovation shifts to incre- increases.
mental component innovation (Henderson and Clark
1990, Ulrich 1995). Such a technological shift reduces Digital Control Hierarchies and
market uncertainty and allows assemblers to incorporate the Systems Integration Perspective
key suppliers into their product development activities The systems integration perspective (Prencipe et al.
strategically (Venkatesan 1992, Anderson and Tushman 2003) can be used in contrast to the mirroring hypoth-
1990), thus stimulating the rise of specialized suppliers esis. According to the systems integration perspective,
that contribute to subsequent product innovations. These as a system evolves, the product gains complexity with
suppliers take on a larger share in product development a greater number of interfaces and linkages (Murmann
and obtain valuable, focused experience that dramati- and Frenken 2006, Tushman and Murmann 1998). It is
cally increases their knowledge about the component up to the systems integrators (i.e., “OEMs”) to build
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1433

the requisite knowledge for managing these systems Moreover, the practices, competencies, and forms of
and innovating across both components and systems. knowledge associated with digital systems development
OEMs, as systems integrators, are firms that possess are fundamentally different from the design of other arti-
technological and organizational capabilities to integrate facts (Turner 1987, Kettunen and Laanti 2005). The pro-
changes and improvements in internally and externally cess of design for digital systems is different from that
designed and produced inputs within existing product for electromechanical products in both the steps of the
architectures (Brusoni et al. 2001, p. 614). A stream process and in the way tools are used in the design activ-
of research associated with this view provides empiri- ity (Whitney 1990). In addition, digital system develop-
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

cal evidence that OEMs possess product design knowl- ers necessarily bring capabilities outside the application
edge and related capabilities for a much broader set of domain that they must span and integrate with the appli-
components than those they directly design and manu- cation domain (Curtis et al. 1988).
facture (Brusoni et al. 2001; Fine and Whitney 1999; Thus, as digital control systems become more preva-
Grandstrand et al. 1997; Prencipe 2000; Takeishi 2001, lent, controls of inclusionary product hierarchies become
2002). OEMs function as all-round knowledgeable firms, decoupled from those hierarchies, and the systems inte-
capable of coordinating changes in a variety of relevant gration perspective would indicate that there should then
technical fields (Brusoni et al. 2001, Prencipe 2002). be a shift toward innovative practices that reflect this
They also maintain the capabilities required for introduc- spanning or integrating of the inclusionary product hier-
ing new product subarchitectures. OEMs must not only archy (Prencipe 2000, Brusoni et al. 2001). Therefore,
maintain component knowledge and architectural inte- we hypothesize that technological shifts in the digital
gration knowledge but also organizational knowledge on control hierarchy are likely to require OEMs to expand
how to integrate different specialist suppliers in the sup- their relative share of component innovations, whereas
ply chain (Christensen 2006). their suppliers would become increasingly involved in
When a major technological shift occurs at a sub- architectural innovation. Thus, we expect technological
system level, OEMs need to develop internal capabili- shifts in the digital control hierarchy to be consistent
ties in components to gain better understanding of the with the systems integration perspective.
inner workings of the new subarchitecture and compo-
nents. OEMs cannot rely solely on external component Hypothesis 2A (H2A). After a technological shift in
suppliers to adequately stay current about new develop- the digital control hierarchy, the suppliers’ overall share
ments (Fine and Whitney 1999, Takeishi 2002). Critical in architectural innovation increases.
engineering knowledge may leak to potential competi- Hypothesis 2B (H2B). After a technological shift in
tors, and OEMs may lose negotiating power to sup- the digital control hierarchy, the OEMs’ overall share in
pliers if they rely too heavily on suppliers for new component innovation increases.
knowledge (Fine and Whitney 1999, Takeishi 2002).
Consequently, as a given system evolves with increas- To test these hypotheses across technological shifts
ing complexity, OEMs need to gain better understanding involving both the inclusionary and the digital control
of the component functionality to assess, test, and inte- hierarchies, we study the patterns of organizational inno-
grate components produced from both in-house research vation associated with nearly two decades of automotive
and development (R&D) functions and external suppliers emissions control systems.
(Prencipe 2000). According to the systems integration
perspective, after a technological shift, suppliers also
Research
look to gain architectural expertise to move downstream
in the supply chain and potentially improve their market Research Context: Automotive Emissions Control
position by providing aggregated, or “bundled,” subsys- Automobile emissions control technology, between 1970
tems (Davies 2003, Dosi et al. 2003). and 1998, provides a suitable context for exploring com-
As indicated above, innovations associated with dig- plex technological shifts involving both the inclusion-
ital control hierarchies represent a relatively different ary and the digital control hierarchies for three reasons.
form of innovation for firms that manufacture complex First, emissions control systems are the first automotive
electromechanical products. Digital control systems can subsystems where the control functionality demanded
be decoupled from traditional inclusionary product hier- by environmental forces (i.e., regulation) is not met
archies (Findeisen et al. 1980, Drouin et al. 1991, Stein- with existing technologies, necessitating digital control
mueller 2003, Pop et al. 2004), integrate across these (Jurgen 1999). Second, the architecture, with respect to
hierarchies (Hobday 1998), and allow for ever-greater the inclusionary product hierarchy, after intense com-
integrative combinations in the future (Franklin et al. petition across a variety of architectures in the 1970s,
1998). Digital controls are necessary for breaching the remains completely stable since the dominant design
ceiling of limited intermodule coordination associated emerged in 1981. Third, largely because of environ-
with inclusionary product hierarchies (Prencipe 2000). mental concerns, digital control hierarchies related to
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1434 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

Figure 2 Digital Control and Inclusionary Hierarchies in Emissions Control Systems

Exhaust system
Air
meter Air
Exhaust Exhaust Resonators
Engine Catalytic converter pipe (mufflers/silencers) Tailpipe
pipe

Gasoline
Sensor(s)
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

Fuel
meter
Key

Inclusionary hierarchy

Electronic
control unit Control hierarchy
(ECU)

emissions control are under increased pressure for more 1998 is shown in Figure 3. Next, we describe each of
complex controls in response to regulation and consumer these architectural innovations.
tastes (Powers and Nicastri 2000, Papodopoulos and
Grante 2005). Emissions control systems underwent two Technological Shift in the Inclusionary Product Hier-
readily identifiable technological shifts over the course archy. Catalyst technology was the primary focus of
of two decades. The first involved a dominant design innovation in automobile emissions control systems
in the inclusionary product hierarchy, and the second throughout the 1970s. High automobile tailpipe emis-
involved a technological shift of the digital control hier- sions reduction requirements mandated by the Clean
archy that integrated across these systems over a decade Air Act Amendments in 1970 (1970 CAAA) led to the
later (see Figure 2). development of catalyst-based emissions reduction tech-
In the development of automobile emissions control nology, or what is commonly referred to as catalytic
systems, a dominant product-level design in the inclu- converter technology. The regulation required automak-
sionary product hierarchy emerged in 1981: an architec- ers to reduce three pollutants simultaneously, namely,
ture described as a “catalyst-based three-way catalytic hydrocarbon (HC), carbon monoxide (CO), and nitro-
converter” (TWC). This technological shift was followed gen oxides (NO, NO2 ; referred to as NOx ). Potential
by a period of incremental change until 1994 with the solutions to emissions control problems abounded in
advanced onboard diagnostic electronic control module the 1970s. An architecture involving “add-in-type” cat-
(OBD), a technological shift in the digital control hier- alytic converters emerged as the viable, standard techni-
archy. The subsystem-level technological evolution in cal solution for reducing all three pollutants. Eventually,
automobile emissions control technologies from 1970 to the TWC converter technology using monolithic-type

Figure 3 Technology Evolution of the Automobile Emissions Control Technology

Inclusionary hierarchy Control hierarchy

Emergence of a Emergence of a subproduct-level

dominant design technological shift
(catalytic converters) (onboard diagnostic system)

Era of ferment Era of incremental Era of product

innovation optimization

1970 1981 1990 1994

Enactment of the Enactment of the

1970 CAAA 1990 CAAA
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1435

catalysts became the dominant design for automobile Patent Search. As the first step in the analysis,
emissions control systems in 1981 and finally satisfied the first author identified relevant patents in the auto
the 90% reduction requirement originally mandated in emissions control technology using successfully applied
the Amendments of the Clean Air Act in 1970. patents in the U.S. Patent and Trademark Office
(USPTO) public database from 1970 to 1998. He used
Technological Shift in the Digital Control Hierar- two approaches in building a patent database on auto-
chy. Many subsequent innovative efforts following the mobile emissions control systems, namely, the class-
TWC standard in 1981 focused on making incremental based search and the abstract-based keyword search.
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

improvements in the inclusionary product hierarchy, on The goal of using both search strategies is to cap-
components such as catalysts and fuel injection systems. ture relevant patents comprehensively. The class-based
Accompanying these incremental innovations, however, search and the abstract-based keyword search comple-
was a growing focus on developing electronics subsys- ment each other. The class-based search, with its detailed
tems. Much of this focus on electronics was due to pres- subclasses, helps researchers capture a broad range of
sure from the U.S. Environmental Protection Agency relevant patents that may not otherwise be captured
(EPA), which was looking to impose more stringent alone under the abstract-based keyword search. How-
automobile emissions reduction requirements (Lee and ever, when the set of relevant technologies is diverse, the
Veloso 2008). Amendments to the Clean Air Act in class-based search approach could miss relevant patents
1990 (1990 CAAA) mandated short-term lowering of within its identified classes. The abstract-based key-
both HC and NOx by 39% and longer-term lowering of word search complements the class-based approach by
HC by 70%, CO by 50%, and NOx by 80% relative allowing researchers to double-check findings and iden-
to the 1990 emissions reduction levels (Mondt 2000). tify potentially important patents not found under class-
These pressures eventually resulted in the commercial- based search.
ization of the advanced electronic onboard diagnostic In the class-based search approach, we adopted the
module (OMD). OMD’s main functions were to mon- patent classes for the catalytic converter technology pre-
itor all emissions-related devices mounted on a vehi- viously used by the Battelle research group (Camp-
cle, which included not only hydrogen reduction devices bell and Levine 1984). Battelle research used patent
during the cold-start but also catalyst efficiency, engine subclasses within two patent classes, namely, Class 55
misfire, oxygen sensor response, abnormality in fuel (gas separation) and Class 423 (inorganic chemistry).
supply system, exhaust gas recirculation (EGR) valve, Class 55 represents the internal structure and con-
secondary air system, and abnormal engine sensor out- trol systems used for catalytic converters, whereas
put (Mondt 2000). Class 423 is mostly related to the chemical composi-
tion of a catalyst or its supporting materials (Camp-
bell and Levine 1984). In the abstract-based keyword
Research Approach
search, we used different combinations of seven key-
We used these two technological shifts to gain insight
words, namely, “catalytic converter,” “emission,” “auto-
into the subsequent division of innovative labor and mobile,” “catalyst,” “pollution,” “exhaust,” and “engine,”
thereby determine whether the evidence supports our to obtain an automobile emissions control patent set.
two sets of hypotheses (H1A/H1B, inclusionary shift Two searched patent sets from both methods were then
consistent with mirroring; and H2A/H2B, digital con- combined, and duplicate patents were eliminated. The
trol shift consistent with systems integration). To do this, abstract-based keyword search was helpful in identi-
we looked at all the patents associated with automo- fying potentially relevant patents not captured under
bile emissions control systems to see whether (1) the the class-based search method. Moreover, we screened
source of innovation was the OEM or a supplier, and out irrelevant patents by carefully reading abstracts of
whether (2) the specific type of change was compo- searched patents. Our purpose is to generate a cleaner
nent or architectural. In analyzing the data, we used patent set by identifying the patents that belong to the
logistic regression analysis to isolate the effects of the search criteria but are not necessarily related to the
technological change on the overall division of inno- technology of interest. For example, a patent for cat-
vative labor between automotive OEMs and suppli- alyst pollution control technology specifically designed
ers. Both groups continually engage in both forms of for an electric power plant could belong to the same
innovation (component and architectural). Therefore, we patent class as other catalyst patents granted for automo-
endeavored to measure the effect of the overall proclivity bile applications. After screening the searched patents,
for these groups to engage in different forms of innova- we generated a time-series patent data set on auto-
tion in the wake of a technology shift. Much of the data mobile emissions control technologies from 1970 to
were drawn from an in-depth patent search. In the fol- 1998. We focused our search on 237 firms that received
lowing, we describe this patent search and the measures 2,100 patents, accounting for approximately 93% of the
used in the analysis. overall patenting activities associated with automotive
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1436 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

emission controls from 1970 to 1998. Next we describe studies on patents can provide detailed and consistent
the variables used in our analysis. information of how firms solve technological problems
especially when patent statistics are combined with qual-
Dependent and Explanatory Variables itative knowledge of technology.
Innovation. We used the firms’ successful U.S. patent Innovation Type. We developed a category for inno-
applications in automobile emissions control technology vation to distinguish its types, namely, architectural and
as a measure of innovation activities. The use of patents component. We coded the patents using the generic def-
initions of component and architectural innovation pro-
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

has various limitations (Acs et al. 2002, Archibugi and

Pianta 1996, Basberg 1987, Griliches 1990, Lanjouw vided by Henderson and Clark (1990; see also Ulrich
et al. 1998, Popp 2005, Scherer 1983). Not all inven- 1995). The patents that capture the identifiable phys-
tions are patented, and the proportion and quality of ical part of the product represent component innova-
inventions patented may vary depending on the sector tion, whereas those that address the interfacial problems
and the period being investigated (e.g., Basberg 1987, among components and how the components can be
Lanjouw et al. 1998, Popp 2005). In some instances, linked together into a whole system were coded as archi-
firms apply for patents only to prevent other firms from tectural innovation.2 Table 1 shows the key features of
“inventing around” a patented invention or exploiting the patent categorization and relevant patent classes for
the patents (Cohen 1995, Pavitt 1985). Moreover, high automobile subsystems.
variances and skewing exist in the distribution of patent Inclusionary and Digital Control Product Hierarchies.
values (Griliches 1990, Lanjouw et al. 1998, Scherer To capture the effect of a technological shift in the inclu-
and Harhoff 2000). The propensity to patent also varies sionary product hierarchy (catalytic converter in 1981) as
across industries and countries (Basberg 1987, Cohen well as in digital control technology (electronic onboard
et al. 2000, Pavitt 1985). However, with these limitations diagnostic module in 1994) on automakers’ propensity
in mind, there exists a substantial amount of precedent for architectural (or component) innovation, we defined
for using patents as an operationalizable proxy for inno- two periods: the inclusionary hierarchy dominant design
vation (e.g., Griliches 1990, Lanjouw et al. 1998, Popp period (1981–1993) and the digital control hierarchy
2003). Patent statistics represents a “unique resource technology-shift period (1994–1998). Because innova-
for the analysis of the process of technical change,” tion in auto emissions control regulation was driven by
providing an abundant data with potential industrial, a government regulation enacted in 1970 (i.e., the Clean
organizational, and technical details (Griliches 1990, Air Act of 1970), we defined the inclusionary hierar-
Lanjouw et al. 1998). Pavitt (1985) also claimed that chy dominant design era as the period that captures the

Table 1 Automotive Emission Control Subsystems (AECS) and Relevant Patent Classes

AECS subsystema Classes Example technologies Innovation-type category

Catalyst 422/171 Porous catalyst carrier, catalyst support materials, Component

Palladium TWC catalyst
422/177 Catalyst support design, absorber system, catalyst Component
support materials
422/213.2 Catalyst materials, wash-coat materials Component
Electronics 60/274 Air-to-fuel ratio control, determining efficiency of cat- Architectural
alytic converter, electronic heating control
60/276 Air-to-fuel ratio control, determining efficiency of cat- Architectural
alytic converter
60/277 Determining efficiency of catalytic converter Architectural
60/278 EGR control, spark timing control, absorber system Architectural
337/382 Temperature sensor Component
Manufacturing 29/890 Assembly method for catalytic converter, catalytic Architectural
converter housing design, manufacturing process
422/179 Catalytic converter housing design, mounting materi- Architectural
als design
422/176 Catalyst support design, catalyst converter housing Architectural
design
425/461 Extrusion die for substrates Component
422/174 Heated cellular substrates Component
a
We further disaggregated the automobile emission control technology patents into three subsystem technology
categories: catalyst, electronics, and manufacturing technologies to help the reader understand better how the three
subsystem technologies comprising the automobile emission control technology are related to the innovation cate-
gories.
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1437

date of the enactment of the technology-forcing auto Figure 5 Proportion of Digital Electronics and
emissions control government regulation (i.e., the Clean Nondigital Technology
Air Act of 1970) until the emergence of the digital Electronics (digital)
control technology in 1994 (1970–1993). We name the Catalyst
0.7
period following the emergence of the digital control Manufacturing

Proportion of technology
technology as the digital control hierarchy technology 0.6
shift era (1994–1998). We then generated interaction
0.5
terms between these two dominant design periods with
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

the automaker dummy to examine whether automak- 0.4

ers’ propensity for different types of innovation (i.e.,
0.3
architectural and component) is related to the two dif-
ferent technological shifts (i.e., inclusionary or control 0.2
hierarchy).
0.1
To verify 1981 as the point of stabilization in the
inclusionary hierarchy of emission control systems, we 0
examined the degree to which the technology was self- 1970 1975 1980 1985 1990 1995
Year
contained at that point. According to Campbell and
Levine (1994), as the technology is being developed, the
degree to which the technology is self-contained—that is, Furthermore, we looked to empirically verify the
the proportion of patent citations given to prior technol- 1994 start to the second era, following the architec-
ogy from its own technology domain—increases until the tural innovation in the digital control hierarchy. We
technology becomes commercialized. The timing of an examined how the proportion of digital electronics tech-
emergence of a dominant design is closely related to the nologies compared with that of nondigital technolo-
commercialization of a new technology (Anderson and gies (i.e., catalyst and manufacturing) over time (see
Tushman 1990). Therefore, what Campbell and Levine’s Figure 5). As expected, we found that the proportion
(1994) study implies is that the timing at which the tech- of digital electronics technology becomes higher than
nology self-containment approaches its maximum value that of nondigital technologies starting in 1993. Prior
captures the emergence of the dominant design. We esti- to 1993, the proportion of digital electronics technol-
mated and plotted technology self-containment over time ogy remained lower than that of nondigital technologies.
and found that the technology self-containment measure Again, we marked the beginning of the “era” following
increased until 1980 and then leveled off (see Figure 4). the architectural innovation in the digital control hierar-
This finding strongly supports the use of 1981 as the first chy in 1994.
year of the “era” following the establishment of a dom-
inant design. When a dominant design emerged in auto Control Variables
emissions control technology—that is, a catalytic con-
verter based on three-way catalysts—nearly 70 % of new Regulatory Compliance Expenditure. We used the
cost estimates for automobile emission control devices
vehicles were equipped with this device in 1981 (Gerard
to control for the effect of regulatory pressures on the
and Lave 2005).
development of automobile emission control technolo-
gies. We applied a one-year lag structure to the regula-
Figure 4 Technology Self-Containment from 1970 to 1998 tory compliance expenditure variable (Jaffe and Palmer
50
1997). We used a number of different sources, such as
the EPA (1990) and the California Air Resource Board
45
reports, to compile the cost data for emission control
40
Percentage of technology

devices (CARB 1996). The EPA’s (1990) cost estimate

35 provides aggregated capital cost estimates for emissions
self-containment

30 control devices, including evaporative emissions canis-

25 ters, high-altitude emissions control, catalytic converters,
20 exhaust gas recirculation units, and air pump units (EPA
15
1990, McConnell et al. 1995).
10 Auto Sales. We included the U.S. retail automobile
5 sales to control for the potential market force in the
0 development of emission control devices. Ward’s publi-
1970 1975 1980 1985 1990 1995 cation on motor vehicle facts and figures was our source
Year for the U.S. automobile retail sales data.
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1438 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

Automotive Patents. We included total patent counts In the fixed-effects analysis, we focused on the top 10
in automotive technology to control for the potential patenting firms, including five automakers, namely, Gen-
impact of innovation trend in the automobile industry eral Motors, Toyota, Ford, Nissan, and Honda; and five
on innovation in automobile emission control technolo- suppliers, namely, Engelhard, Corning, Nippondenso,
gies. To obtain the overall patent counts in automo- EMITEC, and NGK Insulator. We selected the top 10
tive technology, we used the U.S. Patent Classification patenting firms for the fixed-effect regressions because
(USPC), selected subclasses listed under “Automobile” firms’ patenting activity is skewed. For example, among
in the USPC index and Class 180 (Motor Vehicles), the 237 firms that patented, these top 10 patenting firms
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

and counted patents applied under these subclasses from accounted for about 51% of the total patenting activities
1970 to 1998. (1,065 patents),4 and about 54% of firms (130 firms) had
fewer than 10 patents over the 30-year period. Thus, we
Technology Self-Containment. We control for the
dependency of automobile emission control technology reduced the risk of obtaining inconsistent fixed-effects
on other fields. According to Campbell and Levine estimates by using the top representative patenting firms’
(1994), the proportion of citations given to other techni- data (Katila and Chen 2008).5 We ran logistic regres-
cal fields decreases as the technology evolves until the sions for each of the two distinct time periods and com-
timing of commercialization. We calculated the changes pared the results. For architectural innovation in the
in the degree of technology self-containment over time inclusionary hierarchy, we used the time frame between
by examining the percentage of citations given to previ- 1970 and 1993. In the digital control hierarchy analysis,
ously granted patents in the automobile emission control we used the period from 1981 to 1998.
technology.
Total Patents. We included a general economy-wide Results
patenting activity trend variable using the applied total Table 2 presents the descriptive statistics and correlation
patent counts using the USPTO database from 1970 matrix for the variables used for the analysis, indicating
to 1998. that some variables are highly correlated. High correla-
Technology Choice. We control for the functional tions among variables cause significant muliticollinear-
attributes of the three main subsystems of automo- ity and may bias the regression estimates (Kennedy
tive control technology (see Table 1), namely, catalyst, 1985). This study approaches these roblems of muliti-
electronics, and manufacturing. The innovation-type collinearity by comparing the nested regression mod-
variable (component and architectural innovation) cap- els. This approach has been found to be useful in other
tures the physical attributes of technology, whereas this research on ecological dynamics (Baum et al. 1995,
technology-choice variable captures the technology’s Tucker et al. 1990).61 7
functional attributes. Thus, the innovation-type variable Table 3 reports the results of eight logistic regres-
is not confounded by the technology-choice variable. sions on the component innovation. Models 1–6 present
the results of the pooled logistic regressions, and mod-
Data Analysis els 7 and 8 show the results of the firm fixed-effects
To analyze the relative propensity for OEMs to engage logistic regression. Models 1 and 2 add the main effect
in architectural and component innovation following a “technological shift” variables—that is, the inclusionary
technological shift, we fitted logistic regressions where hierarchy (‚1 5 and the digital control (‚2 5—to the con-
the innovation-type variable (i.e., component and archi- trol variables, respectively. Model 1 indicates that the
tectural innovation were coded as 1 and 0, respectively) coefficient for the inclusionary hierarchy variable is pos-
was used as a dependent variable. Because there are itive (‚1 1 0044) but insignificant. In contrast, model 2
multiple observations for each firm, observations within indicates that the coefficient for the digital control vari-
firms may not be independent. To address this potential able is negative and significant (‚2 1 −1055). These find-
problem, we reported the results with robust standard ings weakly indicate that different levels of technological
errors using the Huber–White estimator for variance. shifts have differing effects on firms’ overall propen-
This methodological approach relaxes the assumption sity for component innovation. The technology shift in
that observations within firms are independent (White the inclusionary hierarchy increases the likelihood of
1980). We also performed firm fixed-effect logit regres- the overall component innovation, whereas the technol-
sions to build more robustness into our analysis.3 The ogy shift in the digital control hierarchy decreases the
advantage of adopting firm fixed effects is that they likelihood of component innovation. As such, this may
control for firm-associated unobservable heterogeneities increase the likelihood of architectural innovation. Com-
arising from firms’ potential differences in strategies and ponent innovation is emphasized more following the
capabilities (Penner-Hahn and Shaver 2005), as well as technology shift in the inclusionary hierarchy, as shown
the different propensities for patenting (Pavitt 1985). by a positive coefficient in model 1 (‚1 5. The negative
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1439

Table 2 Descriptive Statistics and Correlations

Variables Mean SD 1 2 3 4 5 6 7 8 9 10 11

1. Inclusionary hierarchy 0034 0047

dummy (1981–1993)
2. Digital control hierarchy 0040 0049 −0057∗
dummy (1994–1998)
3. Innovation typea 0058 0049 0016∗ −0005∗
4. Automaker dummy 0045 0050 −0002 0007∗ −0024∗
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

5. Catalyst technology 0033 0047 0008∗ −0016∗ 0058∗ −0016∗

6. Electronics technology 0038 0049 −0006∗ 0022∗ −0054∗ 0023∗ −0056∗
7. Manufacturing technology 0028 0045 −0002 −0007∗ −0003 −0008∗ −0045∗ −0050∗
8. Regulatory compliance 648057 268070 0024∗ 0059∗ 0008∗ 0010∗ −0012∗ 0020∗ −0009∗
expenditureb
9. Auto sales (106 ) 6076 1020 −0034∗ −0040∗ −0006∗ −0007∗ 0012∗ −0017∗ 0005∗ −0074∗
10. Patenting in automotive 1021 0035 −0021∗ 0087∗ 0002 0007∗ −0015∗ 0021∗ −0007∗ 0075∗ −0060∗
technology (103 )
11. Technology self-containment 0035 0009 −0006∗ 0066∗ 0006∗ 0011∗ −0011∗ 0016∗ −0006∗ 0084∗ −0053∗ 0072∗
12. Total patenting (106 ) 0017 0005 −0012∗ 0082∗ 0000 0008∗ −0018∗ 0025∗ −0008∗ 0080∗ −0073∗ 0096∗ 0071∗

Note. n = 21100.
a
Component innovation = 1 and architectural innovation = 0.
b
Per vehicle (US$).
∗
p < 0005.

Table 3 Results of the Logistic Regression on Component Innovation, 1970–1998

Pooled Fixed effects

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

Variables 1970–1993 1981–1998 1970–1993 1981–1998 1970–1993 1981–1998 1970–1993 1981–1998

‚1 Inclusionary hierarchy 0044 1012† 1094∗∗ 4012∗

(1981–1993) 400595 400655 400575 410785
‚2 Digital control hierarchy −1055∗∗ −1086∗∗ −1080∗∗ −0083
(1994–1998) 400545 400575 400575 400905
‚3 Inclusionary hierarchy −1031∗∗ −1028∗∗ −2069†
× Automaker dummy 400395 400395 410555
‚4 Digital control hierarchy 0088∗∗ 0088∗∗ 0006
× Automaker dummy 400295 400295 400515
‚5 Automaker dummy −0014 −1042∗∗∗ −0015 −1043∗∗∗
400315 400235 400325 400235
‚6 Catalyst technology 5000∗∗∗ 2056∗∗∗ 4098∗∗∗ 2059∗∗∗ 5000∗∗∗ 2060∗∗∗ 4007∗∗∗ 2077∗∗∗
400425 400335 400435 400335 400435 400335 400525 400445
‚7 Electronics technology −2009∗∗∗ −2023∗∗∗ −1097∗∗∗ −2013∗∗∗ −1097∗∗∗ −2011∗∗∗ −1088∗∗∗ −1089∗∗∗
400215 400155 400225 400165 400235 400155 400315 400235
‚8 Regulatory compliance 1013 12061∗ 1014 10057† 0053 7057 −2011 3061
expenditure (1/103 5 410015 450395 410035 450535 410065 460465 410675 470435
‚9 Auto sales (1/106 5 −0009 −0001 −0008 0003 0005 0012 0040† 0029
400115 400165 400115 400175 400125 400255 400195 400245
‚10 Automotive patents (1/103 5 2091∗∗ −0056 2081∗∗ −0035 3064∗∗∗ 1040†
400895 400695 400925 400715 400965 400855
‚11 Technology self-containment 3030† 3058 3027† 2096 4009∗ 2072 6021 −2019
410795 430205 410835 430345 410875 430345 430505 440905
‚12 Total patents (1/106 5 1001† 1060
450485 460405
‚o Intercept −3096∗∗∗ −8054∗ −3089∗∗ −6066∗ −3092∗∗ −5058†
410145 430605 410185 430705 410415 430445
N 1,269 1,538 1,269 1,538 1,269 1,538 634 834
Log-likelihood −402034 −649040 −381029 −624014 −384040 −624023 −180001 −297019
Model • 2 904055∗∗∗ 760021∗∗∗ 946065∗∗∗ 810073∗∗∗ 940043∗∗∗ 810054∗∗∗ 333059∗∗∗ 303090∗∗∗

Notes. Dependent variable: Innovation type (component innovation). Robust standard errors are in parentheses.
†
p < 001; ∗ p < 0005; ∗∗ p < 0001; ∗∗∗ p < 00001.
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1440 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

coefficient for the technological shift in the digital con- on architectural innovation following the shift of the dig-
trol hierarchy in model 2 (‚2 5 denotes the firms’ overall ital control hierarchy, as shown by coefficients ‚1 and
propensity for architectural innovation decreased. ‚2 in models 1 and 2, respectively.
The catalyst and electronics technology regression The interaction between the inclusionary hierarchy
coefficients (i.e., ‚6 and ‚7 5 are positive and negative, variable and the automaker dummy (‚3 5 in model 3 rep-
respectively, in all models (models 1–7). These regres- resents the differential effect of the technological shift
sion findings largely capture the component nature of the on automakers’ innovation in relation to that of the
catalyst and architectural nature of the electronics emis- suppliers rather than the effect solely on automakers’
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

sion control technologies. The coefficient for U.S. retail innovation9 (Yip and Tsang 2007). Similarly, the interac-
automobile sales (‚9 5 suggests that automobile sales did tion term between the technological shift of the control
not have significant influence on firms’ propensities for hierarchy and the automaker dummy (‚4 5 in model 4
either component or architectural innovation. Moreover, represents a differential effect of the shift in the digital
regulatory compliance expenditure (‚8 5 is only weakly control hierarchy on automakers’ innovation over that of
related to firms’ propensity for component (or archi- the suppliers rather than the sole effect of the technol-
tectural) innovation because its impact is found to be ogy shift of the digital control hierarchy on automakers’
weakly significant only during the period of 1981–1998. innovation. Thus, the effect of the inclusionary dominant
The regression coefficient for the overall patenting trend design on automakers’ component innovation is −0009
in the automotive technology (‚10 , Model 1) is positive rather than −1031. With the component innovation as
and significant during the period of 1970–1993.8 The the dependent variable, the negative coefficient for the
firms’ propensity for component innovation in automo- interaction term (−0009) suggests that automakers had
bile emissions control technology is significantly related higher propensity for architectural innovation following
to the overall patenting trends in the automotive tech- a technological shift in the inclusionary hierarchy. Sim-
nology until the early 1990s. The regression coefficient ilarly, the effect of the shift in the digital control hier-
for the technology self-containment variable was found archy on automakers’ component innovation is −0098
to be positive and significant during 1970–1993 but not
rather than 0.88, implying that automakers also focused
during 1981–1998 (models 1–6). This implies that the
more on architectural innovation following the emer-
firms’ R&D decisions on the development of technol-
gence of a digital control hierarchy in auto electronics in
ogy during the period of 1970–1993 were related to the
1994. However, because of the nonlinear nature of the
rapidly established body of developing knowledge about
logit model, interpreting the interaction effect by sim-
automotive emissions control.
ply relying on the significance and signs of the logic
In models 3 and 4, we added the interaction of the
coefficients could be misleading (Bowen and Wiersema
dominant design and the automaker dummy (‚5 5 to
models 1 (‚3 5 and 2 (‚4 5, respectively. Regression coef- 2004, Hoetker 2007). Rather, it would be more infor-
ficients for the automaker dummy are negative, which mative and appropriate to estimate the marginal effects
indicates that automakers are mainly involved in archi- of interaction terms by setting meaningful values for the
tectural innovation rather than component innovation. other variables in the model (Hoetker 2007). To esti-
Moreover, the dominant design coefficients for both the mate the marginal effects of the hierarchical changes on
inclusionary hierarchy (‚1 5 and digital control hierar- the division of innovative labor, we set the automaker
chy (‚2 5 in models 3 and 4 measure the effect of the dummy to zero and one, respectively, while constraining
inclusionary and the digital control hierarchy architec- the other variables at their mean values. The marginal
tural innovations for the base (suppliers) rather than effects analysis indicated that the technology shift of
the effect of different levels of architectural innova- the inclusionary hierarchy increased the probability of
tion on the overall component innovation. Thus, the suppliers’ component innovation and automakers’ archi-
positive coefficient for inclusionary hierarchy dominant tectural innovation by 13.8% and 27.6%, respectively,
design dummies (‚1 1 10125 in model 3 indicates that confirming the interpretations based on the significance
the architectural innovation in the inclusionary hierar- and signs of the interaction terms (see model 3). Simi-
chy increased the probability for suppliers’ involvement larly, the marginal effect analysis also confirmed that the
in component innovation. The negative coefficient for emergence of a digital control hierarchy increased the
the shift in the digital control hierarchy (‚2 1 −1086) in probability of suppliers’ architectural innovation (‚2 ; see
model 4 suggests that suppliers paid relatively more model 4). However, the marginal analysis indicated that
attention to architectural innovation after the technolog- the architectural innovation associated with the digital
ical shift in the digital control hierarchy. The division control hierarchy increased the probability of automak-
of innovative labor in the 1980s and 1990s indicates ers’ component innovation by 8.9%. The results of the
that suppliers focused increasingly on component inno- marginal effect analysis are summarized in Table 4.
vation following the technological shift in the inclusion- In models 5 and 6, we replaced the automotive patent
ary product hierarchy and then focused relatively more control variable (Automotive patents) with a general
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1441

Table 4 Propensity for Component (Architectural) Innovation not perfect substitutes for many forms of advanced tech-
by Automakers and Suppliers After Architectural Inno- nological knowledge (Steinmueller 2003). Large OEMs
vation in Inclusionary and Digital Control Hierarchies need to maintain coordinating, integrative knowledge in
Automakers (%) Suppliers (%) order to produce their products, and researchers in this
tradition offer a variety of explanations for when this
Inclusionary hierarchy
(1970–1993) (27.6) 13.8
is required (see Prencipe et al. 2003). Just as the work
Digital control hierarchy on modularity provided an explanatory mechanism for
(1980–1998) 8.9 (39.7) why suppliers are increasingly engaged in innovation,
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

in this paper we offer an explanatory mechanism for

why OEMs need to maintain competencies beyond the
economy-wide patenting (Total patents) to test the sen- products that they manufacture: the widespread emer-
sitivity of the regression models 3 and 4, respectively. In gence of digital control systems in contemporary com-
models 7 and 8, we tested the robustness of the regres- plex products.
sion results of models 3 and 4 against the unobserved Although our explicit emphasis on digital controls
heterogeneities by performing firm fixed-effects logit is new to organizational scholarship, examples of digi-
regressions.10 The results of regressions and marginal tal controls abound in the research. In particular, work
effect analyses performed following models 5–8 consis- that has questioned the modularity explanation analyzes
tently supported the major findings of this study.11 contexts where digital controls take a central role. In
After a technological shift in the inclusionary hierar- domains as diverse as aircraft engine design (Brusoni
chy, automaker OEMs and component suppliers focus
et al. 2001, Sosa et al. 2004), computer design (Lan-
more on architectural and component innovation, respec-
glois 1997), and chip design (Ernst 2005), each has
tively, which supports H1A and H1B (in accordance
found that the mirroring hypothesis does not seem to
with the mirroring hypothesis). However, with regard
apply. In all cases they argue that the division of inno-
to technological shift in the digital control hierar-
vative labor does not involve delegation of component
chy, automaker OEM propensity for component inno-
knowledge to suppliers because OEMs need to main-
vation increases together with the supplier propensity
tain integrative and coordinating competencies in order
for architectural innovation, supporting H2A and H2B
to bring these complex products to market. Our expla-
(in accordance with the systems integration perspective).
nation extends this observation. Because digital controls
Consequently, the “mirroring hypothesis” seems to hold
are increasingly important to complex products, it is pre-
for technological shifts associated with the inclusionary
cisely these digital controls that address much of that
hierarchy, whereas the “systems integration” perspec-
coordinating and integrating work within the complex
tive holds for technological shifts in the digital control
products (Drouin et al. 1991). Furthermore, digital con-
hierarchy.
trols have an unpredictable, continuously evolving aspect
that is not present in traditional electromechanical prod-
Discussion uct structures.
In their seminal work on the division of innovative Because of this unpredictable, continuously evolving
labor, Arora and Gambardelli (1994) indicated that as aspect of digital controls, complex product structures are
knowledge associated with product innovation becomes never completely stable from a functional perspective—
increasingly abstract and generally applicable, barriers including situations where there are stable modular inter-
to innovation become dramatically reduced and more faces between the physical components of the digital
suppliers will innovate. This knowledge generalization control system. Even if sensors and local control units
can, in part, be facilitated by product modularity. Stan- may be stabilized over a period of time, there are
dardized interfaces between modules substitute for inte- two points of continuous change that persist in a dig-
grative, coordinating knowledge and offer a stabilized, ital control hierarchy that are typically not evident in
explicit form of “advanced technological knowledge” a traditional electromechanical inclusionary hierarchy.
(Sanchez and Mahoney 1996) to which suppliers can The first involves software. Because digital controls are
build competencies. Thus, the work on modularity fur- programmable, there is the ever-present opportunity to
ther fulfills Arora and Gambardelli’s thesis by offer- change and improve controls—even after the complex
ing an explanatory mechanism by which knowledge can systems are in use (e.g., firmware updates). Second,
become generalized. there is always the possibility that integrative control
Following this work on modularity and the division of units will be added further up the hierarchy that coordi-
innovative labor, Brusoni et al. (2001, p. 597) indicated nate across ever-broader reaches of the complex system
that even for modular products, firms “know more than and integrate this complex system with other complex
they make.” Firm competencies, especially on the part of systems. In this sense, over time digital control sys-
large OEMs, go beyond the products they actually man- tems become increasingly decoupled from the compo-
ufacture, indicating that modular product interfaces are nents they control (Findeisen et al. 1980).
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1442 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

As evolving decision systems (Mesarovic et al. 1970) (e.g., Baldwin 2008) and the systems integration per-
that are simultaneously decoupled from the inclusion- spective (e.g., Prencipe et al. 2003) can be reconciled.
ary hierarchy, yet also work to integrate that hierarchy, Although these two views are often thought to be con-
digital controls offer insight into the management and tradictory (Brusoni 2005), our research shows that they
design of complex systems. One of the core precepts are both relevant, given a dual-product hierarchy view
of complex systems theory is Ashby’s (1958) law of of complex products.
requisite variety, which in essence states that the con-
troller must have as much variety (or be as complex) Implications for Practice
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

as the system it controls in order to control that sys- The practical implications of this research involve clar-
tem adequately. Following Simon (1996), much of orga- ifying key capabilities that different firms might nur-
nizational scholarship is based on the observation that ture given the increased role of digital controls in
design complexity can be managed through partially innovative activity around complex products. First, it
decomposable subsystems, which involves reducing the is important to note that design practices associated
complexity of the system under control by decompos- with digital technologies are fundamentally different
ing that system into manageable modules (see Baldwin than those associated with traditional electromechan-
and Clark 2000). However, with this focus on simplify- ical complex products (Turner 1987, Whitney 1990),
ing products, organizational scholars overlook the other and certain organizations, particularly OEMs, may wish
strategy for attaining requisite variety: by increasing the to proactively develop capabilities in software develop-
complexity of the controller. With the radical innova- ment. Product design and development firms rooted in
tion in speed, size, and reliability, digital systems can traditional electromechanical logic follow linear or cas-
now dramatically expand the scope of automated con- cading development methodologies in designing com-
trol (Pop et al. 2004). Large-scale integrative control plex physical systems (Ulrich and Eppinger 1995). In
activities that in the past existed only in theory can these firms, development problems typically involve
now be realized in practice (Warwick and Rees 1988). managerial solutions or changes to intrafirm or inter-
Moreover, continued advancements in embedded tech- firm structures (e.g., Clark and Fujimoto 1991). In con-
nology and related standards enable unprecedented lev- trast, software development requires firms to have the
els of data integration and convergence that can monitor technical knowledge about fast-paced digital innovations
and control increasingly diverse subsystems and compo- and about a wide array of application domains (Curtis
nents in real time (Yoo et al. 2010). These decoupled, et al. 1988). This process of integrating multiple appli-
integrative digital control hierarchies enable organiza- cation domains with diverse sets of knowledge are one
tions to control an unprecedented level of complexity. of the central issues that research on software develop-
Thus, to develop increasingly complex systems, organi- ment has explored for decades (e.g., Hirschheim et al.
zational researchers can also think in terms of increasing 1995). As such, the literature on software development
the complexity of the controller through digital controls may offer managers a number of ways to enhance their
(that is, in addition to decomposition). product design and development activities. For example,
These observations about digital controls may explain information systems scholars have detailed a number of
why the systems integration perspective (Prencipe et al. methods and notations for executing design projects cov-
2003) has gained so much attention in recent years. ering a range of topics specifically geared toward man-
In the past, the mirroring hypothesis (Brusoni 2008) aging complexity, including risk mitigation (Lyytinen
was accepted as fact—and this view held well enough et al. 1998), user involvement (Mumford 2003), itera-
because control systems were tightly coupled with the tive development (Berente and Lyytinen 2009), design
components that they controlled. A single-product hier- requirements (Hansen et al. 2009), and a variety of
archy was adequate for characterizing complex prod- notations including object-oriented techniques (Booch
ucts, and product components were largely integrated et al. 2007).
through their direct interfaces with each other. Only in In addition to building capabilities in domains such
recent decades have digital control technologies matured as software development, OEMs may wish to maintain
enough to begin transforming industries. Now, given the capabilities in component design in order to more effec-
broadly integrative capabilities associated with digital tively integrate components through digital control sys-
control hierarchies, researchers have begun documenting tems. Although modular architectures are intended, in
the transformations accompanying this revolution. This part, to relieve OEMs from the need to maintain certain
research provides a language for dealing with this par- component-level competencies, our research shows that
ticular form of digital innovation, as well as an initial this may not be the case to the degree that some schol-
explanation for the implications of digital control hier- ars have argued. Instead, this work adds to the body of
archies on interfirm innovation. More importantly, how- research that advises OEMs to maintain competencies in
ever, this research explains how the mirroring hypothesis certain key components (Christensen 2006).
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1443

Of course, our analysis highlights how the strate- established firms are disadvantaged when new techno-
gies associated with relative efforts in component versus logical changes have digital-type innovation because of
architectural innovation depend, to an extent, on an orga- established firms’ difficulty in switching to a new mode
nization’s position in the supply chain and the point in of learning that requires building new digital knowledge
the evolution of the relevant product structures. Just as on existing digital knowledge base. However, we found
OEMs may increase their component innovation to best that the OEMs led much of the digital control-related
capitalize on a technological shift in the digital control innovation. What does this mean to suppliers? Future
hierarchy, their suppliers may want to move upstream research might reexamine the thesis of Henderson and
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

and focus more on architectural innovation—even the Clark in the context of digital controls.
innovations associated with digital controls—in order to Future work should also address different system con-
position themselves for bundling multiple components texts in different industries and should include alternative
together. measures for innovation. Also, it would make sense to
Our research does not address which forms of innova- relate these measures to organizational performance over
tion firms should be pursuing and when, but we do high- time. Finally, some researchers have indicated that com-
light what they might expect before and after shifts in the petitive forces and the division of innovative labor do
inclusionary and digital control hierarchies. However, not play out in all complex product development efforts
based on our theoretical lens, when innovating around in the same way (Argyres and Bigelow 2010). Certain
the digital control hierarchy, OEMs need to maintain complex systems are not subject to Schumpeterian mar-
work across a variety of domains—digital technologies, ket forces and are coordinated through powerful actors
architectural, and component level—because digital hier- coming together to realize specific product designs in a
archies span all of these realms. As products continue to way that does not reflect traditional competitive dynam-
become increasingly complex and digital control hierar- ics (Miller et al. 1995). Therefore, one might expect
chies continue to expand, it will be critical for OEMs to the relationship between architectural innovation and the
continue aggressively innovating at the component level division of innovative labor to perhaps be different in
to be prepared to capitalize on the never-ending stream different contexts.
of opportunities associated with digital technologies.

Future Research Conclusions

Future research should do more to distinguish digital The digital revolution is transforming the way that
controls from other forms of digital innovation. In auto- firms innovate in a supply chain, and there is a grow-
motive contexts, for example, digital innovation is trans- ing body of organizational literature on the relation-
forming the driver and passenger experience through ship between information technologies and the interfirm
multimedia entertainment and communication systems process of innovation associated with complex products
(Leen et al. 1999). Digital controls are already beginning (e.g., Argyres 1999, Majchrzak et al. 2000, Yoo et al.
to integrate with entertainment systems and other elec- 2006, Boland et al. 2007). However, this is not the extent
tronic subsystems, and it will be important to determine of the transformational effect of the digital revolution on
when and how we might distinguish digital controls innovative interfirm practices. Complex products them-
from other digital systems. Beyond this, however, a vari- selves are increasingly made up of digital components,
ety of digital systems aid in integrating, coordinating, a trend that Yoo et al. (2010) refer to as “digital inno-
and controlling interfirm and intrafirm processes. For vation,” which emphasizes the digital elements of prod-
example, Brusoni (2005) looks at chemical processing uct innovation (in contrast to this work on information
as a (more or less) modular system, and the digital con- technology innovation that focuses on the process of
trols that span those modular practices may change the innovation). In recent years, digital product innovations
way such organizational practices are conceived. More have brought transformative change to a variety of indus-
research in the way different forms of digital controls tries, including the music, film, and automotive indus-
might integrate across products and interfirm practices tries (Yoo 2010). In this paper we add to the growing
is also in order. body of literature on digital technologies in products of
Furthermore, much of the research on the interfirm innovation and in the associated effects on organization
division of innovative labor generally positions the OEM (Yoo et al. 2010).
in a privileged role in supplier networks, and our theoriz- Specifically, we extend the body of literature that
ing is no different. Future work should focus more on the attends to the relationship between complex product
impact of digital control innovation from a supplier per- architectures and the division of innovative labor in the
spective. For example, our finding that automakers have supply chain. Although recent research does identify the
an central role in innovating after a technological shift importance of the digital revolution to the way firms
in the digital control hierarchy seems to contradict the maintain their capabilities, none of the organizational lit-
view of Henderson and Clark (1990). They argued that erature specifically distinguishes the important class of
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1444 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

11
digital technologies known as digital control systems. In Results of additional marginal effect analyses for models 5–8
emphasizing the importance of digital control systems in are available from the authors upon request.
our work, we conceive of complex product architectures
in terms of two product hierarchies: the inclusionary and References
the digital control hierarchy. Furthermore, we use this Abernathy, W. J., J. M. Utterback. 1978. Patterns of industrial inno-
distinction to reconcile two competing views about the vation. Tech. Rev. 80(7) 40–47.
relationship between complex product structures and the Acs, Z. J., L. Anselin, A. Varga. 2002. Patents and innovation counts
interfirm division of innovative labor. as measures of regional production of new knowledge. Res. Pol-
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

icy 31(7) 1069–1085.

Acknowledgments Almeida, P., B. Kogut. 1999. Localization of knowledge and the
The authors thank Ann Majchrzak, Dick Boland, Kalle Lyyti- mobility of engineers in regional networks. Management Sci.
nen, Youngjin Yoo, and three anonymous reviewers for their 45(7) 905–917.
invaluable direction in crafting this paper. They also appre- Anderson, P., M. L. Tushman. 1990. Technological discontinuities and
ciate the guidance and support of James Costantini, Susan dominant designs: A cyclical model of technological change.
Helper, John Leslie King, Joseph Mahoney, Anja Schultze, Admin. Sci. Quart. 35(4) 604–633.
Francisco Veloso, and seminar participants of the 2009 Inter- Archibugi, D., M. Pianta. 1996. Measuring technological change
national Motor Vehicle Program Automotive Research Confer- through patents and innovation surveys. Technovation 16(9)
ence. This research was partially funded by the International 451–468.
Motor Vehicle Program (http://www.imvpnet.org). Argyres, N. S. 1999. The impact of information technology on coor-
dination: Evidence from the B-2 “Stealth” bomber. Organ. Sci.
Endnotes 10(2) 162–180.
1
The general theory of Mesarovic et al. (1970) is rooted in Argyres, N., L. Bigelow. 2010. Innovation, modularity, and verti-
cybernetic system theory and is concerned with generic coor- cal deintegration: Evidence from the early U.S. auto industry.
dination principles intended to be applicable to any com- Organ. Sci. 21(4) 842–853.
plex system domain—from industrial and biological systems
Arora, A., A. Gambardella. 1994. The changing technology of techno-
to organizations and societies. This work has been influential
logical change: General and abstract knowledge and the division
in process control engineering, operations research, complex
of innovative labour. Res. Policy 23(5) 523–532.
systems, and computer science, and it has been leveraged
much less often in research in information systems, eco- Ashby, W. R. 1958. Requisite variety and its implications for the
nomics, and social and organizational theories. control of complex systems. Cybernetica 1(2) 83–99.
2
Coding for both innovation-type and technology-choice vari- Baldwin, C. K. 2008. Where do transactions come from? Modular-
ables was cross-checked by an independent coder. ity, transactions, and the boundaries of firms. Indust. Corporate
3 Change 17(1) 155–195.
We also ran count-based negative binomial fixed-effects
regressions to test the sensitivity of the model. The negative Baldwin, C., K. Clark. 2000. Design Rules, Volume 1: The Power of
binomial fixed-effects estimator confirmed the key results. The Modularity. MIT Press, Cambridge, MA.
results are available from the authors upon request. Basberg, B. L. 1987. Patents and the measurement of technological
4
Patent counts for the individual top patenting firms are avail- change: A survey of the literature. Res. Policy 16(2–4) 131–141.
able from authors upon request.
5
This research methodological approach of selecting a repre- Baum, J. A. C., H. J. Korn, S. Kotha. 1995. Dominant designs and
population dynamics in telecommunications services: Found-
sentative subset of a population, notably the major firms’ data
ing and failure of facsimile transmission service organizations,
for analysis, is widely adopted in patent data-based studies
1965–1992. Soc. Sci. Res. 24 97–135.
(e.g., Rosenkopf and Nerkar 2001, Almeida and Kogut 1999).
6
We used the Stata command by Collin to calculate the vari- Berente, N., K. Lyytinen. 2009. Iteration in systems analysis and
ance inflation factor for independent variables; none of them design: Cognitive processes and representational artifacts. R. H.
was higher than 10. L. Chiang, K. Siau, B. C. Hardgrave, eds. Systems Analysis and
7 Design: Techniques, Methodologies, Approaches, and Architec-
The relatively high correlation between overall patenting in
tures. M. E. Sharpe, New York, 37–57.
automotive technology and digital control hierarchy suggests
that the digital revolution in the 1990s may account for the Boland, R. J., Jr., K. Lyytinen, Y. Yoo. 2007. Wakes of innovation
nontrivial influence on patenting activities in automotive tech- in project networks: The case of digital 3-D representations in
nology. Correlation is also relatively high between regulatory architecture, engineering, and construction. Organ. Sci. 18(4)
631–647.
expenditure and technology self-containment. This relatively
high correlation supports the idea that high regulatory strin- Booch, G., B. Maksimchuk, M. Engle, B. Young, J. Conallen,
gency may have influenced the innovation in auto emissions K. Houston. 2007. Object-Oriented Analysis and Design with
control. Applications. Addison-Wesley, Upper Saddle River, NJ.
8
A similar result was observed for model 5 with the total Bowen, H., M. Wiersema. 2004. Modeling limited dependent vari-
patent control variable (‚11 5. ables: Methods and guidelines for researchers in strategic man-
9 agement. D. Ketchen, D. Bergh, eds. Research Methodology in
‚3 = (Effect on automakers)−‚1 in model 3, and ‚4 = (Effect
on automakers) − ‚2 in model 4. Strategy and Management. Elsevier, New York, 87–134.
10
The OEM dummy variable is subsumed within the firm fixed Brusoni, S. 2005. The limits to specialization: Problem solving
effects and is therefore not shown in the specification for mod- and coordination in “modular networks.” Organ. Stud. 26(12)
els 7 and 8. 1885–1907.
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1445

Brusoni, S., A. Prencipe. 2006. Making design rules: A multidomain Grandstrand, O., P. Patel, K. Pavitt. 1997. Multitechnology corpora-
perspective. Organ. Sci. 17(2) 179–189. tions: Why they have “distributed” rather than “distinctive core’
Brusoni, S., A. Prencipe, K. Pavitt. 2001. Knowledge specialization, capabilities.” Calif. Management Rev. 39(4) 8–25.
organizational coupling, and the boundaries of the firm: Why Griliches, Z. 1990. Patent statistics as economic-indicators: A survey.
do firms know more than they make? Admin. Sci. Quart. 46(4) J. Econom. Literature 28(4) 1661–1707.
597–621.
Handfield, R. B., G. L. Ragatz, K. J. Peterson, R. M. Monczka. 1999.
California Air Resource Board. 1996. Low-Emission Vehicle and Involving suppliers in new product development? Calif. Man-
Zero-Emission Vehicle Program Review. Report, Mobile Divi- agement Rev. 42(1) 59–82.
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

sion, CARB, Sacramento, CA.

Hansen, S., N. Berente, K. Lyytinen. 2009. Requirements in the
Campbell, R. S., L. O. Levine. 1984. Technology Indicators Based 21st century: Current practice and emerging trends. K. Lyyti-
on Patent Data: Three Case Studies. Report, Battelle Pacific nen, P. Loucopoulos, J. Mylopoulos, B. Robinson, eds. Design
Northwest Laboratory, Richland, WA. Requirements Engineering: A Ten-Year Perspective. Lecture
Christensen, J. F. 2006. Wither core competency for the large corpo- Notes in Business Information Processing, Vol. 14. Springer-
ration in an open innovation world? H. Chesbrough, W. Van- Verlag, Berlin, 44–87.
haverbeke, J. West, eds. Open Innovation: Researching a New Helper, S., J. P. MacDuffie, C. Sabel. 2000. Pragmatic collaborations:
Paradigm. Oxford University Press, Oxford, UK, 35–61. Advancing knowledge while controlling opportunism. Indust.
Clark, K. B. 1989. Project scope and project performance: The effect Corporate Change 9(3) 443–488.
of parts strategy and supplier involvement on product develop- Henderson, R. M., K. B. Clark. 1990. Architectural innovation: The
ment. Management Sci. 35(10) 1247–1263. recognition of existing product technologies and the failure of
Clark, K. B., T. Fujimoto. 1991. Product Development Performance: established firms. Admin. Sci. Quart. 35(1) 9–30.
Strategy, Organization, and Management in the World Auto Henfridsson, O., L. Mathiassen, F. Svahn. 2009. Reconfiguring mod-
Industry. Harvard Business School Press, Boston. ularity: Closing capability gaps in digital innovation, sprouts.
Cohen, W. M. 1995. Empirical studies of innovative activity. P. Stone- Sprouts: Working Papers Inform. Systems 9(22) http://sprouts
man, ed. Handbook of the Economics of Innovation and Techni- .aisnet.org/9-22.
cal Change. Basil Blackwell, Oxford, UK, 182–264. Hirschheim, R. A., H. K. Klein, K. Lyytinen. 1995. Information
Cohen, W. N., R. R. Nelson, J. P. Walsh. 2000. Protecting their intel- Systems Development and Data Modeling: Conceptual and
lectual assets: Appropriability conditions and why U.S. man- Philosophical Foundations. Cambridge University Press, Cam-
ufacturing firms patent (or not). NBER Working Paper 7522, bridge, UK.
National Bureau of Economic Research, Cambridge, MA.
Hoang, H., F. Rothaermel. 2010. Leveraging internal and external
Curtis, B., H. Krasner, N. Iscoe. 1988. A field study of the soft- experience: Exploration, exploitaion, and R&D project perfor-
ware design process for large systems. Comm. ACM 31(11) mance. Strategic Management J. 31(7) 734–758.
1268–1287.
Hobday, M. 1998. Product complexity, innovation and industrial orga-
Davies, A. 2003. Integrated solutions: The changing business of nization. Res. Policy 26(6) 689–710.
systems integration. A. Prencipe, A. Davies, M. Hobday, eds.
Hoetker, G. 2007. The use of logit and probit models in strategic
The Business of Systems Integration. Oxford University Press,
management research: Critical issues. Strategic Management J.
Oxford, UK, 333–368.
28(4) 331–343.
Dorf, R. C., R. H. Bishop. 2008. Modern Control Systems, 9th ed.
Pearson Prentice Hall, Upper Saddle River, NJ. Inkpen, A. C., A. Dinur. 1998. Knowledge management processes and
international joint ventures. Organ. Sci. 9(4) 454–468.
Dosi, G., M. Hobday, L. Marengo, A. Prencipe. 2003. The eco-
nomics of system integration: Toward an evolutonary interpre- Jaffe, A. B., K. Palmer. 1997. Environmental regulation and innova-
tation. A. Prencipe, A. Davies, M. Hobday, eds. The Business tion: A panel data study. Rev. Econom. Statist. 79(4) 610–619.
of Systems Integration. Oxford University Press, Oxford, UK, Jurgen, R. K. 1999. Automotive Electronics Handbook. McGraw-Hill,
95–113. New York.
Drouin, M., H. Abou-Kandil, M. Mariton. 1991. Control of Complex Katila, R., E. Chen. 2008. Effects of search timing on innovation: The
Systems: Methods and Technology. Planum Press, New York. value of not being in sync with rivals. Admin. Sci. Quart. 53(4)
Ernst, D. 2005. Limits to modularity: Reflections on recent develop- 593–625.
ments in chip design. Indust. Innovation 12(3) 303–335. Kennedy, P. 1985. A Guide to Econometric Methods, 2nd ed. MIT
Findeisen, W., F. N. Bailey, M. Bryds, K. Malinowski, P. Tatjewski, Press, Cambridge, MA.
A. Wozniak. 1980. Control and Coordination in Hierarchical Kettunen, P., M. Laanti. 2005. How to steer an embedded software
Systems. John Wiley & Sons, Chichester, UK. project: Tactics for selecting the software process model. Inform.
Fine, C., D. Whitney. 1999. Is the make-buy decision process a Software Tech. 47(9) 587–609.
core competence? M. Muffato, K. Pawar, eds. Logistics in the King, J. L., K. Lyytinen. 2005. Automotive informatics: Informa-
Information Age—Proc. 4th Internat. Sympos. Logistics, Servizi tion technology and enterprise transformation in the automo-
Grafici Editoriali, Padova, Italy, 31–63. bile industry. W. H. Dutton, B. Kahin, R. O’Callaghan, A. W.
Franklin, G. F., J. D. Powell, M. L. Workman. 1998. Digital Control Wychoff, eds. Transforming Enterprise: The Economic and
of Dynamic Systems. Addison-Wesley, Upper Saddle River, NJ. Social Implications of Information Technology. MIT Press,
Cambridge, MA, 283–312.
Gerard, D., L. B. Lave. 2005. Implementing technology-forcing poli-
cies: The 1970 Clean Air Act Amendments and the introduction Kogut, B., U. Zander. 1992. Knowledge of the firm, combinative
of advanced automotive emissions controls in the United States. capabilities, and the replication of technology. Organ. Sci. 3(3)
Tech. Forecasting Soc. Change 72(7) 761–778. 383–397.
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
1446 Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS

Koopman, P. 2002. Critical embedded automotive networks. IEEE Popp, D. 2005. Lessons from patents: Using patents to measure tech-
Micro 22(4) 14–18. nological change in environmental models. Ecological Econom.
Langlois, R. N. 1997. Cognition and capabilities: Opportunities seized 54(2–3) 209–226.
and missed in the history of the computer industry. R. Garud, Powell, W. W., K. W. Koput, L. Smith-Doerr. 1996. Interorganiza-
P. Nayyar, Z. Shapira, eds. Technological Entrepreneurship: tional collaboration and the locus of innovation: Networks of
Oversights and Foresights. Cambridge University Press, New learning in biotechnology. Admin. Sci. Quart. 41(1) 116–145.
York, 71–94.
Powers, W. F., P. R. Nicastri. 2000. Automotive vehicle control
Langlois, R. N. 2003. The vanishing hand: The changing dynamics of
challenges in the 21st century. Control Engrg. Practice 8(6)
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

industrial capitalism. Indust. Corporate Change 12(2) 351–385.

605–618.
Lanjouw, J. O., A. Pakes, J. Putnam. 1998. How to count patents
and value intellectual property: The uses of patent renewal and Prencipe, A. 2000. Breadth and depth of technological capabilities
application data. J. Indust. Econom. 46(4) 405–432. in CoPS: The case of the aircraft engine control system. Res.
Policy 29(7–8) 895–911.
Lee, J., F. Veloso. 2008. Interfirm innovation under uncertainty:
Empirical evidence for strategic knowledge partitioning. J. Prod- Prencipe, A. 2002. Strategy, Systems, and Scope: Managing Systems
uct Innovation Management 25(5) 418–435. Integration in Complex Products. Sage, London.
Leen, G., D. Heffernan. 2002. Expanding automotive electronic sys- Prencipe, A., A. Davies, M. Hobday. 2003. The Business of Systems
tems. IEEE Computer 35(1) 88–93. Integration. Oxford University Press, Oxford, UK.
Leen, G., D. Heffernan, A. Dunne. 1999. Digital network in the auto- Ragatz, G. L., R. G. Handfield, T. V. Scanneil. 1997. Success factors
motive vehicle. Comput. Control Engrg. J. 10(6) 257–266. for integrating suppliers into new product development. J. Prod-
Lichtenthaler, U. 2008. Open innovation in practice: An analysis of uct Innovation Management 14(3) 190–202.
strategic approaches to technology transactions. IEEE Trans.
Rosenkopf, L., A. Nerkar. 2001. Beyond local search: Boundary-
Engrg. Management 55(1) 148–157.
spanning, exploration, and impact in the optical disk industry.
Lyytinen, K., L. Mathiassen, J. Ropponen. 1998. Attention shaping Strategic Management J. 22(4) 287–306.
and software risk—A categorical analysis of four classical risk
management approaches. Inform. Systems Res. 9(3) 233–255. Sanchez, R., J. T. Mahoney. 1996. Modularity, flexibility, and knowl-
edge management in product and organization design. Strategic
Majchrzak, A., R. E. Rice, A. Malhotra, N. King, S. Ba. 2000. Tech- Management J. 17(Special issue) 63–76.
nology adaptation: The case of a computer-supported interorga-
nizational virtual team. MIS Quart. 24(4) 569–600. Scherer, F. M. 1983. The propensity to patent. Internat. J. Indust.
Organ. 1(1) 107–128.
McConnell, V. D., M. A. Walls, W. Harrington. 1995. Evaluat-
ing the costs of compliance with mobile source emission Scherer, F. M., D. Harhoff. 2000. Technology policy for a world of
control requirements: Retrospective analysis. Discussion paper, skew-distributed outcomes. Res. Policy 29(4–5) 559–566.
Resources for the Future, Washington DC.
Simon, H. A. 1996. The Sciences of the Artificial. MIT Press, Cam-
Mesarovic, M. D., D. Macko, Y. Takahara. 1970. Theory of Hierar- bridge, MA.
chical, Multilevel Systems. Academic Press, New York.
Sosa, M. E., S. D. Eppinger, C. M. Rowles. 2004. The misalignment
Miller, R., M. Hobday, T. Leroux-Demers, X. Olleros. 1995. Innova-
of product architecture and organizational structure in complex
tion in complex systems industries: The case of flight simulation.
product development. Management Sci. 50(12) 1674–1689.
Indust. Corporate Change 4(2) 363–400.
Mondt, J. R. 2000. Cleaner Cars: The History and Technology of Steinmueller, W. E. 2003. The role of technical standards in coor-
Emission Control Since the 1960s. Society of Automotive Engi- dinating the division of labor in complex systems industries.
neers, Warrendale, PA. A. Prencipe, A. Davies, M. Hobday, eds. The Business of
Systems Integration. Oxford University Press, Oxford, UK,
Mumford, E. 2003. Redesigning Human Systems. Idea Publishing, 133–151.
New York.
Takeishi, A. 2001. Bridging inter- and intra-firm boundaries: Man-
Murmann, J. P., K. Frenken. 2006. Toward a systematic framework
agement of supplier involvement in automobile product devel-
for research on dominant designs, technological innovations, and
opment. Strategic Management J. 22(5) 404–433.
industrial change. Res. Policy 35(7) 925–952.
Papodopoulos, Y., C. Grante. 2005. Evolving car designs using model- Takeishi, A. 2002. Knowledge partitioning in the interfirm division of
based automated safety analysis and optimization techniques. J. labor: The case of automotive product development. Organ. Sci.
Systems Software 76(1) 77–89. 13(3) 321–338.
Pavitt, K. 1985. Patent statistics as indicators of innovative activities: Tucker, D. J., J. V. Singh, A. G. Meinhard. 1990. Organizational
Possibilities and problems. Scientometrics 7(1–2) 77–99. form, population dynamics, and institutional change: The found-
ing patterns of voluntary organizations. Acad. Management J.
Penner-Hahn, J., J. M. Shaver. 2005. Does international research and
33(1) 151–178.
development increase patent output? An analysis of Japanese
pharmaceutical firms. Strategic Management J. 26(2) 121–140. Turner, J. 1987. Understanding the elements of system design. R. J.
Pop, P., P. Eles, Z. Peng. 2004. Analysis and Synthesis of Distributed Boland, R. Hirschheim, eds. Critical Issues in Information Sys-
Real-Time Embedded Systems. Kluwer Academic Publishers, tems Research. John Wiley & Sons, Chichester, UK, 97–111.
Boston. Tushman, M., J. P. Murmann. 1998. Dominant designs, technol-
Popp, D. 2003. Pollution control innovations and the Clean Air Act ogy cycles and organizational outcomes. Res. Organ. Behav. 20
of 1990. J. Policy Anal. Management 22(4) 641–660. 231–266.
 Lee and Berente: Digital Innovation and the Division of Innovative Labor
Organization Science 23(5), pp. 1428–1447, © 2012 INFORMS 1447

Ulrich, K. 1995. The role of product architecture in the manufacturing Yip, P. S. L., E. W. K. Tsang. 2007. Interpreting dummy variables and
firm. Res. Policy 24(3) 419–440. their interaction effects in strategy research. Strategic Organ.
5(1) 13–30.
Ulrich, K., S. Eppinger. 1995. Product Design and Development, 4th
ed. McGraw-Hill, New York. Yoo, Y. 2010. Computing in everyday life: A call for research on
experiential computing. MIS Quart. 34(2) 213–231.
U.S. Environmental Protection Agency (EPA). 1990. Environmental
investments: The cost of a clean environment. Report EPA-230- Yoo, Y., R. J. Boland Jr., K. Lyytinen. 2006. From organization design
1 1-90-083, Washington, DC. to organization designing. Organ. Sci. 17(2) 215–229.
Yoo, Y., O. Henfridsson, K. Lyytinen. 2010. The new organizing
Van de Ven, A. H. 2005. Running in packs to develop knowledge-
Downloaded from informs.org by [129.93.16.3] on 02 April 2015, at 21:28 . For personal use only, all rights reserved.

logic of digital innovation: An agenda for information systems

intensive technologies. MIS Quart. 29(2) 365–378.
research. Inform. Systems Res. 21(4) 724–735.
Venkatesan, R. 1992. Strategic sourcing: To make or not to make.
Harvard Bus. Rev. 70(6) 98–107.
Jaegul Lee is an assistant professor of management at
von Bertalanffy, L. 1968. General System Theory: Foundations, the School of Business Administration at Wayne State Uni-
Development, Applications. George Braziller, New York. versity. He received his Ph.D. in technology management
Von Hippel, E. 1988. The Sources of Innovation. Oxford University and entrepreneurship from Carnegie Mellon University. His
Press, Oxford, UK. research interests rest in the intersection of technological inno-
Warwick, K., D. Rees. 1988. Industrial Digital Control Systems, vation, entrepreneurship, and firm strategy.
revised ed. Peter Peregrinus, Stevenage, UK. Nicholas Berente is an assistant professor of management
information systems at the Terry College of Business at the
White, H. 1980. A heteroskedasticity-consistent covariance matrix University of Georgia. He received his Ph.D. in information
estimator and a direct test for heteroskedasticity. Econometrica
systems from Case Western Reserve University’s Weatherhead
48(4) 817–830.
School of Management and conducted his postdoctoral studies
Whitney, D. 1990. Designing the design process. Res. Engrg. Design at the University of Michigan’s School of Information. His
2(1) 3–13. research focuses on organizing for innovation and the role of
Wilson, D. 1969. Forms of hierarchy: A selected bibliography. Gen- digital technologies in organizations, particularly with respect
eral Systems 14 3–15. to innovative design practices.
 Materials for Sustainable Energy

Research in materials science is contributing to progress towards a sustainable future based on

renewable energy sources and on a better management of energy utilization and storage.

Civilization continues to be transformed by our ability to harness energy beyond human and
animal power. The series of industrial and agricultural revolutions have allowed an increasing
fraction of the world population to heat and light their homes, fertilize and irrigate their crops,
and to connect and travel around the world. All of this progress is fueled by our ability to find,
extract and use energy with ever increasing dexterity.

The world energy consumption in 2012 was 549 Quadrillion British thermal units (or Quad
BTUs, with 1 Quad BTU = 293 TWh)1, the equivalent of continuously having 25 billion horses
working for us. The modern-day equivalent of the biological waste of horses is represented by
the pollutants created from fossil fuels that currently supply us with 85% of our energy. The CO2
in the atmosphere has risen from 278 ppm at the beginning of the industrial revolution to ~ 410
ppm today. In addition to CO2, other significant greenhouse gases (GHGs) are CH4, N2O and
fluorinated gases, and the total CO2 equivalent concentration (CO2 equiv) is 496 ppm. Notably,
75% of this rise has taken place since 19502. Burning fossil fuels also creates other
environmental pollutants such as SOx, NOx, particulate matters (PM), volatile organic
compounds and toxic heavy metals.

As a result of our GHG emissions, our climate is changing. The increasing heat waves and
droughts, torrential heavy rainfalls and accompanying floods, and the encroaching seas are early
but ominous warning signs. The Earth’s climate system is extremely complex and there are
considerable uncertainties in predicting future risks An important cause for such uncertainty is
the huge thermal inertia of cold deep oceans3, which are slowly warming up as a result of more
heat trapped by the increase of GHGs in the atmosphere. This means that, even if human
emissions stop tomorrow, it could take centuries to reach a new equilibrium temperature and
appreciate the full extent of the damage created. For these reasons, mitigating the climate risks
by dramatically decreasing our GHG emissions is a prudent insurance policy; waiting for
additional evidence is foolhardy.

“If you don’t change direction, you will end up where you are heading”, said Lao Tzu, the father
of Taoism. Where are we heading? Energy use is projected to climb to 629 Quad BTUs by 2020,
and 813 Quad BTUs by 20401. The global emissions of CO2 doubled from 1975 to 2015, and are
projected to triple by 2040 if we stay on our present course4.

In 2012, we discussed the opportunities and challenges in the transition to a sustainable future5,
focusing on carbon emissions in the transportation and electricity generating sectors with account
for less than half of GHG emissions. We note that 30% of GHG emissions and most of the CH4
and N2O emissions are from agriculture, forestry and land use4, and the remainder from
industrial processes and heat. Because anthropogenic GHG emissions are coming from virtually
every sector of our economy, many solutions are needed to meet this global challenge.

1
Although fossil fuels will remain a significant source of energy and materials for decades to
come, a path to zero emissions is envisioned in Fig. 1. An increasing amount of clean renewable
and nuclear energy is being converted into electricity for household applications such as lighting
and air conditioning, and for industrial production, while advances in materials science are
lowering the carbon footprint in buildings. Transmission and distribution of electricity is
becoming more efficient thanks to breakthroughs in wide bandgap semiconductors and dielectric
materials. Electricity stored in batteries is increasingly being used to power electric vehicles and
back up intermittent renewables. We will transition from a chemical industry based on oil and
gas feedstocks to one that is increasingly based on renewable agricultural residues. As the cost of
clean energy continues to plunge, new catalysts will enable us to economically convert CO2 and
H2O directly into fuels and chemicals. Moreover, metal-organic frameworks (MOFs) and other
novel materials are being used to capture carbon from power generation, cement, aluminum,
plastic and steel production, and ultimately, the atmosphere. After recycling as much of the
captured carbon as possible, advances in materials science will allow us to permanently sequester
the rest of the carbon6.

Here, we give an overview of the opportunities that recent exciting results in materials science
may open in selected areas of energy research. Some of the recent important scientific
developments are reviewed in more detail in the accompanying review articles7–11.

2
Figure 1 | Renewable energy future. Advances in materials science and engineering are
increasing the efficiency of electricity generation from clean renewable energy sources — for
instance using solar panels (a) and wind turbines (b) — as well as electricity distribution (c) and
storage (d). Improved energy management in buildings (e) and widespread diffusion of electric
vehicles (f) are decreasing our carbon footprint. Better catalysts and photoelectrochemical
devices allow a more efficient generation of hydrogen and oxygen (g), and CO2 conversion to
fuels and chemicals (h). Carbon-capture materials (i) will decrease the amount of CO2 released in
the atmosphere. With thanks to Paloma Liu for the design of the images.

Photovoltaic devices

The price of solar power modules has plunged from 3 US $ W-1 in 2008 (the price of a module to
generate 1 watt of power under a defined illumination intensity of one “sun”) to 0.50 US$ W-1 in
201612. In 2011, the U.S. Department of Energy (DOE) SunShot 2020 goals for the full levelized
cost of electricity (L.C.O.E., representing the total cost to build and operate an energy producing
source over its anticipated lifetime divided by the total amount of electricity produced) were 0.06
US$ kWh-1 and 0.09 US$ kWh-1 for utility and residential solar power, respectively13. It appears
likely that the SunShot goals will be met, and indeed, the L.C.O.E. of solar energy in locations
such as Dubai with excellent solar insolation is already as low as 0.03 - 0.04 US$ kWh-1,
according to recent price bids submitted by developers of renewable energy systems in reverse
auctions (Fig. 2a)14. As a result, the DOE recently announced their 2030 goals of 0.03 US$ kWh-
1
and 0.05 US$ kWh-1 L.C.O.E. for utility and residential solar power.

Fig. 2b shows one pathway of reducing prices from the 2016 benchmark of 0.07$ kWh-1 to 0.03
US$ kWh-1, leveraging on the reductions expected for the fabrication and maintenance costs of
solar panels, and on the expected increase in lifetime15. In 2015, 69.5% of the production of solar
modules was covered by polycrystalline silicon technology, 23.9 % by single crystalline silicon
and 6.6% by CdTe. The power-conversion efficiencies of solar modules are as high as 21% for
single crystal silicon, typically 17% for polycrystalline, and 15.6% for CdTe, whereas at the
laboratory cell level the highest efficiencies are 25.6%, 21.3% and 21%, respectively16. Steady
reductions in capital and manufacturing costs, along with financing and policies such as reverse
auction bidding are expected to drive the photovoltaic electricity cost down to 0.03 US$ kWh-1.
Future reductions will come from higher module efficiency, longer module and electronics
lifetimes (>35 years) and lower efficiency degradation (< 0.3%/year).

While conventional Si and CdTe will continue to drive down the cost of PV electricity, to reach
values lower than 0.02 US$ kWh-1 new mechanisms to increase the efficiency of solar cells and
materials having reduced production and processing costs will be needed. Many of the factors
that affect the performance of single and multi-junction solar cells are discussed in ref. 8. The
limit of single junction cell efficiency, known as the Shockley-Queisser Limit, can be overcome
by concentrating sunlight, by a better management of direct incident and diffuse light — for
instance, using nanocone structures to enhance light absorption in solar cells17 —, or by the use
of multi-junction cells. Other possible improvements include intermediate bandgap solar cells,
photon up-conversion or down-conversion, luminescence concentrators, carrier multiplication,

3
and thermophotovoltaic schemes. Developing methods of coherently transferring energy, as is
done in biological photosynthetic systems, is another possibility to improve solar cells
efficiency7. The use of novel materials may open promising alternative paths to traditional
technologies. For instance, perovskite materials18 such as CH3NH3PbI3 have demonstrated their
potential within a very short time (less than 5 years), reaching a power conversion efficiency of
22.1% for small-scale solar cells19. In addition to their good performance, perovskites are
advantageous for their low-cost processing and for the possibility to tune their absorption
spectrum for an optimized integration in multi-junction cells; however, long-term durability
remains a major challenge.

Finally, self-cleaning, super hydrophobic surfaces based on nanostructured materials are being
investigated to keep the solar module surfaces clear of dirt and sand with minimal water use, as
well as to prevent ice formation on airplanes and wind turbine blades, and as antifouling coatings
for the hulls of marine vessels. Many of these anti-wetting and anti-icing surfaces mimic the
micro and nanostructures of lotus leaves (Fig. 2c and 2d) and insect wings20–22.

Figure 2 | Photovoltaic technologies. a, Reverse auction prices of unsubsidized solar contracts.

The bids in Texas and New Mexico have been multiplied by 1.4 to account for the 30%
investment tax credit. b, A DOE example of a modeled pathway to 0.03 US$ kWh-1 by 203015.
O&M: operation and maintenance costs. c, A drop of water on a Lotus leaf (bottom) and a
cartoon of how a lotus leaf efficiently cleans itself with minimal water use (top), with a non-
4
wetting droplet of water efficiently picking up dirt20. d, Water helps cleaning dirt from a lotus
leaf (top), whereas dirt, snow and ice remain deposited on the surface of untreated solar panels
(bottom)21.

Batteries

Robust, rechargeable batteries are essential for electric vehicles (EV) and grid storage. The DOE
has set a cost target for EV battery packs of 125 US$ kWh-1 to be reached in 2022, which would
assure the wide scale deployment of clean transportation technologies23. The current EV battery
cells have a specific energy of 200-250 Wh kg-1, but an aggressive goal to reach 500 Wh kg-1
with a >10-year life and total mileage of ~150,000 miles has been proposed by the Battery500
Consortium. In grid-storage applications, high energy density is less crucial, but requirements on
performance stability — more than 6000 cycles and lifetimes surpassing 20 years — are more
stringent.

Lithium ion batteries (LIB) have been the dominating technology for applications in
transportation and consumer electronics. The current LIB is based on graphite anodes and
lithium metal oxide cathodes such as layered LiNixCoyMnzO2 (NMC), LiMn2O4, LiFePO4.
Li4Ti5O12 has also been used as anode for fast charging in electric buses but it has lower energy
density due to a high anode potential24. Further increase in the energy density and power density
calls for new electrode materials.

Figure 3 | Materials for batteries. a, Si nanowire electrodes25. b Sulfur/mesoporous carbon

electrode material26. c, Practical specific energy of cells versus different combinations of
electrodes (Graphite/NMC: 300 Wh kg-1, Si/NMC: 400 Wh kg-1, Li/NMC: 500 Wh kg-1, Li/S:
>600 Wh kg-1).

On the anode side, silicon has 10 times higher specific capacity (4200 mAh g-1 for silicon, 2100
mAh g-1 for Li4.4Si) than graphite (372 mAh g-1), but it undergoes a four-times volume change
between charged and discharged states that may cause mechanical failure. In the past decade,
solutions to overcome mechanical fracture and instability of the solid-electrolyte interphase have
been proposed based on a variety of nanoscale designs (Fig. 3a)25. Lithium metal anodes are
another interesting choice due to their high capacity (3860 mAh g-1). However, dendrite
5
formation, low cycling Coulombic efficiency, short battery life and safety concerns have so far
prevented commercialization. Several recent approaches, such as the use of nanoscale design,
electrolyte additives, solid state electrolytes, and the combination of Li metal with graphene
oxide hosts27, show promise to overcome these limitations.

On the cathode side, the continuous improvement of high Ni and Li-rich LiNixCoyMnzO2 might
result in an appreciable improvement of specific capacity (~250 mAh g-1) of batteries, due to the
possibility for both transition metal atoms and oxygen to change their oxidation state. However,
the chemical and structural stability of these materials needs to be improved28. A potentially
large capacity improvement could come from sulfur (~1700 mAh g-1) and O2 (~1700 mAhg-1 for
Li2O2) cathodes, although these technologies are affected by long-term stability issues too. Also
in this case, the use of nanostructured architectures26 (Fig.3b) may reveal effective in increasing
the cycle life and overall performance of these cathodes. As shown in Fig. 3c, proper
combination of some of these anodes and cathodes may allow the realization of Lithium-based
batteries with practical specific energy ranging from 200 to more than 600 Wh kg-1.

In addition, replacement of lithium with more sustainable chemistries based on Na, K, Mg and
Al are being investigated, though all the battery components - including anode, cathode, solvent,
salt and additives - need to be reconsidered and developed ad hoc to make these alternative
technologies competitive.

For grid scale applications, the cost, scalability and long cycle/calendar life are more important
figures of merit. Redox flow batteries show promise for grid storage, but the cost of the ion-
exchange membranes that separate the negative and positive electrolytes used in such devices
must be reduced, and the solubility of redox molecules needs to increase29. Semiflow batteries
(using one solid and one liquid electrode) such as Zn-halogen30 and Li-polysulfide31 systems
remove the needs of ion-selective membranes, potentially lowering costs. Aqueous electrolyte-
based batteries are attractive due to the low electrolyte cost and excellent safety32. A detailed
overview of the current status of research on batteries and the development of sustainable
materials and architectures for battery applications is reported in ref. 9.

Solar and chemical fuels

The above-mentioned progress notwithstanding, it should be considered that the volumetric and
gravimetric energy density of current electrochemical batteries is almost 100 times lower than
gasoline or diesel fuel. Even a fuel with relatively low energy density such as methanol is likely
to contain at least 10-fold higher chemical energy than any foreseeable sealed battery. The
advantages of hydrocarbon fuels that can be stored as liquids near standard temperature and
pressure conditions are enormous. The compression and refrigeration costs of liquefying natural
gas, plus the need to transport it quickly and store it as a cryogenic liquid are much higher than
the associated cost of dealing with liquid fuels. Slow moving oil tankers can be regarded as
transcontinental energy transmission conduits, where remarkably, the cost of moving oil around
the world only adds a few pennies to the cost of gallon of gasoline. As a consequence, when the
L.C.O.E. of clean electricity will drop to 0.02-0.03 US$ kWh-1, it may be more convenient to

6
store such electricity in the form of chemical fuels and high valuable chemicals through
electrocatalytic reactions such as water splitting and CO2 reduction.

Water splitting, 2𝐻2 𝑂 ↔ 2𝐻2 + 𝑂2, has four related half reactions: hydrogen evolution reaction
(HER), oxygen evolution reaction (OER), hydrogen oxidation reaction (HOR) and oxygen
reduction reaction (ORR). HER and OER are key for hydrogen fuel production, and HOR and
ORR occur in fuel cells. These four reactions traditionally all use precious metal catalysts, yet it
may be possible to achieve efficient conversion with earth abundant materials by understanding
the catalysis mechanism. For HER, platinum has been the most well-known catalyst, but recent
studies also show that 3d transition metals and their alloys, metal phosphides, sulfides, and
hydroxides are also very active11. For instance, layered MoS2 (Fig. 4a)33 has been shown to
present active Mo edge sites34,35. For OER, in addition to benchmarked RuO2 and IrO2, other
transition metal oxides and oxyhydroxides have shown very high activities. Pt and Ir are the most
active HOR catalysts, although NiMo is an attractive choice due to its low cost. Pt-based
catalysts are the most used in ORR36 including Pt alloys with transition metals (Fig.4b).
Multimetallic nanostructures offer extremely high surface to volume ratios. In the case of a Pt3Ni
nanoframe as shown in Fig. 4c, a 36x enhancement in mass activity and 22x enhancement in
specific activity for ORR reactions were observed37. In addition to exploring different
compositions of catalysts, a new approach was recently demonstrated based on electrochemical
tuning during lithium insertion and extraction (Fig. 4d) to control the chemical potential,
oxidation state and electronic structure, and efficient catalysts have been discovered38. The
contributions of theoretical simulations and fundamental characterization techniques to the
understanding of the catalytic mechanisms and to the design of new catalysts are discussed in ref.
10
.

7
Figure 4 | Materials for catalysis. a, Schematic drawing of MoS2 structure33. b, Different crystal
facets of Pt3Ni alloys and their ORR catalytic activity36. c, Nanoframes with three dimensional
electrocatalytic surfaces37. d, Electrochemical tuning of electrocatalytic activity38.

The electrochemical reduction of CO2 to other carbon compounds having higher chemical
energy: 𝐶𝑂2 + 𝐻2 𝑂 ↔ 𝐻𝑥 𝐶𝑦 𝑂𝑧 + 𝑂2 , is an attractive method to produce chemical fuels while
reducing carbon emissions. The recycling of carbon dioxide into methanol and derived products
(such as carbon monoxide39 and formic acid40) in a “methanol economy” has been touted as an
alternative to the hydrogen economy because of the inherently higher energy density41–43 of the
reaction products. A recent review of catalysts for the electro-reduction of carbon dioxide into
fuels is reported in ref. 44.

Alternative to electrochemical reactions are photoelectrochemical reactions which utilize

sunlight directly to convert water and CO2 into chemical fuels and feedstocks. As discussed in
refs.9,10, the photochemical reactions present numerous material requirements to simultaneously
combine photon absorption, charge separation, catalysis and electronic structure alignment with
chemical reactions. From an economic perspective, it is important to realize that in all schemes
to convert electrical energy or direct solar energy into chemical fuels the capital expense (and the
amortized cost of money) of the commercial plant is a significant fraction of the levelized cost of
any chemical plant. An electrochemical facility powered only by local PV will have a seasonal
duty cycle of 25% if every day is sunny. To maximize the return on the capital investment, it is
desirable to operate the facility as close to full capacity as possible.

8
Further opportunities

In addition to the research areas discussed above and in the Reviews in this Insight7–11, other
areas of research also deserve attention.

Materials for gas separation and storage

In the petro chemical industry, the initial cracking of long chain hydrocarbons occurs at high
temperatures. The further fractionations of light hydrocarbons of similar volatility and molecular
sizes such as ethylene/ethane (C2H4 and C2H6) and propylene/propane (C3H6 and C3H8) are
currently done at low temperatures and high pressures, and their separation is one of the most
energy-intensive industrial processes. MOFs have been recently shown to separate small chain
hydrocarbons, from CH4 to C3H8 with unprecedented selectivity by tuning pore size with
nanometer precision and attaching appropriate functional groups45.

The capture of CO2 in power generation and in industrial processes such as chemical, cement and
steel production is a necessary part of the required transition to deep decarbonization46. Carbon
capture methods include post-combustion capture of CO2 from the flue gas (such as liquid
amine), and pre-combustion separation where the fuel is converted into CO + H2 and then to CO2
+ H2 (such as in integrated gasification combined cycle plants). There is very limited operational
experience with carbon capture methods for mid-scale power plants in the range of 300 - 600
MW. Estimates of the cost of a new carbon capture power plant (based on the operational
experience of the first half a dozen commercial-scale plants) vary, but are generally believed to
increase the cost of electricity by 30% - 50%. The cost of retrofitted existing plants will be
higher, and in the case of lower thermal efficiency plants the additional parasitic energy load
would be prohibitive47,48. New materials such as MOFs49, zeolitic imidazolate frameworks
(ZIFs)50 and other nano and microstructured materials offer the possibility of reducing the cost
of electricity with CO2 capture to 15% - 20% additional cost. Computational methods using
genetic algorithms to search for high performing MOFs have been used51. These same classes of
materials may also be used in an oxy-combustion configuration, where O2 is separated from air,
and combustion of fuel in a CO2 + O2 atmosphere yields CO2 + H2O, which are easily separated.
This process may be particularly advantageous in natural gas combustion52. Oher advanced CO2
capture approaches include ionic liquids, chemical looping and biofixation53. Since the
likelihood of going over 600 ppm of CO2, equiv. is reasonably high, we will need direct air capture
of CO2 as well.

Unfortunately, there is no financial incentive beyond enhanced oil recovery to deploy

technologies for carbon capture, utilization or sequestration (CCUS). Even with policy
incentives, the cost of the first mid-size commercial scale carbon capture power plants — even
with new carbon capture materials — will be over U.S.$2 billion. For this reason, international
cooperation and information sharing is needed to assist the early testing of CCUS technologies.

Materials for thermal energy conversion and manipulation

Heat management is an important part of energy technology. Direct conversion of waste heat
into electricity has been pursued for several decades54. A figure of merit of thermoelectric
materials is the dimensionless parameter ZT = (2/)T, where , the Seebeck coefficient,

9
represents the electric potential generated for a given temperature difference T,  is the electrical
conductivity, and  is the thermal conductivity. Nanostructuring, heterostructuring and alloying
can be used to improve ZT. Recently, thermogalvanic materials have increased the effective ZT
to 3.5 and higher55 since electrochemical reactions often have much larger entropy change and
thus higher temperature coefficient (~1 mV/K) than usual thermoelectric materials.

Effective management of heat through radiation also offers opportunities to decrease the global
energy demand. For example, devices and clothes have been engineered to facilitate transmission
of 7-14 m infrared radiation for passive radiative cooling56,57. Smart-window systems using
electro-, photo-, and thermochromic mechanisms to dynamically control infrared transmission
through liquid crystals, nanoparticles such as ITO and other materials are used to enhance
building energy efficiency58,59.

Materials for power electronics

Power electronics is essential for the control and conversion of electricity in power systems,
motor drives, inverters and converters60,61. Silicon has been the predominant semiconductor for
these applications, but higher bandgap semiconductors such as SiC and GaN permit more
thermally efficient and higher voltage transistors. The major challenge lies in how to grow high
quality materials and junctions over large area. Diamond, with its wider bandgap, high
breakdown voltage and extraordinary high thermal conductivity is ideal for insulated-gate
bipolar transistors used in high voltage, high power switching applications; however, doping
processes to realize high conductivity in this material are challenging.

Microbial energy conversion

Microbes have been recently explored as biomaterials for energy conversion. In microbial fuel
cells, a microbial anode and an air cathode are used to convert organics in solution into
electricity62. The use of nanotechnologies to increase the surface area and roughness of the
anodes can improve the interface with microbes and enhance the electron collection efficiency.
Further optimizations are required to decrease voltage and charge losses due to the oxygen gas
reduction and diffusion in these cells. Microbial biosynthesis is also being used to convert CO2,
water and low-cost energy sources into chemicals, for example by using CdS nanoparticles to
increase the photoresponse of microbes and boost their CO2 reduction efficiency when exposed
to solar light63.

Development and deployment

Materials science is an enabling science and is playing a central role in our transition to
sustainable energy. The advances in the past decades have been breathtaking. While accidental
discovery remains a major source of new materials, the empirical trial-and-error method of
innovation has been profoundly supplemented by a deeper understanding of the basic physics
and chemistry of materials and our ability to measure and fabricate materials at the molecular
and nanoscale.

10
Such continued scientific advances must be adequately supported by the public sector since
much of the materials science research is too long-term to garner private sector investments.
Beyond research, there are two “valleys of death” that have to be crossed in the transition from
scientific discovery to innovation and wide scale deployment. The first valley crossing requires
patient private sector investors that are found in most venture capital or equity investment firms.
The second valley is the scale-up from small pilot projects to commercial size facilities where
economies of scale and economies of large quantity production can be used to dramatically
reduce costs.

For certain technologies, these crossings require government subsidies. As an example, the
purchase price of solar PV modules has plunged by 50x since 1980, and by an additional 10x
since 200564. This remarkable drop in price was largely stimulated by a generous feed-in tariff by
the German government that created a large market, which in turn drew huge investments in
manufacturing plants.

Wind technology also received substantial government support. Support initially came from the
U.S., but it was the long-term, consistent sustenance from Denmark and Germany that ultimately
made wind energy economically competitive. Those subsidies led to improved manufacturing
processes of composite materials for wind blades and more robust power electronics. In the
future, substitute materials for rare earth magnets and/or more environmentally friendly rare
earth refining methods are needed to lower the cost of reliable direct drive turbines.

The development of lithium ion battery technology followed a different path. The initial market
pull was created by laptop computers. The cell phone market has become the major stimulant for
high volumetric energy density and fast-charge batteries, while the rapidly emerging drone
market seeks the highest gravimetric energy density. The emerging EV and plug-in hybrid EV
markets were aided by purchase rebates, and in some areas in the U.S., by access to (high-
occupancy vehicle lanes, regardless of the number of passengers in the vehicle. As a result of
these policies, the manufacturing cost of EV batteries has decreased 3-fold from 2006 to 2015 to
~ 300 U.S. $ kWh-1 for a battery format originally designed for laptop computers. By 2018, the
manufacturing cost is expected to drop by another factor of 2.

Currently there is neither a market draw nor a regulatory push to dramatically reduce carbon
emissions, which is instead essential to avoid human-induced climate change becoming the
ultimate “tragedy of the commons.” The 2015 Paris Agreement was a necessary first step, yet the
potentially disastrous long-tail climate risks demand stronger actions.

Most economists agree that the most efficient mechanism to stimulate the shift to a very low
carbon economy is a meaningful price on carbon. If the cost of carbon emissions is included in
the price of fossils fuels and primary energy intensive industries, they argue that market forces
will be more efficient than regulations.

The European Union has put in place a cap-and-trade policy since 2005, where carbon emission
is capped but emitters can trade emission allowances. The price of carbon emissions would
increase as the cap is progressively lowered in order to provide continuing economic incentive to
reduce emissions. By early 2013, the allowance price has been hovering around 5-6 €/tonne of

11
CO2, however most experts agree that the price needed to stimulate widespread decarbonization
is between 70 and 80 €/tonne of CO2. In hindsight, too many carbon credits were given away in
the EU trading scheme. The method of assigning allocation limits for each country is a
regulatory headache, open to gamesmanship, and the political will to aggressively lower the cap
on carbon remains challenging.

An alternative to a cap-and-trade system is a simple carbon tax. A tax maximizes transparency,

minimizes market manipulation and regulatory complexity, and provides investment certainty.
However, a carbon tax will raise the cost of energy. A politically more attractive route is a
revenue neutral scheme where the money collected with this tax is returned equally to every
working adult. The price on carbon can rise gradually to 80 €/tonne by 2040 to allow industries
and individuals time to adjust. Fossil fuel subsidies should be eliminated as soon as possible and,
as the carbon price rises, subsidies for renewable energy can be phased out. Nuclear energy, a
carbon-free, energy-on-demand source, can compete with fossil energy with the added costs of
carbon capture and utilization or sequestration.

A carbon tax avoids the problem of how to allocate carbon emissions credits between developed
and developing countries. It levies the highest taxes on the biggest emitters, and global
agreement is not needed. Leaders such as the EU, China and hopefully the U.S. need to lead. The
playing field can be leveled with suitable border tariffs supported by the World Trade
Organization. This option is especially important to assure leading countries that their economies
will not be handicapped.

The most important aspect of a carbon tax rising inexorably to ~ 80 €/tonne is that it will rapidly
unleash scientific ingenuity, innovation, and market investments that are still needed. In the last
six years, the cost of clean energy has plunged and in many areas of the world, the life-cycle cost
of wind and solar energy is reaching parity with fossil energy. However, the cost of
decarbonizing the first 25% of the world economy is far less than the cost of decarbonizing the
last 25%. Achieving affordable and fully clean energy in our electrical and chemical
transmission, distribution and storage systems present significant technical and political
challenges. We must and shall overcome these challenges: otherwise, to quote Martin Luther
King, “There is such a thing as being too late.”

Steven Chu is at the Department of Physics and Department of Molecular and Cellular
Physiology, Stanford University, Stanford, California 94305, USA. Yi Cui is at the Department
of Materials Science and Engineering, Stanford University, Stanford, California 94305, USA,
and Stanford Institute for Materials and Energy Sciences, SLAC National Accelerator
Laboratory, 2575 Sand Hill Road, Menlo Park, California 94025, USA. Nian Liu is at
Department of Physics and Department of Materials Science and Engineering, Stanford
University, Stanford, California, 94305, USA.
Email: [email protected]

12
References
1. U.S. Energy Information Administration | International Energy Outlook 2016. (2016).
2. The NOAA Annual Greenhouse Gas Index (AGGI). (2016). Available at:
http://www.esrl.noaa.gov/gmd/aggi/aggi.html. (Accessed: 24th November 2016)
3. Roemmich, D. et al. Unabated planetary warming and its ocean structure since 2006. Nat.
Clim. Chang. 5, 240–245 (2015).
4. IPCC Climate Change 2014 Synthesis Report. (2014). Available at:
https://www.ipcc.ch/report/ar5/syr/.
5. Chu, S. & Majumdar, A. Opportunities and challenges for a sustainable energy future.
Nature 488, 294–303 (2012).
6. Boot-Handford, M. E. et al. Carbon capture and storage update. Energy Environ. Sci. 7,
130–189 (2014).
7. Brédas, J.-L., Sargent, E. H. & Scholes, G. D. Photovoltaic concepts inspired by
coherence effects in photosynthetic systems. Nat. Mater. Page# to be added (2017).
8. Green, M. A. & Bremner, S. P. Energy Conversion Approaches and Materials for High
Efficiency Photovoltaics. Nat. Mater. Page# to be added (2017).
9. Grey, C. P. & Tarascon, J. M. Battery research towards sustainable chemistries and in-situ
monitoring. Nat. Mater. Page# to be added (2017).
10. Montoya, J. H. et al. Materials for Solar Fuels and Chemicals. Nat. Mater. Page# to be
added (2017).
11. Stamenkovic, V. R., Strmcnik, D., Lopes, P. P. & Markovic, N. M. Energy and Fuels from
Electrochemical Interfaces. Nat. Mater. Page# to be added (2017).
12. The September 2016 spot market price of solar modules. (2016). Available at:
http://www.solarserver.com/service/pvx-spot-market-price-index-solar-pv-modules.html.
(Accessed: 24th November 2016)
13. DOE Office of Energy Efficiency and Renewable Energy SunShot Vision Study. (2012).
14. Low Solar Prices Scaring Companies Away From Solar Auctions. (2016). Available at:
https://cleantechnica.com/2016/07/27/low-solar-prices-scaring-companies-away-solar-
auctions/. (Accessed: 7th November 2016)
15. DOE SunShot White Paper. Available at: http://energy.gov/eere/sunshot/sunshot-2030.
(Accessed: 24th November 2016)
16. Photovoltaics Report. (2016). Available at:
https://www.ise.fraunhofer.de/de/downloads/pdf-files/aktuelles/photovoltaics-report-in-
englischer-sprache.pdf. (Accessed: 24th November 2016)
17. Zhu, J., Hsu, C.-M., Yu, Z., Fan, S. & Cui, Y. Nanodome Solar Cells with Efficient Light
Management and Self-Cleaning. Nano Lett. 10, 1979–1984 (2010).
18. Burschka, J. et al. Sequential deposition as a route to high-performance perovskite-
sensitized solar cells. Nature 499, 316–319 (2013).
19. Seo, J., Noh, J. H. & Seok, S. Il. Rational Strategies for Efficient Perovskite Solar Cells.
Acc. Chem. Res. 49, 562–572 (2016).
20. Nishimoto, S. & Bhushan, B. Bioinspired self-cleaning surfaces with
superhydrophobicity, superoleophobicity, and superhydrophilicity. RSC Adv. 3, 671–690
(2013).
21. Zhang, P. & Lv, F. Y. A review of the recent advances in superhydrophobic surfaces and
the emerging energy-related applications. Energy 82, 1068–1087 (2015).
22. Gogolides, E., Ellinas, K. & Tserepi, A. Hierarchical micro and nano structured,

13
 hydrophilic, superhydrophobic and superoleophobic surfaces incorporated in
microfluidics, microarrays and lab on chip microsystems. Microelectron. Eng. 132, 135–
155 (2015).
23. EV Everywhere Grand Challenge Blueprint. (2013). Available at:
http://energy.gov/sites/prod/files/2016/05/f31/eveverywhere_blueprint.pdf. (Accessed:
24th November 2016)
24. Whittingham, M. S. Lithium batteries and cathode materials. Chem. Rev. 104, 4271–4301
(2004).
25. Chan, C. K. et al. High-performance lithium battery anodes using silicon nanowires. Nat.
Nanotechnol. 3, 31–35 (2008).
26. Ji, X., Lee, K. T. & Nazar, L. F. A highly ordered nanostructured carbon-sulphur cathode
for lithium-sulphur batteries. Nat. Mater. 8, 500–6 (2009).
27. Lin, D. et al. Layered reduced graphene oxide with nanoscale interlayer gaps as a stable
host for lithium metal anodes. Nat. Nanotechnol. 11, 626–632 (2016).
28. Sathiya, M. et al. Reversible anionic redox chemistry in high-capacity layered-oxide
electrodes. Nat. Mater. 12, 827–35 (2013).
29. Li, X., Zhang, H., Mai, Z., Zhang, H. & Vankelecom, I. Ion exchange membranes for
vanadium redox flow battery (VRB) applications. Energy Environ. Sci. 4, 1147 (2011).
30. Li, B. et al. Ambipolar zinc-polyiodide electrolyte for a high-energy density aqueous
redox flow battery. Nat. Commun. 6, 6303 (2015).
31. Yang, Y., Zheng, G. & Cui, Y. A membrane-free lithium/polysulfide semi-liquid battery
for large-scale energy storage. Energy Environ. Sci. 6, 1552–1558 (2013).
32. Kim, Y. J., Wu, W., Chun, S.-E., Whitacre, J. F. & Bettinger, C. J. Biologically derived
melanin electrodes in aqueous sodium-ion energy storage devices. Proc. Natl. Acad. Sci.
U. S. A. 110, 20912–7 (2013).
33. Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Single-layer MoS2
transistors. Nat. Nanotechnol. 6, 147–150 (2011).
34. Hinnemann, B. et al. Biomimetic Hydrogen Evolution:  MoS2 Nanoparticles as Catalyst
for Hydrogen Evolution. J. Am. Chem. Soc. 127, 5308–5309 (2005).
35. Jaramillo, T. F. et al. Identification of active edge sites for electrochemical H2 evolution
from MoS2 nanocatalysts. Science 317, 100–102 (2007).
36. Stamenkovic, V. R. et al. Improved oxygen reduction activity on Pt3Ni (111) via
increased surface site availability. Science 315, 493–497 (2007).
37. Chen, C. et al. Highly Crystalline Multimetallic Nanoframes with Three-Dimensional
Electrocatalytic Surfaces. Science 343, 1339 LP-1343 (2014).
38. Wang, H. et al. Electrochemical tuning of vertically aligned MoS2 nanofilms and its
application in improving hydrogen evolution reaction. Proc. Natl. Acad. Sci. U. S. A. 110,
19701–19706 (2013).
39. Asadi, M. et al. Nanostructured transition metal dichalcogenide electrocatalysts for CO2
reduction in ionic liquid. Science 353, 467–470 (2016).
40. Gupta, K., Bersani, M. & Darr, J. A. Highly efficient electro-reduction of CO2 to formic
acid by nano-copper. J. Mater. Chem. A 4, 13786–13794 (2016).
41. Kuhl, K. P. et al. Electrocatalytic conversion of carbon dioxide to methane and methanol
on transition metal surfaces. J. Am. Chem. Soc. 136, 14107–14113 (2014).
42. Goeppert, A., Czaun, M., Jones, J.-P., Surya Prakash, G. K. & Olah, G. A. Recycling of
carbon dioxide to methanol and derived products – closing the loop. Chem. Soc. Rev. 43,

14
 7995–8048 (2014).
43. Studt, F. et al. Discovery of a Ni-Ga catalyst for carbon dioxide reduction to methanol.
Nat. Chem. 6, 320–324 (2014).
44. Qiao, J. et al. A review of catalysts for the electroreduction of carbon dioxide to produce
low-carbon fuels. Chem. Soc. Rev. 43, 631–675 (2014).
45. Bloch, E. D. et al. Hydrocarbon Separations in a Metal-Organic Framework with Open
Iron(II) Coordination Sites. Science 335, 1606 LP-1610 (2012).
46. Chu, S. Carbon Capture and Sequestration. Science 325, 1599 LP-1599 (2009).
47. IPCC Special Report on Carbon Dioxide Capture and Storage. (2005). Available at:
https://www.ipcc.ch/pdf/special-reports/srccs/srccs_wholereport.pdf. (Accessed: 24th
November 2016)
48. Cost and Performance Baseline for Fossil Energy Plants Volume 1: Bituminous Coal and
Natural Gas to Electricity. (2013). Available at: http://www.netl.doe.gov/energy-
analyses/pubs/BitBase_FinRep_Rev2a-3_20130919_1.pdf. (Accessed: 24th November
2016)
49. Mason, J. A. et al. Application of a High-Throughput Analyzer in Evaluating Solid
Adsorbents for Post-Combustion Carbon Capture via Multicomponent Adsorption of
CO2, N2, and H2O. J. Am. Chem. Soc. 137, 4787–4803 (2015).
50. Banerjee, R. et al. High-throughput synthesis of zeolitic imidazolate frameworks and
application to CO2 capture. Science 319, 939–943 (2008).
51. Chung, Y. G. et al. In silico discovery of metal-organic frameworks for precombustion
CO&lt;sub&gt;2&lt;/sub&gt; capture using a genetic algorithm. Sci. Adv. 2, (2016).
52. Allam, R. J. et al. High efficiency and low cost of electricity generation from fossil fuels
while eliminating atmospheric emissions, including carbon dioxide. in Energy Procedia
37, 1135–1149 (2013).
53. D’Alessandro, D. M., Smit, B. & Long, J. R. Carbon dioxide capture: Prospects for new
materials. Angewandte Chemie - International Edition 49, 6058–6082 (2010).
54. Snyder, G. J. & Toberer, E. S. Complex thermoelectric materials. Nat. Mater. 7, 105–114
(2008).
55. Lee, S. W. et al. An electrochemical system for efficiently harvesting low-grade heat
energy. Nat. Commun. 5, 3942 (2014).
56. Raman, A. P., Anoma, M. A., Zhu, L., Rephaeli, E. & Fan, S. Passive radiative cooling
below ambient air temperature under direct sunlight. Nature 515, 540–544 (2014).
57. Hsu, P.-C. et al. Radiative human body cooling by nanoporous polyethylene textile.
Science 353, 1019–1023 (2016).
58. Baetens, R., Jelle, B. P. & Gustavsen, A. Properties, requirements and possibilities of
smart windows for dynamic daylight and solar energy control in buildings: A state-of-the-
art review. Solar Energy Materials and Solar Cells 94, 87–105 (2010).
59. Llordes, A., Garcia, G., Gazquez, J. & Milliron, D. J. Tunable near-infrared and visible-
light transmittance in nanocrystal-in-glass composites. Nature 500, 323–326 (2013).
60. Iacopi, F., Van Hove, M., Charles, M. & Endo, K. Power electronics with wide bandgap
materials: Toward greener, more efficient technologies. MRS Bull. 40, 390–395 (2015).
61. DOE Wide Bandgap Power Electronics Technology Assessment. Available at:
http://energy.gov/sites/prod/files/2015/02/f19/QTR Ch8 - Wide Bandgap TA Feb-13-
2015.pdf. (Accessed: 24th November 2016)
62. Xie, X., Criddle, C. & Cui, Y. Design and fabrication of bioelectrodes for microbial

15
 bioelectrochemical systems. Energy Environ. Sci. 8, 3418–3441 (2015).
63. Sakimoto, K. K., Wong, A. B. & Yang, P. Self-photosensitization of nonphotosynthetic
bacteria for solar-to-chemical production. Science 351, 74 LP-77 (2015).
64. Current and Future Cost of Photovoltaics. (2015). Available at: https://www.agora-
energiewende.de/fileadmin/Projekte/2014/Kosten-Photovoltaik-
2050/AgoraEnergiewende_Current_and_Future_Cost_of_PV_Feb2015_web.pdf.
(Accessed: 24th November 2016)

16
 Technological Forecasting & Social Change 135 (2018) 66–74

Contents lists available at ScienceDirect

Technological Forecasting & Social Change

journal homepage: www.elsevier.com/locate/techfore

China's manufacturing locus in 2025: With a comparison of “Made-in-China T

2025” and “Industry 4.0”
Ling Li
Old Dominion University, Norfolk, VA 23454, USA

A R T I C L E I N F O A B S T R A C T

Keywords: In this study, we have compared Germany's “Industry 4.0” and China's “Made-in-China 2025” and estimated
Made-in-China 2025 China's locus in “Made-in-China 2025”. “Made-in-China 2025” has clear goals, measures and sector focus. Its
Industry 4.0 guiding principles are to enhance industrial capability through innovation-driven manufacturing, optimize the
Emerging economy structure of Chinese industry, emphasize quality over quantity, train and attract talent, and achieve green
Cyber-physical systems (CPS)
manufacturing and environment. Data show that currently China is no longer the lowest–cost labor market; it is
Internet of Things (IoT)
Manufacturing capability
being squeezed by newly emerging low-cost producers such as Vietnam, Cambodia, and Laos. Meanwhile, China
Human capital is not the strongest player in the high-tech arena; well-established industrialized nations, the US, Germany, and
R&D Japan, have all effectively deployed digital technology to create new industrial environments, produce new
Smart factory products, and improve their well-established brands. Having analyzed the data from the World Bank and China's
Collaborative robots National Bureau of Statistics, we find an upward trajectory in China in manufacturing capability development,
research and development commitment, and human capital investment. However, implementing an ambitious
strategic plan such as “Made-in-China 2025” is coupled with challenges. This research helps us understand the
relationship between technological entrepreneurship and socio-economic changes in emerging economies such
as China. Furthermore, the experience accumulated in China can be referenced by both emerging economies and
developed nations to advance their technological entrepreneurship.

1. Introduction value chain and reinvent itself from a world manufacturing production
workshop into a world-class industrialized power (Chinadaily.com.cn,
Over the past decade, China has emerged as one of the most sig- 2015).
nificant manufacturing miracles since the industrial revolution began in In 2013, Germany, a world leading industrialized nation published
Great Britain in the eighteenth century (Li, 2013). By the end of 2012, its “Industry 4.0” strategic plan (Branger and Pang, 2015; GTAI, 2014;
China became a global leader in manufacturing operations and the Lu, 2017; Xu, in press). Well-known for many of its prestigious brands,
second largest economic power in the world. The Made-in-China Volkswagen, BMW, and SAP, to name a few, Germany's leading in-
paradigm has been evidenced by products made in China ranging from dustries have emphasized their innovative strength which allows them
high-tech goods such as personal computers, mobile phones to con- to reinvent themselves time and again. “Industry 4.0” is one more ex-
sumer goods such as air conditioners. According to a report from China ample of Germany's manufacturing strategy to compete in the new
Daily (Chinadaily.com.cn, 2015), in 2014, China produced 286.2 mil- round of industrial revolution that focuses on industrial integration (Xu
lion personal computers, which was about 90% of the world total et al., 2016; Lu, 2016), industrial information integration (Xu, 2016;
(Table 1), 109 billion air conditioners counting for 80% of the world Chen, 2016), manufacturing digitization (Xu et al., 2014b; CPS (Gürdür
total, 4.3 billion energy-saving lamps approximately 80% of the world et al., 2016), Internet of Things (IoT) (Liu et al., 2017; Lai et al., 2017;
total, and its mobile phone production counted for a little over 70% of Xu et al., 2014a), and artificial intelligence.
the world total. In this study we have analyzed China's manufacturing potential in
These statistics indicate that China is good at making many things, 2025 using data from the World Bank and China's National Bureau of
from daily consumer goods to high-tech gadgets, from children's toys to Statistics. We focus on the following two research questions, (i) what is
giant vessels. China is looking forward to climbing new heights in the difference between China's “Made-in-China 2025” plan and
manufacturing. In 2015, China issued a 10-year national plan, “Made- Germany's “Industry 4.0” plan? And (ii) what are the critical factors
in-China 2025”. This strategic plan states China's will to move up the that will affect and support the implementation of the “Made-in-China

E-mail address: [email protected].

http://dx.doi.org/10.1016/j.techfore.2017.05.028
Received 22 February 2017; Received in revised form 12 May 2017; Accepted 19 May 2017
Available online 10 August 2017
0040-1625/ © 2017 Elsevier Inc. All rights reserved.
L. Li Technological Forecasting & Social Change 135 (2018) 66–74

Table 1
Top ten products China makes.
Source: Based on “Top 10 products China manufactures most in the world.” Chinadaily.com.cn, 9/16/2015.

No Products produced in China Volume produced Account for % of world Total

1 Personal computers 286.2 million personal computers in 2014 90.60%

2 Air conditioners 109 billion air conditioners annually 80%
3 Energy-saving lamps 4.3 billion energy-saving lamps 80%
4 Solar cell 21.8 million solar cell kw annually 80%
5 Mobile Phone 1.77 billion mobile phone annually 71%
6 Shoes 12.6 billion number of shoes 63%
7 Cement 1.8 billion tons of cement annually 60%
8 Pork 1.5 million tons of pork annually 49.80%
9 Coal 1.8 billion tons of coal (oil equivalent) annually 48.20%
10 Ship 766 million tons of vessels annually 45.10%

2025” plan? (Schwab, 2016). The 4th industrial revolution is due to significant
Identifying the critical indicators that impact China's economic de- technological development through ICT, cyber-physical systems (CPS),
velopment in the digital era is one of the major contributions of this and IoT (Xu, in press). The vision of economies of scope realized in mass
study. Three critical indicators (manufacturing capability, human ca- customization will lead to a batch size of one while maintaining mass
pital, and R & D) are found to be the major source of China's social- production's economies of scale (Lasi et al., 2014). Paring with ad-
economic change in the past 30 years. Manufacturing capability, vanced digitalization and manufacturing technologies, the 4th in-
human capital, and R & D commitment will continue influence China's dustrial revolution will result in a new fundamental paradigm shift,
implementation of “Made-in-China 2025”. These indicators help us much more than simple manufacturing advancement. The concept of
understand the relationship between technological entrepreneurship the 4th industrial revolution stimulated the introduction of Germany's
and socio-economic changes in emerging economies such as China. “Industry 4.0” plan which was announced in April 2013 (Lasi et al.,
Furthermore, the experience accumulated in China can be referenced 2014) and the “Made-in-China 2025” plan released in May 2015. We
by both emerging economies and developed nations to advance their may say that “Made-in-China 2025” is a Chinese version of “Industry
technological entrepreneurship. 4.0”. Both are described in detail below.

2. Background 2.2. The Made-in-China 2025 plan

2.1. Industry revolutions If the goal of China's economic reform, which began in 1978, was to
lift hundreds and thousands of people from poverty, that goal has been
The world has experienced three industrial revolutions and each has achieved. After 30 years of economic development, China's manu-
benefited mankind and moved society forward. The first industrial re- facturing growth has entered a new era. New opportunities and chal-
volution began in the 1770s in England and spread to the rest of Europe lenges have now emerged. Resources and environmental constraints
and the United States in the 19th century (Stevenson, 2015). Prior to continue to intensify, labor and material costs increase, environmental
that time, products were made in a family workshop by craftsmen and responsibility continues to rise, foreign direct investment flow and ex-
their apprentices. The introduction of the steam engine during the 1st port growth have slowed down. Rethinking and planning manu-
industry revolution provided a source of power to mechanize produc- facturing strategy is inevitable.
tion in factories, which provided jobs for millions of people who mi- In response to the recent global reindustrialization tide and
grated from rural farms to urban areas. Germany's high-tech strategy “Industry 4.0”, the State Council of China
Mass production was the key contribution to the 2nd industrial re- announced the “Made-in-China 2025” Plan in May 2015. This plan laid
volution. When demand for Ford T-model cars came at sonorous pace, out strategic goals for economic development of the next 10 years from
the company experienced a difficult time to keep up with the orders. 2016 to 2025. This blueprint was developed jointly by China's National
Subsequently, the moving assembly line was invented with the in- Development and Reform Commission (NDRC) and by the Ministry of
tensive use of electrical power to produce a large volume of standar- Science & Technology (MOST), with additional inputs from the Ministry
dized products with division of labor and specialized skills at a low unit of Industry and Information Technology (MIIT) and other con-
cost. This revolution provided many well-paid manufacturing jobs and stituencies (State Council of People Republic of China, 2017).
significantly improved ordinary people's living standards. The “Made-in-China 2025” plan is China's industrial development
Then computer technology, information technology, and the wide- master plan for the next 10 years. The plan signals China's intention to
spread digitalization cast a major influence on the digital revolution launch an industrial transformation from labor intensive production to
which is often called the 3rd industrial revolution. Digital technology knowledge intensive manufacturing, and usher in a major breakthrough
has allowed automation of production and service. Manufacturing has at a fast speed. “Made-in-China 2025” is the first-stage of a “three-
evolved from mass production to mass customization, a strategy of phase” grand plan, which will guide China to become a world manu-
manufacturing products with the help of programmable machines to facturing power from the current grand production workshop of the
produce standardized products with some degree of flexibility at the world. The plan focuses on improving the quality of products made in
sub-assembly or final assembly. The third industrial revolution ushered China, creating China's own brands, building a solid manufacturing
in globalization. The concept of a manufacturing supply chain is no capability by developing cutting-edge advanced technologies, re-
longer simply a vertical integration within a company. Instead, a searching new materials, and producing key parts and components of
manufacturing supply chain has become virtual integration around the major products. According to the State Council of People Republic of
world. This revolution benefited more people in the world than that of China (2017), ten industries have been prioritized: information tech-
the previous two industrial revolutions. Wealth has been redistributed nology, high-end numerical control machinery and automation, aero-
among industrialized nations, emerging economies, and developing space and aviation equipment, maritime engineering equipment and
countries. high-tech vessel manufacturing, rail equipment, energy-saving vehicles,
The term of “fourth industrial revolution” emerged recently electrical equipment, new materials, biomedicine and high-

67
L. Li Technological Forecasting & Social Change 135 (2018) 66–74

performance medical apparatus, and agricultural equipment (State 2.4. Comparison of “Made-in-China 2025” and “Industry 4.0”
Council of People Republic of China, 2017).
Based on China's own experience of economic reform which took a Both “Industry 4.0” and “Made-in-China 2025” focus on the new
step-by-step implementation procedure, the “Made-in-China 2025” plan round of industrial revolution and employ manufacturing digitization,
is anticipated to be extended to three phases. Phase One covers year CPS, IoT, and intelligent manufacturing. The core of “Industry 4.0” is
2015 to 2025; during this period, China strives to be included in the list cyber-physical system (CPS) and integration in dynamic value-creation
of global manufacturing power countries. Phase Two covers year 2026 networks. “Made-in-China 2025”, in addition to an action plan of
to 2035; in this period, China will rise to the medium level in the “Internet Plus Industry”, has a broader scope on consolidating existing
world's manufacturing power camp. Phase Three, from year 2036 to industries (Gorkhali and Xu, 2016), promoting diversity and broad-
2049, the time when the People's Republic of China celebrates her 100- ening the range of various industries, enhancing regional cooperation
year anniversary, China dreams to be a leading manufacturing power in (Lima, 2016), using Internet of Things to realize manufacturing without
the world. boundaries (Bi et al., 2014; Xu et al., 2012), innovating new products
China has accumulated rich experience in implementing nation- and improving product quality.
wide strategic plans. Its economic reform in 1978 started with a pilot One of the common priorities of both “Industry 4.0” and “Made-in-
city in Shenzhen Special Economic Zone (SEZ). The special economic China 2025” is to speed up automation and develop collaborative in-
zone experimented with free market business models paired with a state dustrial robots. Advanced robots are a form of complex machines that
planning policy, providing subsidiary, and preferential tax benefits to support and relieve the human operator, improve productivity, increase
foreign firms that were interested in and doing business in China (Li, flexibility, reduce cost, and increase security. Collaborative industrial
2013). This approach has been adopted to implement “Made-in-China robots are good devices to be used in small parts assembly and the
2025”. Ningbo, a port city, was chosen to be the first pilot city to speed sorting of materials (Yang, 2017). Another priority of both plans is
up the construction of its own industrial and manufacturing capability, Internet of Things, Internet of Services, Internet of Media, big data and
collaborate with regional innovation systems, personnel training sys- data analytics. Germany's “Industry 4.0” has emphasized cyber-physical
tems, and policy support systems, to create a healthy ecological en- systems, interoperability (Avanzi et al., 2017), decentralization and full
vironment and achieve diversity in development. Then, a 2nd cohort of virtualization of value chain (Magruk, 2016; Xu, in press).
20 to 30 cities will be selected to join the development effort The new technological innovations, such as cyber-physical systems,
(ifeng.com, 2017). Internet of Things, and virtualization of value chain that stated in
Industry 4.0 and Made-in-China 2025, have already led to socio-eco-
nomic change in emerging economies. Collaborative influence and legal
2.3. Industry 4.0 value of patents, and research and development (R & D) in advanced
nations and emerging economies have demonstrated positive effect on
Germany is well-known for the design and quality of its products better R & D outcomes because diverse knowledge and complementary
(Selko, 2015). Seven companies are the engines of Germany's GDP and competences from different countries generate inventions that may not
source of corporate revenue. Three of its seven giant companies are happen in a single country. Additionally, collaborative R & D promotes
Volkswagen (the biggest corporation in Germany), Daimler (the second the probability of product commercialization and international trading
largest in Germany), and BMW. All three are famous for the design of activities. The increasing research collaborative projects in East Asian
automobiles. Each has substantial exports and global markets as well. countries has internationalized R & D activities cross country borders to
Volkswagen is tied with General Motors Co. for sales in China, the enhance competitive advantages by obtaining foreign resource or
world's largest car market. BWM and Mercedes are the two leading knowledge (Su, 2017). In a study of legal value of international R & D,
luxury car brands in the world (McIntyre, 2012). Germany is strong in Su (2017) analyzed legal value of collaborative research and develop-
electronics and automation industry. Founded in 1847, Siemens is a ment between China and Japan, China and Korea, Japan and Taiwan,
German conglomerate company with principal businesses in industry, Japan and Korea, and Korea and Taiwan, and concluded that interna-
energy, healthcare, and infrastructure. BASF, a chemical firm in Ger- tional R & D collaboration generated higher cross-country patent in-
many, has annual sales exceed $80 billion. Its primary divisions are in fringement (Su, 2017). The virtual supply chain described in Industry
agriculture, chemistry and other practices. In the financial sector, 4.0 will have strong impact on the economic development of emerging
Germany has Deutsche Bank, the largest bank in Europe. Another economies.
Germany's financial group is Allianz, an insurance company, but it ac- A recent score-card of China's infrastructure, scientific and techno-
tually is a financial services holding company that conducts businesses logical achievements lists six areas that have amazing results (Gu,
in the global market. 2017); these six areas include new energy, numerous recent-built mega
“Industry 4.0” which was first introduced in 2013 by the German length bridges, aerospace equipment (e.g. satellite launch), e-business
government, focuses on the concept of the smart factory and cyber- (e.g. Alibaba.com), transportation network (e.g. high speed rail, high
physical systems (CPS) which integrates advanced technologies such as way, tunnels, etc.), and supercomputers. These achievements have
automation, data exchange in manufacturing technology, 3D printing, improved the quality of life of Chinese people and laid a solid foun-
cloud computing, Internet of Things (IoT), and goods and people. This dation for the next level industrialization and economic development.
type of industrial integration (Xu et al., 2016) includes a wide range of However, China's strength mirrors the areas that it needs to improve.
current concepts including cyber-physical systems, smart factory, de- Most of these six developed areas are in the domestic infrastructure
centralized self-organization, new systems in distribution and pro- development scope and are not world-wide recognized brands like
curement, product and service systems that will be individualized, and Germany's Volkswagen, Mercedes, BMW, and Siemens, which produce
corporate social responsibility (Lasi et al., 2014). The core concept of many consumer goods that are sold internationally.
“Industry 4.0” is integration (GTAI, 2014): Integration of the physical Table 2 provides a comparison of “Industry 4.0” and “Made-in-
basic system and the software system, integration with other branches China 2025”. The two plans were issued within a short period of time:
and economic sectors, integration with other industries and industry one in 2013, and the other 2015. The aim of “Industry 4.0” is to en-
types (Fettke, 2013; Hermann et al., 2016). In this sense, “Industry 4.0” hance a new level of organization and control over the entire value
is interpreted as a new level of organization and control over the entire chain of product life cycle, while “Made-in-China 2025” aims at moving
value chain of the lifecycle of products (Ganzarain and Errasti, 2016). from “Made-in-China” to “Designed-in-China”, and making China's
manufacturing strong and innovative. Germany's size of GDP in 2015 is
$3.363 trillion and China's is $11.007 trillion; China's GDP is more than

68
L. Li Technological Forecasting & Social Change 135 (2018) 66–74

Table 2
Comparison of Industry 4.0 and Made-in-China 2025.

Country Germany China

Issue time April 2013 May 2015

Aim Intelligent manufacturing, cyber-physical systems, i.e. From Made-in-China to Designed-in-China; Make China's manufacturing
applying the tools of ICT to production strong & innovative
GDP (in Year 2015) $3.363 trillion $11.007 trillion
Industry, value-added (% of GDP 30.95% 40.92%
in 2015)
Change of GDP 2007 to 2015 − 2.22% 209.89%
Current well-known industry Volkswagen, Daimler, BMW, Siemens, BASF, Deutsche New energy, bridges, aerospace equipment, e-business, transportation network
Bank, Allianz
Strength Well-established brands, global reputation of its Innovation driven, emphasize quality over quantity, emphasize achieving green
products development, emphasize optimizing the structure of Chinese industry
Implementation period 10–15 years 10 years; extend to 2049
Implementing phases Not clearly explained Three Phases
Pilot plan N/A City of Ning Po be the 1st pilot city

3 times larger than that of Germany's. According to World Bank's data, premises of “Industry 4.0” to the focus of “Made-in-China 2025”,
China's industry value added as percentage of GDP in 2015 is about building a solid manufacturing capability via developing cutting-edge
10% more than that of Germany's. This implies that China is the world's advanced technologies is a common goal of these two grand plans.
largest manufacturing workshop, while Germany's financial firms, such The comparison of ‘Industry 4.0’ and ‘Made-in-China 2025’ pre-
as Deutsche Bank and Allianz contribute a good portion to the nation's sented in Table 2 comprehensively discusses the two plans, compares
GDP. and contrasts the similarities and differences of the two plans. From the
Germany's well established brands are all known for their superior discussion in this section, we have answered the first question “what is
design and production quality, and each is a major exporter, or has the difference between ‘Made-in-China 2025’ and ‘Industry 4.0’”?
large operations in other countries. China's score card lists bridges it
built in recent years, aerospace accomplishments, high speed rail,
3. Data and analysis
highways, transportation network, and e-business. These are remark-
able accomplishments. However, China needs to develop sustainable
3.1. Data
brands, reduce its one-time transaction mentality that inhibits China's
reputation as a global manufacturer. As for the time frame, Germany
Operations Management and strategic management literature has
plans to realize the goal of “Industry 4.0” in ten to fifteen years; China
long identified a number of indicators for analyzing a nation's manu-
sets the similar time frame, but this is only the first phase of its grand
facturing capability. These indicators included low-cost production,
industry plan. The other difference between the two plans is that China
differentiation strategies such as new product development and tech-
implements “Made-in-China 2025” with a pilot city, Ning Po, and will
nology innovations, and manufacturing soft-power development such
follow up with more cities, while Germany does not have a pilot study.
as workforce development and technology (Li, 2005, 2013). In this
Implementing “Industry 4.0” requires digital transition which will
study, soft power competencies include human resources, R & D in-
lead to significant innovations. 3D printing, an important part of the
vestment, foreign direct investment net flows, and high-tech exports. To
future of additive manufacturing, is one of the examples of cyber-
make a nation great in manufacturing in a digital world, the twenty-
physical systems. For example, applying 3D technology, manufactures
first century calls for new technology innovation, new product devel-
will manage spare parts and critical parts using different methods. Low
opment, and new market deployment.
demand and none critical parts are manufactured using 3D technology
We use data and information from the World Bank Data (2017) and
at a local site near the end user when there is a demand, while critical
China's National Bureau of Statistics to analyze the potential of ad-
parts are produced at factory with quality specifications and process
vancing the plan from “Made-in-China” to “Designed-in-China”. Table 3
control skills (Jiang et al., 2017). Making spare parts using additive
shows the data that are used in this study as well as the data sources.
manufacturing will simplify logistics and change the current supply
With the availability of data, we group the variables into three major
chain network configuration which can affect social, economic, and
categories: (i) manufacturing capability, (ii) research and development,
supply chain structure in emerging economies which serve as major
and (iii) human capital (Li, 2012). Four items are included in the ca-
manufacturing hub for the industrialized nations now. A 3D focus could
tegory of manufacturing capability: they are GDP, “industry value
help reduce slow moving spare parts in storage, so as to increase effi-
added as a percentage of GDP”, “high technology exports as percentage
ciency and reduce distribution costs (Jiang et al., 2017). Extending the
of manufacturing exports”, and “foreign direct investment net flow”.

Table 3
Data.

Indicator Indicators Data source

Manufacturing capability Industry value added as percentage of GDP The World Bank data (2017)
High technology exports as percentage of manufacturing exports The World Bank data (2017)
Foreign direct investment net flow The World Bank data (2017)
R&D Expenditure on R & D
Number of patent applications The World Bank data (2017)
Human capital Urban and rural residents China's National Bureau of Statistics
Primary and secondary high school graduates China's National Bureau of Statistics
College graduates China's National Bureau of Statistics
Graduate students China's National Bureau of Statistics
Returned study abroad students China's National Bureau of Statistics

69
L. Li Technological Forecasting & Social Change 135 (2018) 66–74

Fig. 1. GDP: China and Germany. Data source: The World Bank Data (2017); data are in
1,000,000 US$. Fig. 3. Data source: The World Bank Data (2017); data are in 1000 US$.

Two items are shown in the R & D category, expenditure on R & D and
number of patent applications. Five indicators are included in the
human capital category. Urban and rural residents, primary and sec-
ondary high school graduates, college graduates, graduate students,
and returned study abroad students.

3.2. Manufacturing capability

Gross Domestic Product (GDP) is a well-accepted measure of a na-

tional's economic performance. By 2007, the size of China's GDP was
the world's third-largest, in 2009 its GDP was ranked the second largest
and the same year, China's GDP surpassed that of Japan and became the
world's 2nd largest economy. China became a member of World Trade
Organization (WTO) in 2001. In 2015, China's GDP was $11 trillion and
was 3.27 times that of Germany's (Fig. 1). Fig. 4. Data source: The World Bank Data (2017); % of manufactured exports.
At the dawn of the 21st century, more industrialized nations made
strategic turns to focus on high-tech products and technology, and re-
number stabilizes in the range of 25.3% to 27.5% of the percentage of
duce the proportion of labor-intensive, low-value-add, low-profit-
manufactured export (Fig. 4). The reduction in high-tech exports can be
margin production. This decision led to a global manufacturing re-
explained in a couple of ways. First, China's domestic consumption has
configuration (Li, 2013). The epicenter of manufacturing moved from
increased, and second, the ripple effect of reduction in foreign invest-
industrialized nations in North America and Europe to Southeast Asian
ment.
countries and South American countries. This industry evolution can be
illustrated using a couple indicators, GDP, percentage of GDP as in-
dustry value added, foreign direct investment net flow, and high-tech- 3.3. Research and development
nology exports.
In China, the percentage of GDP as industry value added reaches its Two items are shown in the R & D category: expenditure on R & D
peak at the value of 47.77% in year 2006 (Fig. 2). In 2015, the per- and the number of patent applications. According to the World Bank,
centage of GDP as industry value added dropped to 40.93%. The R & D expenditures have been made on creative work to increase
highest foreign direct investment net flow happened in 2013 with a knowledge, including knowledge of the humanities, culture, and so-
value of $290.928 billion and then curved down in 2014 and 2015 ciety, and the use of knowledge for new applications. R & D covers basic
(Fig. 3). research, applied research, and experimental development. In the
The largest high-technology exports as percentage of manufactured period of 2002–2015, the amount of investment in R & D has increased
exports occurred in 2006, accounts for 30.52%, and then the export dramatically. In 2015, China spent 14,169.88 million Yuan on research

Fig. 2. Data source: The World Bank Data (2017); data are in %
of GDP.

70
L. Li Technological Forecasting & Social Change 135 (2018) 66–74

and development, which is 10 times as much as that of in 2002. During

the period of 2002–2015, the expedition on R & D has increased by
1000% (Fig. 5).
Patent applications are defined as “applications filed through the
Patent Cooperation Treaty procedure or with a national patent office
for exclusive rights for an invention - a product or process that provides
a new way of doing something or offers a new technical solution to a
problem (the World Bank Data).” According to World Bank Data re-
trieved in 2017, the number of patent applications filed by Chinese
citizens in 2014 reached more than 800,000 (Fig. 6), an increase of
1912.6% compared with 2002.
Both Figs. 5 and 6 show an S curve. The S curve usually describes
the growth of a variable over time. In addition to describing growth, the
S curve predicts the performance of a company or a product over a Fig. 5. Data sources: China's National Bureau of Statistics.

period of time. The growth curve for both expenditure on R & D and
patent application started in 2002 with a modest growth, and then a
rapid growth. The R & D expenditure and patent applications curves are
still in a growth momentum.

3.4. Human capital

To realize the goal of “Made-in-China 2025”, human resource de-

velopment is a key. Human resource practices with commitment or-
ientation help build an adequate work environment to enhance human
capital creativity and innovativeness. Human capital is critical to fa-
cilitate the appearance of exploitative and explorative practice (Popa
et al., 2017).
Research and innovation rely on human talent, therefore, human
capital should be nurtured by sound education. People who operate
cyber physical systems (“Industry 4.0”) or Internet-Plus manufacturing
Fig. 6. Data source: The World Bank Data (2017).
systems in “Made-in-China 2025” need to be educated to be able to run
the system. Revitalization through education is a critical step to move
up to further growth. Some authors hypothesized that Chinese manu-
facturers would keep producing low value, labor-intensive products
while their counterparts in the industrialized countries would dominate
the high-tech industries (Filippetti and Innocenti, 2012). The graphs
shown in Figs. 7–9, and 10 refute this conjecture. To China, the global
competition in the twenty-first century is no longer in the area of low-
cost labor production. It has evolved into a stage of Internet-Plus
Manufacturing that competes for talent, human capital, and intellectual
resources (Li, 2013). China's goal is to be a power of Designed-in-China.
In 2002, the urban residents of China were 39% of the total popu-
lation and the rural population was 61% (Fig. 7). By 2010, half of the
population was urban residents and the other half rural residents. 2011
ushered in the first year when the urban population exceeded that of
the rural (51% versus 49%). In 2015, China's urban population reached Fig. 7. Data sources: Chinese National Bureau of Statistics.
56%. The rural population was a major source of low-cost labor. In the
past 30 years, China had a good supply of “farmer-workers” or migrant
workers who served as a pool of flexible workforce that covered peak
demand seasons and smoothed ripple effects of demand forecast errors.
Farmer-workers also encompass those who “worked in the factories but
had residence registration in the rural areas”. They worked in the city
when factories needed them and returned to their farms when there was
no need of their labor in the factories,” (Li, 2013). The migration of
rural residents to urban areas reduces the supply of low cost labor. In
turn, China has since had to adjust its manufacturing strategy that fo-
cused on low-cost labor.
Looking at the overall picture of China's education in period
2002–2015, the number of primary school students is decreasing stea-
dily. This age group has decreased about 39% in the period 2002–2015
(Fig. 8). The number of senior secondary high school students has in-
creased from 2002 to 2008, and then decreased about 4.59% from 2009 Fig. 8. Data sources: Chinese National Bureau of Statistics.
to 2015. These statistics indicate that China's one-child policy is im-
pacting the size of K-12 enrollment.
On the other hand, the number of students who have completed
college and graduate education has increased dramatically. In 2015,

71
L. Li Technological Forecasting & Social Change 135 (2018) 66–74

4.1. Manufacturing capability

Manufacturing capabilities that include GDP, percentage of GDP as

industry value added, foreign direct investment net flow, and high-
technology exports are key competencies to successfully implement
“Made-in-China 2025”. In the past three decades, China has trans-
formed itself from an agriculture-based economy into an industrial
machine that has processed and assembled about 90% of the manu-
factured goods in the world (Li, 2013). The data from the World Bank
indicate that although China's GDP is continuously increasing (Fig. 1),
the foreign direct investment net flow (Fig. 3) and high-technology
exports (Fig. 4) started to phase down in 2013. These statistics signal
that foreign investors have gradually pulled out of the Chinese market.
As a result, China has relied more on its own investment and domestic
Fig. 9. Data sources: Chinese National Bureau of Statistics. consumption of high-tech products. Since 2011, the percentage of GDP
as industry value added has decreased from 47.56% in 2006 to 40.93%
China graduated 6,808,866 four-year college students (Fig. 9), which is in 2015. The absolute dollar value may not change due to the increase
a little more than the size of the total labor force of Denmark and in GDP but the proportion has changed. The other possible explanation
Norway combined (the World Bank Data).1 In 2015, 1,911,406 grad- is that China has increased service operation.
uate students finished their degree in Chinese universities; this number
is 282 times the size of the graduate student population who completed 4.2. Research and development
degrees in 2002. In the past 30 years, many Chinese students and
scholars have studied abroad. They acquired international experience, Both ‘expense on research and development’ (Fig. 5) and ‘number of
improved language capabilities, and developed cross-cultural commu- patent applications’ (Fig. 6) take an S curve shape, starting in 2002 with
nication skills; all of these are fundamental qualities to succeed in the a modest growth, and then a rapid upward increase in 2013. R & D
global economy. This group of educated talent forms a unique category investment and activities are an essential foundation to build China's
of human resource, which has been named “the returned study-abroad brands. To date, Chinese manufacturing firms established fewer well-
students”. The number of students in this group increased about 2190% known brands than its counterparts in Germany, the United States, and
in the period 2002–2015 (Fig. 9). These globally minded, highly edu- Japan. The increase in investments and patent applications provides
cated people, who have promoted economic, technological, cultural noticeable evidence indicating that China is moving away from low-
and social exchanges, have benefited China and the rest of the world in tech content manufacturing, and strives to support the goal of in-
the past 30 years. novation, high-tech, and integration.
Based on the statistics presented above, China won't have as many Manufacturing is the core component of a nation's economy. Since
young people who are willing to take low cost blue-collar jobs in the its economic reform in 1978, China's manufacturing industries have
next ten years. China is a country which values white-collar profes- progressed not only in labor intensive industries such as footwear and
sional career. The 21st century is an era when industries are competing apparel, but also in advanced technology industries such as electronic
for qualified professionals and talents. In order to sustain the growth, appliances and personal computers. As a result, China's manufacturing
manufacturing industries, especially, high-tech industries need quali- industry experienced internal growth. There is a 370% increase of net
fied professionals from around the world to exploit their expertise and foreign direct investment in China from the year 2002 to the year 2015
creative potential. (Fig. 3). In the same period, Germany had a decrease of 9.75% in the
Data presented in this section help us answer the second question, same category. The effects of high net inflow of foreign investment are
“what are the critical indicators that will affect the implementation of on high value-added industrial output and increased export volume.
the ‘Made-in-China 2025’ plan”? The critical indicators are manu- China's value-added manufacturing and mining (as percent of GDP) was
facturing capability, research and development, and human capital. 40.9% in 2015, while Germany's was 30.5%. A noticeable phenomenon
is that China has kept a stable rate of high-tech exports as a percentage
of manufacturing exports in the period 2008 to 2015 (Fig. 4). In this
4. Discussion period, China's high-tech export accounts for 25.75% of its total ex-
ports; conversely, Germany's high-tech export was 16.7%.
For China, the challenge is not merely continuing the trajectory of As stated by a few authors (Zhang et al., 2016), in 2010, China's
“Made-in-China” from big to mega, rather, it is to advance from “Made- manufacturing outputs accounted for 19.8% of the world total, and in
in-China” to “Designed-in-China” and “Innovated-in-China” as 2014, China's manufacturing outputs accounted for 25% of the world
Germany and the United States have successfully done for decades. In total.
“Made-in-China 2025”, the goals and implementation ideas have been
laid out, including advancing cutting edge new technologies through 4.3. Human capital
investment in R & D, increasing intellectual property accumulation,
creating distinct technical standards, leveraging access to the Chinese China is a nation well-known for its passion for education. This
market in inviting foreign investors, enhancing Internet Plus industry, mentality coincides with the notion of the importance of workforce in
and fostering manufacturing and industrial development. Three major its “Made-in-China 2025”. Workforce availability and competence are
critical factors are identified; they are manufacturing capabilities, re- key attributes that contributed to manufacturing competitiveness.
search and development, and Human Capital. This finding provides an When China started its economic reform in 1978, the urban residents
answer to our 2nd question: the critical factors that will affect and counted for only 18% of the total population (Li, 2013). In 2015, the
support the implementation of the “Made-in-China 2025” plan. percentage of urban residents exceeds that of rural residents by a little
more than 12% in China.
1
According to the definition of the World Bank, total labor force comprises people ages
Educational institutions have cultivated and trained hundreds and
15 and older who meet the International Labor Organization definition of the econom- thousands of college graduates who have contributed to the economic
ically active population. development. The statistics about college graduates, graduate students,

72
L. Li Technological Forecasting & Social Change 135 (2018) 66–74

and returned study abroad students demonstrate China's commitment and the other three are NAFTA countries.
to investing in education, research, technology, and innovation. The world is anticipating the 4th Industrial revolution which is
Primary school students (equivalent to elementary school in the US) propelled by advanced technology. During the process of comparing
have continued to decrease since 2000, while the number of secondary Industry 4.0 and Made-in-China 2025, we find a few research gaps that
high graduates (equivalent to high school in the US) remains stable. worth more research effort. First, more work needs to be done to un-
These numbers imply that the low-cost labor pool is shrinking and will derstand the managerial implications and the strategic advantages of
continue to do so. Historically, the rural population has been a major collaborative scientific activities and innovation that involve more than
source of low-cost labor. Today, replacing the shortage of low-cost labor one country. Second, the chasm between the industrial and societal
is automation (Liu et al., 2016, Mao et al., 2016; Xu et al., 2012, 2014a; impact of collaborative innovation deserves more research. Third, fu-
Yang et al., 2016; Zhai et al., 2016), industry robots, artificial in- ture research may differentiate international R & D results from dif-
telligence, machine learning; all of these are the cornerstone of the 4th ferent business activities, e.g. manufacturing outsourcing, R & D, col-
industry revolution, and the core of “Made-in-China 2025”. laboration, licensing, and join-venture. Finally, in terms of the
Returned study abroad students and scholars is a unique category geographical coverage, more research is needed from regions beyond
reported by China's Bureau of Statistics. This group increased by 2179% those major emerging economies such as BRIC countries (Brazil, Russia,
in the period 2002–2015 (Fig. 9). An ability to attract the world's elite India, and China). New emerging economies such as South Africa,
talents is a symbol of a nation's power and competitiveness. These Vietnam and Hungary that have contributed to the world economy in
highly educated people who have promoted economic and social ex- recent years require more attention.
changes will continue to benefit China and the rest of the world (Li,
2013). The United States has successfully attracted millions of world's References
talents to build its manufacturing and high-tech industries; for example,
eBay was founded by Pierre Omidyar, a French-born Iranian-American ifeng.com, 2017. Made-in-China 2025 Pilot Cities Extended to 20 (中国制造2025试点城市
entrepreneur in 1995 and one of the founders of PayPal, “Max” Lev- 将扩围至20余个), 2017–01-12 07. Retrieved on Feb. 6, 2017. http://finance.ifeng.
com/a/20170112/15139387_0.shtml.
chyn, is a Ukrainian born American. In order to sustain their growth, Avanzi, D., Foggiatto, A., Santos, V., Deschamps, F., Loures, E., 2017. A framework for
Chinese industries need qualified professionals from around the world interoperability assessment in crisis management. J. Ind. Inf. Integr. 5, 26–38. http://
to contribute to their expertise and creative potential to implement dx.doi.org/10.1016/j.jii.2017.02.004.
Bi, Z., Xu, L., Wang, C., 2014. Internet of things for enterprise systems of modern man-
“Made-in-China 2025” blueprint. The dramatically growing number in ufacturing. IEEE Trans. Ind. Inf. 10 (2), 1537–1546. http://dx.doi.org/10.1109/TII.
the categories of college graduates and returned study abroad students 2014.2300338.
shows a promising picture of China's capability of attracting the world's Branger, J., Pang, Z., 2015. From automated home to sustainable, healthy and manu-
facturing home: a new story enabled by the Internet-of-Things and Industry 4.0. J.
top scientists and innovators. Manag. Anal. 2 (4), 314–332. http://dx.doi.org/10.1080/23270012.2015.1115379.
This ambitious strategic plan is coupled with challenges that China Chen, Y., 2016. Industrial information integration-a literature review 2006–2015. J. Ind.
is facing (Zhang et al., 2014) On the one hand, China is no longer the Inf. Integr. 2, 30–64. http://dx.doi.org/10.1016/j.jii.2016.04.004.
Chinadaily.com.cn, 2015. Top 10 Products China Manufactures Most in the World, 9/16/
lowest–cost labor market; it is being squeezed by newly emerging low-
2015. Retrieved on Feb. 6, 2017. http://www.chinadaily.com.cn/business/2015-09/
cost producers such as Vietnam, Cambodia, and Laos. On the other 16/content_21886983.htm.
hand, China is not the strongest player in the high-tech arena. Ad- Deloitte, 2016. 2016 Global Manufacturing Competitiveness Index. Retrieved on Feb. 6,
vanced industrialized economies, the US, Germany, and Japan, have all 2017. https://www2.deloitte.com/global/en/pages/manufacturing/articles/global-
manufacturing-competitiveness-index.html.
effectively deployed digital technology to create new industrial en- Fettke, P., 2013. Big Data, Industrie 4.0 und Wirtschaftsinformatik. Vortrag vom 25.
vironments, produce new products and improve their well-established Oktober 2013 anlässlich der Ernennung zum DFKI Research Fellow. http://www.
brands. A recent New York Times' report provides some evidence of dfki.de/web/ueber/research-fellows/131031_rf_vortrag_fettke_extern.pdf (Accessed
2014-04-02).
China's high-tech capability (Markoff and Rosenberg, 2017). A well- Filippetti, A., Innocenti, S., 2012. 10 Years after: Taking Stock of China's WTO
known Artificial Intelligence researcher gently reminded Microsoft that Membership, Open Democracy, 11 May 2012. Retrieved on Feb. 6, 2017. http://
Baidu (a Chinese version of Google) had achieved similar accuracy in www.opendemocracy.net.
Ganzarain, J., Errasti, N., 2016. Three stage maturity model in SME's toward industry 4.0.
artificial intelligence with the Chinese language two years earlier than J. Ind. Eng. Manag. 9 (5), 1119–1128.
their US counterparts. Chinese companies and the nation's government Gorkhali, A., Xu, L., 2016. Enterprise application integration in industrial integration: a
laboratories are making major investments in artificial intelligence. In literature review. J. Ind. Integr. Manag. 1, 1650014. http://dx.doi.org/10.1142/
S2424862216500147.
summary, enhancing manufacturing capability, research and develop-
GTAI (Germany Trade & Invest), 2014. Industries 4.0-Smart Manufacturing for the Future.
ment investment, and human capital give China a head start to trans- Gu, Y., 2017. Editor: This 6 “Chinese Business Card” Shock the World! XinHua News.
form the “Made-in-China 2025” plan into action. https://news.xinhuanet.com/politics/2017-02/04/c_1120409513.htm&prev=
search (Feb. 4).
Gürdür, D., El-Khoury, J., Seceleanu, T., Lednicki, L., 2016. Making interoperability
5. Conclusions visible: data visualization of cyber-physical systems development tool chains. J. Ind.
Inf. Integr. 4, 26–34. http://dx.doi.org/10.1016/j.jii.2016.09.002.
In this study, we compared China's “Made-in-China 2025” with Hermann, M., Pentek, T., Otto, B., 2016. Design principles for industrie 4.0 scenarios. In:
2016 49th Hawaii International Conference on System Sciences, 2016, pp.
Germany's “Industry 4.0” and estimated China's locus in Made-in-China 3928–3937.
2025. “Made-in-China 2025” has clear goals, measures and sector focus. Jiang, R., Kleer, R., Piller, F.T., 2017. Predicting the future of additive manufacturing: a
Its guiding principles are to enhance industrial capability through in- Delphi study on economic and societal implications of 3D printing for 2030. Technol.
Forecast. Soc. Chang. 117, 84–97.
novation-driven manufacturing, emphasize quality over quantity, Lai, C., Jackson, P., Jiang, W., 2017. Shifting paradigm to service-dominant logic via
achieve green development, optimize the structure of Chinese industry, Internet-of-Things with applications in the elevators industry. J. Manag. Anal. 4 (1),
and nurture human talent. Both plans have recognized the use of the 35–54. http://dx.doi.org/10.1080/23270012.2016.1259967.
Lasi, H., Fettke, P., Kemper, H., Feld, T., Hoffmann, M., 2014. Industry 4.0. Bus. Inf. Syst.
Internet of Things in manufacturing, employing digital networking of Eng. 6 (4), 239–242.
production to create smart manufacturing systems within and beyond Li, L., 2005. Assessing intermediate infrastructural manufacturing decisions that affect a
the factory to both customers and suppliers, and creating a highly re- firm's market performance. Int. J. Prod. Res. 43 (12), 2537–2552.
Li, L., 2012. Effects of enterprise technology on supply chain collaboration: analysis of
sponsive, innovative, and competitive global manufacturing landscape. China-linked supply chain. Enterp. Inf. Syst. 6 (1), 55–77.
A recent study on global manufacturing competitiveness by Deloitte Li, L., 2013. The path to made-in-China: how it was done and future prospects. Int. J.
Global predicts that the U.S. will be the most competitive manu- Prod. Econ. 146 (1), 4–13.
Lima, R., 2016. Economic growth and human Capital in the Post-Knowledge era: a focus
facturing nation in the world in 2020, followed by China, Germany,
on positive externalities and spillover effects of knowledge in Italy and the emergency
Japan, India, South Korea, Mexico, Taiwan, Canada and Singapore of the less developed areas. J. Ind. Integr. Manag. 1, 1650010. http://dx.doi.org/10.
(Deloitte, 2016). Six out of the ten nations are in Asia, one in Europe, 1142/S242486221650010X.

73
L. Li Technological Forecasting & Social Change 135 (2018) 66–74

Liu, Y., Han, W., Zhang, Y., Li, L., Wang, J., Zheng, X., 2016. An Internet-of-Things so- 10 (4), 2233–2248. http://dx.doi.org/10.1109/TII.2014.2300753.
lution for food safety and quality control: a pilot project in China. J. Ind. Inf. Integr. 3, Xu, L., Wang, C., Bi, Z., Yu, J., 2012. AutoAssem: an automated assembly planning system
1–7. http://dx.doi.org/10.1016/j.jii.2016.06.001. for complex products. IEEE Trans. Ind. Inf. 8 (3), 669–678.
Liu, F., Tan, C., Lim, E., Choi, B., 2017. Traversing knowledge networks: an algorithmic Xu, L., Wang, C., Bi, Z., Yu, J., 2014b. Object-oriented templates for automated assembly
historiography of extant literature on the Internet of Things (IoT). J. Manag. Anal. 4 planning of complex products. IEEE Trans. Autom. Sci. Eng. 11 (2), 492–503. http://
(1), 3–34. http://dx.doi.org/10.1080/23270012.2016.1214540. dx.doi.org/10.1109/TASE.2012.2232652.
Lu, Y., 2016. Industrial integration: a literature review. J. Ind. Int. Manag. 1, 1650007. Yang, M., 2017. China and Germany Cooperate on Collaborative Robots.
http://dx.doi.org/10.1142/S242486221650007X1601001. Chinadaily.com.cn, Updated: 2017-01-16. Retrieved on Feb. 6, 2017. http://www.
Lu, Y., 2017. Industry 4.0: a survey on technologies, applications and open research is- chinadaily.com.cn/m/liaoning/investinshenyang/2017-01/16/content_27968638.
sues. J. Ind. Inf. Integr. 6. http://dx.doi.org/10.1016/j.jii.2017.04.005. htm.
Magruk, A., 2016. Uncertainty in the sphere of the INDUSTRY 4.0 – potential areas to Yang, B., Stankevicius, D., Marozas, V., Deng, Z., Liu, E., Lukosevicius, A., Dong, F., Xu,
research. Bus. Manag. Educ. 14 (2), 275–291. L., Min, G., 2016. Lifelogging data validation model for internet of things enabled
Mao, J., Zhou, Q., Sarmiento, M., Chen, J., Wang, P., Jonsson, F., Xu, L., Zheng, L., Zou, healthcare system. IEEE Trans. Syst. Man Cybern. Syst.. http://dx.doi.org/10.1109/
Z., 2016. A hybrid reader transceiver design for industrial Internet of Things. J. Ind. TSMC.2016.2586075. (online published).
Inf. Integr. 2, 19–29. http://dx.doi.org/10.1016/j.jii.2016.05.001. Zhai, C., Zou, Z., Chen, Q., Xu, L., Zheng, L., Tenhunen, H., 2016. Delay-aware and re-
Markoff, J., Rosenberg, M., 2017. China gains on the U.S. in the artificial intelligence liability-aware contention-free MF-TDMA protocol for automated RFID monitoring in
arms race. New York times, February 4, 2017. Retrieved on Feb. 6, 2017. http://cn. industrial IoT. J. Ind. Inf. Integr. 3, 8–19. http://dx.doi.org/10.1016/j.jii.2016.06.
nytimes.com/world/20170204/artificial-intelligence-china-united-states/. 002.
McIntyre, D., 2012. The Seven Companies That Power Germany's Amazing Economy. Zhang, Z., Liu, S., Tang, M., 2014. Industry 4.0: challenges and opportunities for Chinese
http://247wallst.com/investing/2012/10/18/the-seven-companies-that-power- manufacturing industry. Tech. Gaz. 21 (6), III–IIV.
germanys-amazing-economy/. Zhang, X., Peek, W., Pikas, B., Lee, T., 2016. The transformation and upgrading of the
Popa, S., Soto-Acosta, P., Martinez-Conesa, I., 2017. Antecedents, moderators, and out- Chinese manufacturing industry: based on “German Industry 4.0”. J. Appl. Bus. Econ.
comes of innovation climate and open innovation: an empirical study in SMEs. 18 (5), 97–105.
Technol. Forecast. Soc. Chang. 118, 134–142.
Schwab, K., 2016. The fourth industrial revolution. In: World Economic Forum, 2016, pp. Dr. Ling Li is the Chair of the Department of Information Technology and Decision
51–59. Sciences, Coordinator of Maritime and Supply Chain Management discipline at Strome
Selko, A., 2015. Top 10 manufacturing countries in 2020. In: Industry Week, Dec 9, 2015, College of Business, Old Dominion University, USA. She is university professor and a
Retrieved on Feb 10, 2017. http://www.industryweek.com/competitiveness/top-10- fellow of APICS (the Association for Operations Management).
manufacturing-countries-2020#slide-0-field_images-192471.
State Council of People Republic of China, 2017. Building a World Manufacturing Power-
premier and ‘Made in China 2025’ Strategy, January 31, 2017. Retrieved on Feb 10,
In tribute to her research records, she was awarded the title of Eminent Scholar. She has
2017. http://english.gov.cn/premier/news/2017/01/29/content_
published over 100 peer-refereed research articles in high quality journals, three single-
281475554068056.htm.
authored books on supply chain management and logistics, encyclopedia articles, busi-
Stevenson, W.J., 2015. Operations Management, 12th ed. McGraw-Hill, New York, NY.
ness cases, conference proceeding papers, and book chapters. She is the winner of many
Su, H., 2017. Collaborative and legal dynamics of International R & D - evolving patterns
awards.
in East Asia. Technol. Forecast. Soc. Chang. 117, 217–227.
World Bank Data, 2017. http://data.worldbank.org/ (Accessed Dec. 20, 2016).
Xu, L., 2016. Inaugural issue editorial. J. Ind. Inf. Integr. 1, 1–2. http://dx.doi.org/10.
She serves as the First Secretary (officer) of International Federation for Information
1016/j.jii.2016.04.001.
Processing TC8 WG 8.9, an organization which is under the auspices of UNESCO. She is
Xu, L., 2017. Industry 4.0: state of the art and future trends. Int. J. Prod. Res (in press).
Area Editor of Systems Research and Behavioral Science Journal, Associate Editor of Journal
Xu, L., Cai, L., Zhao, S., Ge, B., 2016. Editorial: inaugural issue. J. Ind. Integr. Manag. 1,
of Management Analytics, and an Editorial Board Member of International Journal of
1601001. http://dx.doi.org/10.1142/S2424862216010016.
Integrated Supply Management.
Xu, L., He, W., Li, S., 2014a. Internet of things in industries: a survey. IEEE Trans. Ind. Inf.

74
 Research Policy 50 (2021) 103943

Contents lists available at ScienceDirect

Research Policy
journal homepage: www.elsevier.com/locate/respol

The role of pre-innovation platform activity for diffusion success: Evidence

from consumer innovations on a 3D printing platform
Jörg Claussen a, b, 1, Maria A. Halbinger c, 1, *
a
Ludwig-Maximilians-Universität München, Fakultät für Betriebswirtschaft, Institut für Strategie, Technologie und Organisation, Kaulbachstr. 45, 80539 Munich,
Germany
b
Copenhagen Business School, Department of Strategic Management and Globalization, Room: KIL/14.A, 2000 Frederiksberg, Denmark
c
Baruch College Narendra P. Loomba Department of Management, Baruch College’s Zicklin School of Business, 55 Lexington Ave. B9-240, New York, NY 10010, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Digital platforms are becoming increasingly important for household sector innovators that seek support for the
Household sector innovation innovation process and that want to make innovations available to large audiences. Innovation development and
Diffusion diffusion is especially challenging for first-time innovators as they cannot build on experiences from prior in­
Platforms
novations. We argue that first-time innovators can increase the diffusion success of their innovations by engaging
3D printing
in pre-innovation platform activities. We use the context of the 3D printing platform Thingiverse to show that a
consumer’s pre-innovation platform activity increases innovation diffusion success and that frequency, quality
and relatedness of a consumer’s pre-innovation platform activity promotes this effect. We find support that
innovation quality, the use of recombinant innovations, and innovation documentation are three mechanisms
through which pre-innovation platform activities translate into higher diffusion success of consumers’ first
innovation.

1. Introduction they provide opportunities for observational learning and for direct
exchange with other platform participants. These opportunities are
Digital platforms are economically important (Dahlander and Mag­ especially valuable for first-time innovators because they cannot build
nusson, 2005; McIntyre and Srinivasan, 2017; Srinivasan and Venka­ on experiences from prior innovations.
traman, 2018). Because of the increasing availability of advanced Further, digital platforms also help diffusing consumer innovation
technologies, design and communication tools (Baldwin and von Hip­ and this is of relevance for those consumers that are motivated to ach­
pel, 2011), digital platforms are also becoming more and more impor­ ieve social recognition or status via innovation diffusion. Individuals
tant for household sector innovators, i.e. individual consumers who generally seek recognition (Lindenberg, 2013) and a high number of
innovate for themselves in their discretionary time (von Hippel, 2017). likes or views on innovations diffused on a platform can be a symbol of
While prior literature found that many household sector innovators are social recognition in the digital space (Burtch et al., 2019). Diffusing
self-rewarded and thus, not interested in actively diffusing their inno­ consumer innovation is important because it has the potential to
vation (de Jong et al., 2015), an increasingly important subgroup of improve social welfare (Gambardella et al., 2016), and the ecosystem of
household sector innovators appears to engage in digital platforms to digital platforms could represent a vehicle through which household
learn and, notably, is interested in successfully diffusing their sector innovation can be made available to a broader audience.
innovations. In this study, we argue that first-time innovators can increase the
While some of these particular consumer innovators use multipur­ diffusion success of their innovations by engaging in pre-innovation
pose platforms such as YouTube or Pinterest for their do-it-yourself ac­ platform activities. Specifically, we propose that frequency, quality,
tivities, others use more specialized platforms such as Github, and relatedness of pre-innovation platform activity will positively shape
Instructables or Thingiverse to develop and share their innovative de­ consumers’ likelihoods of successfully diffusing their first innovations.
velopments. Platforms such as these support the innovation process as We examine the relationship between pre-innovation platform

* Corresponding author.
E-mail addresses: [email protected] (J. Claussen), [email protected] (M.A. Halbinger).
1
Both authors contributed equally to the development of the manuscript with different focus areas.

https://doi.org/10.1016/j.respol.2020.103943
Received 22 April 2019; Received in revised form 3 February 2020; Accepted 6 February 2020
Available online 18 March 2020
0048-7333/© 2020 Elsevier B.V. All rights reserved.
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

activity and innovation diffusion success in the context of the largest 2018). Independent of functional or creative/design value, both types of
digital platform on 3D printing, Thingiverse. Similar to open source innovations are self-rewarded, for example because of a personal need,
software platforms where locally dispersed members interact online, the inherent enjoyment of the activity, or because consumers want to
members on 3D printing platforms can easily exchange comments on learn (von Hippel, 2017).
other members’ work, as well as share and further develop designs in the We consider an innovation successfully diffused when other con­
form of digital files (Kyriakou et al., 2017; Stanko, 2016). Digital files sumers perceive a new development or design as valuable and verifiably
function as blueprints for innovations and can be transformed into 3D, demonstrate their utilization and appreciation of the innovation via
thereby providing enhanced space for innovation and collaboration. positive feedback – and thus, we go beyond pure innovation adoption
Digital files can be created and shared more cost efficiently, faster and (de Jong et al., 2018). Considering something as successful when it is
easier than physical prototypes. The context of Thingiverse is thus valuable to others, i.e. when it passes social evaluation such as accept­
particularly advantageous to explore the factors that enable and shape ing, adopting or using an innovation, has long been a consideration in
successful diffusion of consumers’ first innovations given that members’ innovation research (Landau et al., 1986). Yet, while in the offline
platform activities prior, during and after the first innovation can be world, diffusion success may be accomplished by deliberate or
observed. non-deliberate peer-to-peer sharing, and/or through commercialization
Our results show that platform activity prior to developing an of the innovation via an incumbent firm or via a newly started venture
innovation has a positive effect on innovation diffusion success. (for an overview of diffusion pathways, see de Jong et al., 2015), in the
Furthermore, we show that the frequency, the quality and relatedness of online world, diffusion success may go beyond these more traditional
these pre-innovation platform activities are key factors strengthening diffusion pathways as we explain in the following.
this said effect. We are not only able to show these effects, but are also
able to empirically test mechanisms through which these effects work. 2.2. Self-rewards and knowledge sharing as basis for innovation on
The three mechanisms through which pre-innovation platform activities platforms
drive innovation diffusion success are 1) quality of the first innovation,
2) recombining other platform members’ work into recombinant inno­ In times where consumers are increasingly using advanced digital
vation, and 3) innovation documentation. Finally, we show that diffu­ technologies and social media, it occurs that these individuals feel
sion success on the platform develops in stages from awareness and rewarded by reputation, status and social signals on digital platforms.
evaluation to implementation. The relevance of status and awards has been discussed in the emerging
Our study offers important contributions to research on household literature on user generated content (e.g. Burtch et al., 2018; Gallus,
sector innovation as we highlight the role of digital platforms in pro­ 2016). Particularly the recognition by peers has been found to be an
moting innovation diffusion success for a particular group of household important motivation for individuals to engage on digital platforms
sector innovators that can be expected to become increasingly important (Lerner and Tirole, 2002). This conveys the idea that on digital plat­
in the future. Motivated to learn and diffuse, these consumer innovators forms, and in the online world more generally, the appreciation by
can use platforms to facilitate both the development and the diffusion of others via social signals functions as a (peer) reward and motivator
innovation. Our study thus, on the one hand, provides insights on the (Burtch et al., 2019), and represents an indicator to the focal consumer
transition from platform participants to innovators. On the other hand, (and others) that an innovation is successfully diffused.
our results indicate that merely sharing innovations on platforms is The significance of social signals (e.g. likes, comments, re-use of
necessary but not sufficient for diffusion success. Instead, being active contributions) as a reward for platform users has also been supported by
and engaging on the platform via commenting, copying and familiar­ anecdotal evidence in interviews of Thingiverse platform users.
izing themselves with others’ work will help consumers generating first Repeatedly our respondents stated they select into the setting of the
innovations that are appreciated and re-used by other platform mem­ Thingiverse platform because of the possibility to learn from others to
bers. We hence contribute to current research on consumer innovation solve innovation-related problems. But more importantly, they stated to
because our findings shed light on the genesis and diffusion of innova­ use the Thingiverse platform because of their intention to receive
tion in the household sector, and ultimately, highlight a way to solve the acknowledgement via likes or comments on their shared developments
prevailing diffusion problem. We thus provide insights into the differ­ or, because others find their innovations valuable enough to re-use
ences between consumers who successfully develop and diffuse inno­ them. These social signals are a form of acknowledgement that they
vation and those who do not. find rewarding and socially gratifying.
On platforms knowledge is distributed, created and exchanged and
2. Household sector innovations on digital platforms this provides the basis for the development of innovations, their diffu­
sion and usage (Autio et al., 2013; Dahlander and Frederiksen, 2012).
In this section, we first define household sector innovation and Members of platforms typically share ideas, provide feedback on exist­
diffusion success and then explain how consumers in the Digital Age are ing designs and exchange information on new solutions (von Krogh and
motivated by social signals on platforms, thereby indicating a new form von Hippel, 2006) thereby enabling important knowledge spillovers free
of self-reward in the household sector. We conclude the section with a of charge (Harhoff et al., 2003). The diversity of this information is core
review of the literature on collaborative innovation systems, including to the development of innovation because solution- and need-based in­
platforms. formation can be combined in meaningful ways for innovation (von
Hippel, 1994). Besides information advantages, digital platforms enable
2.1. Household sector innovation and diffusion success the formation of communities that typically evolve around common
interests on hobbies, products, tools and practices.
Innovation in the household sector has been defined as “the ideation Because platform members share their work for others to (re-) use
and development of new or modified products, services, processes, (Haefliger et al., 2008; Stanko, 2016), innovating within these collab­
techniques, or other applications by consumers, practiced in their orative systems is highly efficient (Baldwin and von Hippel, 2011).
discretionary time without pay” (Martin et al., 2020, p. 1; von Hippel, Members of digital platforms have developed sophisticated software
2017). von Hippel (2017) has further emphasized that the outcomes that products in peer-to-peer collaboration (Benkler and Nissenbaum, 2006)
consumers produce may go beyond the strict definition of innovations and have successfully solved scientific problems for organizations in
and may encompass other forms of less-innovative, creative outcomes competitive settings (Jeppesen and Lakhani, 2010). Moreover, the
such as user generated content (Aaltonen and Seiler, 2015; Benkler, feedback and information that members of digital (and physical) plat­
2006), or creative “fan fiction” (Jenkins and Deuze, 2008; Jenkins et al., form communities receive evidently reduces their uncertainty and thus,

2
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

Fig. 1. Effects of pre-innovation platform activity.

Table 1
Underlying mechanisms.
Hypotheses Innovation quality Recombinant innovation Innovation Documentation

H1: Pre-innovation Insights into market trends and technological Existing knowledge pieces and accumulated Understanding of usability as success factor and
platform activity possibility space lead to higher innovation choice sets facilitate recombinant innovation platform norms increase innovation
quality documentation
H2a: Frequency of pre- Learning leads to higher innovation quality Bigger pool of knowledge modules leads to Increased understanding and lower costs lead to
innovation activity recombinant innovation increased innovation documentation
H2b: Quality of pre- Learning from the best leads to higher Higher quality pool of knowledge modules High quality designs score high on ease of use
innovation activity innovation quality leads to more recombinant innovation thereby lowering costs of innovation
documentation
H2c: Relatedness of pre- Development of domain-specific expertise leads More related pool of knowledge modules Lower development costs lead to increased
innovation activity to higher innovation quality leads to more recombinant innovation innovation documentation

can induce the implementation of their innovative ideas (Autio et al., up-to-date and current with the state-of-the-art. Engaging in these ac­
2013; Halbinger, 2018). Overall, the strength of platform-based inno­ tivities helped for example platform users in the music industry to
vation processes resides in the access to a diversity of information from transform discussed trends and technologies into innovative ideas and
comments, questions, and solutions of other platform members to spe­ promoted engagement in commercial diffusion activity in the form of
cific, ready-to-use ideas and designs, typically free of charge. starting up (Autio et al., 2013). On the other hand, members of digital
Although these scholars all share the overall idea, that the key to the platforms are enabled to both create innovations and access other users’
creation of successful innovation is the participation in collaborative creations. Platforms have features and technologies allowing their
innovation systems, the underlying mechanisms remain unclear. We members to visualize their ideas and engage in trial-error processes
thus focus in the following section on understanding the role of platform through virtual prototyping and experimenting (D’Adderio, 2001;
activity prior to innovating and how this promotes the diffusion success Thomke, 1998). New communication and design tools allow for novel,
of the resulting innovation. additional forms of interaction on platforms. Because ideas and inno­
vative developments are typically available to all platform members free
3. Theory and hypotheses of charge (Harhoff et al., 2003), platform users can easily access, copy
and/or alter other members’ work for their own purposes. In this way,
In this section, we first explain typical activities of users on digital consumers gain important insights into the specifics of others’ de­
platforms. We then develop our first hypothesis in Section 3.1. based on velopments, the innovation design themselves, but also gain insight into
three mechanisms through which these platform activities relate to the the development process, i.e. how to develop an innovation.
diffusion success of consumers’ first innovation: its innovation quality, The pre-innovation platform activities will thus function as tem­
recombinant innovation and innovation documentation. In Section 3.2., plates and guidance for innovation type and process.
we develop a set of hypotheses on the characteristics of consumers’ pre-
innovation platform activities: frequency, quality and relatedness and 3.1. Pre-innovation platform activities and innovation diffusion success
elaborate how they relate to innovation diffusion success based on the
above-mentioned mechanisms. Fig. 1 provides an overview of our A priori, it is not clear how pre-innovation platform activities relate
theoretical model. It includes the effects of frequency, quality and to innovation diffusion success. After all, it may be sufficient to simply
relatedness of pre-innovation platform activity on innovation diffusion share an innovation on a platform without further activity on behalf of
success and, takes account of the mechanisms through which these re­ the focal contributor. We hypothesize that consumers’ activities on the
lationships are formed. Table 1 shows in an overview the main mecha­ platform prior to innovating will be positively related to innovation
nisms of all hypotheses developed. diffusion success based on three mechanisms, i.e. innovation quality,
Typical platform activities involve on the one hand, participation in recombinant innovation, and innovation documentation.
discussions on current topics and technologies, posting questions and Innovation quality. When consumers engage in platform activities
solutions, sharing ideas and commenting on each other’s work (Jeppe­ such as participating in discussions and exchanging ideas prior to
sen and Laursen, 2009). Simply put, by being active on platforms, users innovating, they first and foremost have a better market understanding.
support and learn from each other and at the same time, they become They learn about the problems and needs of platform users as well as the

3
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

newest technological requirements. Second, they will gain insights in on a platform, they develop an understanding of the norms and practices
the technological possibility space - that is, they will learn about the prevailing in platform communities. Online communities are governed
technological possibilities of tools used, e.g. 3D printer functionalities. by the practice of free revealing, that is platform members share their
These insights will impact the process of designing the innovation and its innovations with others in the community for free (Harhoff et al., 2003).
features. In sum, the activity on the platform prior to innovating will Thus, consumers who accessed and engaged with other platform mem­
lead to higher quality innovations given that the focal innovator has bers’ innovations are aware that they are expected to document all steps
learned about the platform users’ preferences on design features and of developing the innovation including its features and functionalities.
modules and what their peers consider functionally or aesthetically Put differently, through prior platform activity they know that investing
novel and useful, the core aspects of consumer innovation (von Hippel, this effort in diffusing their innovation and sharing their innovation in
1986, 1988, 2017). We expect that this in turn will lead to high appre­ return is expected and a wide-spread norm in network-based innovation
ciation and accordingly, high diffusion rates on the platform. systems (Faraj and Johnson, 2011; Ryall and Sorenson, 2007; Van der
Recombinant innovation. Furthermore, we expect that the knowledge Vegt et al., 2006). Consequently, we expect that active platform con­
obtained through pre-innovation platform activities is not only abstract, sumers will ensure to share their innovation objects in an understand­
but will lead to direct employment in the form of recombinant innova­ able and “easy to use”, compatible way. In turn, with increased effort on
tion. Recombinant innovation involves drawing on existing resources making the innovation understandable and easy to use, i.e. a more
and recombining them in new, useful ways (Henderson and Clark, 1990; thorough documentation of the innovation, more members are enabled
Schumpeter, 1942). Knowledge recombination in turn is an important to appreciate and further enhance the shared innovation. Ease of use
source for innovation success (Hargadon and Sutton, 1997). By engaging typically attracts users to further adopt designs (Fogliatto et al., 2012;
in various activities on the platform, consumers can build up a choice set Rogers, 2003).
of information and innovation modules that are proven to be valuable In sum, we expect that the platform activity prior to innovating will
and functional. This can foster a modular approach when designing their lead to higher innovation quality, recombinant innovation as well as
first own innovation because they can build on existing parts that they more elaborate innovation documentation and that this in turn, will
deem appropriate and useful for their innovation. increase the chances of higher diffusion success. This is of particular
The recombination will be conducive for innovation diffusion suc­ relevance when consumers become innovators for the first time as it
cess when attempting to innovate their first innovation. Innovators learn lowers the threshold to innovate. Designing an innovation is costly,
from one another and this will influence their subsequent actions demands effort and it is highly uncertain how other users will react to it
(Baldwin et al., 2006). Recombination allows synthesizing insights from (Baldwin et al., 2006). When innovation-related ideas and information
other platform users’ innovation designs, avoids redundancies and in­ have been exchanged on the platform before actually innovating, it is
creases focus on relevant user needs and technologies. Through very likely that this is also conducive for increasing its adoption after­
recombination, innovators can refine and integrate existing information wards (de Jong et al., 2018). Thus, we hypothesize:
pieces into their projects, allowing them to be more effective and effi­
H1. Pre-innovation platform activity increases innovation diffusion
cient in their innovation process and diffusion. Recombining existing
success.
knowledge to achieve innovation success is a common practice in the
development of innovation by users, e.g. in software development
(Haefliger et al., 2008), and in 3D communities where individuals work 3.2. Characteristics of consumers’ pre-innovation platform activities:
with digital artifacts and reuse “models” for innovation (Kyriakou et al., frequency, quality and relatedness
2017; Stanko, 2016). It is further a successful practice in firms (Leipo­
nen and Helfat, 2010). In sum, because using existing knowledge is less We argue in the following sections that the frequency, quality and
uncertain, effortful and costly, we expect that consumers who are active relatedness of the platform activities will matter for the diffusion success
on the platform prior to innovating will recombine existing modules into of consumers’ first innovation. For each of the following hypotheses, we
recombinant innovation, which in turn will foster the diffusion success. will again identify innovation quality, recombinant innovation and
Innovation documentation. Further, we expect that consumers who are innovation documentation as the three mechanisms through which the
active on the platform prior to innovating develop an understanding that characteristics of platform activities drive diffusion success.
those innovations become more successfully diffused that are well
documented. At the same time, an elaborate innovation documentation 3.2.1. Frequency of pre-innovation platform activities
may foster the observability of the innovation, reduce its complexity, We first expect that more frequent engagement in pre-innovation
facilitate compatibility and may also promote trialability – all of which platform activity facilitates higher innovation quality and this in turn,
are known antecedents of successful innovation diffusion (Rogers, increases chances of diffusion success. When individuals perform ac­
2003). Simply put, innovations that are properly documented are less tivities more frequently, i.e. in a series of events, they improve their
effort and easier to use than those that lack proper documentation and performance because each repetition offers an opportunity to learn
thus, generate a positive experience when interacting with them. This (Ellis, 1965). Learning occurs in iterative, dynamic ways (March 2010).
documentation includes describing the innovation, i.e. details on func­ Users can go through trial-and-error-processes and in this way enable
tionality and features, as well as elaborating on the steps of the devel­ learning by doing. By using toolkits for example, user preferences can
opment process. This enables others in turn, to more easily observe and become evident via iterations of the innovation design process until the
understand the innovation, replicate (test) and use it. It is well known optimal innovation outcome is reached (von Hippel and Katz, 2002).
that communication is key in affecting the successful adoption of new Similarly, inventors develop patents of higher value when they engage
products (e.g. de Jong et al., 2018), but when active on the platform in collaborations with others more frequently (Breschi and Lenzi, 2016;
prior to innovating, consumers know from their own experience that Samila and Sorenson, 2017). Thus, frequent activity on the platform
detailed instructions reduce complexity and that investing effort creates and/or exercising the design process makes individuals better in using
positive developer experience (Fagerholm and Münch, 2012). This ease tools and equipment, enabling them to produce outcomes of higher
of use and a positive developer experience is key to receive likes, com­ quality.
ments and to induce others to re-use an innovation - all of which are Second, frequency will also foster recombinant innovation which will
substantial for diffusion success. This is important particularly because increase innovation diffusion success. Being active on platforms more
the type of consumer active on digital platforms cares about those types frequently provides increased information advantages on technologies
of social signals (Gallus, 2016; Lerner and Tirole, 2002). and members’ preferences on functionality, design features and gaps in
Even if consumers are not motivated by social signals, when active the market (Allen, 1983; Autio et al., 2013; Orr, 2016). This is because

4
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

consumers can accumulate an even bigger pool of innovation-related 3.2.3. Relatedness of pre-innovation platform activities
information and choice sets to draw from when innovating them­ Related innovation “artifacts” provide a common basis from which
selves. Thus, it is very likely that consumers will re-use collected inno­ individuals can economically manage their efforts and attention in
vation knowledge and modules into recombinant innovations. Given the problem-solving activities (Carlile, 2002). Thus, we expect that when
efficiency and effectiveness involved with recombining existing knowl­ consumers engage on the platform with innovation objects in domains
edge, this leads to higher innovation diffusion success. related to their own, their chances of innovation diffusion success are
Finally, frequency will promote elaborate innovation documentation increased.
and thereby lead to higher innovation diffusion success. With frequency First, we expect this effect to be facilitated through higher innovation
comes an increased understanding of the successful patterns and norms quality. Consumers who focus on related domains when communicating
on the platform. It will become increasingly evident that those in­ with platform members and collecting information develop expert
novations are appreciated and reused that include an in-depth docu­ knowledge in that area. They develop a deeper understanding of the
mentation of the innovation’s functionality and design process steps domain-specific designs and technologies as well as learn the user needs
because they may promote key drivers of diffusion success such as and aesthetics in that domain and understand what is considered tech­
observability, ease-of-use and reduced complexity (Rogers, 2003). nologically valuable. This will enable them to generate innovations that
Frequent activity will lead to an increased understanding of those factors are more tailored to current demands – an aspect that will likely increase
and at the same time, lower design costs for the focal consumer inducing the chances of developing an innovation that is valuable to others. For
him/her to invest effort in documenting his/her own innovation (i.e. example, in special interest groups, information that is more related to
description length). Thus, taken together, we hypothesize: the focal individual’s domain is more valuable for developing an inno­
vation than information received outside that domain (Franke and
H2a. The frequency of pre-innovation platform activity increases innova­
Shah, 2003). The resulting higher innovation quality will ultimately
tion diffusion success.
lead to higher diffusion success.
Second, we expect that relatedness of pre-innovation platform ac­
3.2.2. Quality of pre-innovation platform activities
tivity will induce recombinant innovation and thus, will lead to innova­
Yet, not all innovation activities are equally valuable on platforms.
tion diffusion success. Innovation sparks a related flow of information
We expect that particularly the engagement with others’ high-quality
and knowledge creation, prompting a “sequence of innovations” in the
innovation objects will lead to higher innovation quality of the focal
wider innovation system (Nonaka, 1994). Put differently, innovations
consumer’s first innovation. This in turn will be linked to higher inno­
stem from related activities and this relatedness allows for economic and
vation diffusion success. On platforms, high quality innovations often
streamlined actions. On free open-source software platforms for
foreshadow market trends because they are typically developed by users
example, members who communicated with other platform members on
who are ahead of a trend (Autio et al., 2013; Jeppesen and Laursen,
a specific topic, pursued also projects in that area, allowing them to
2009). Their innovations address needs that are to date unmet, i.e. no
effectively work on sub-problems within that field (Foss et al., 2016).
product, service or process exists in the market and thus, these inno­
Although the recombination of unrelated and related knowledge pieces
vation have high market value (Lilien et al., 2002; von Hippel, 1986).
for innovation can be per see a challenging balancing act (Laursen and
Consumers who are engaging with high quality innovations prior to
Salter, 2006), we believe that the active engagement of the individual
innovating have accordingly substantial information advantages, simply
user with the platform community will provide indirect guidance into
because they learn from the best. The high levels of technical sophisti­
the “right” combination of knowledge pieces and prevent myopia
cation of these innovations will inform them about the (high) standards
because user needs are constantly updated. Further, by engaging in ac­
of manufacturing, i.e. what is considered aesthetically pleasing to others
tivities that are related to the focal innovation, more related knowledge
on the platform. This will induce the focal consumer to develop
modules and choice sets are available to the individuals and this in turn,
advanced innovations and will likely lead to higher innovation success.
will facilitate an effortless recombination of those modules. Thus,
Second, engaging with high quality innovations prior to innovating
although relatedness can, at times, produce myopia and potentially
will also induce consumers to engage in recombinant innovation. This is
decrease quality, we expect that the recombinant innovation achieved
because high quality innovation parts and modules are more likely to be
via prior platform engagement will be overall better fitted to the plat­
selected by the focal individual to be reused and recombined for a new
forms needs and thereby increases the chances for diffusion success.
innovation project – because they have received a lot of appreciation on
Finally, we expect that recombinant innovation will increase inno­
the platform and thus, attracted attention. For example, one study
vation documentation, which will prompt diffusion success. This is
showed that when innovation objects have a relative advantage
because individuals can realize increasing returns from innovating in
compared to other objects, they are more likely to be remixed (Stanko,
related domains given that the costs of transferring information from
2016). In turn, when a consumer uses only high-quality ingredients for
one domain to a related domain are decreasing with proximity. An
building his or her own innovation, we expect that this will also attract
important enabler of innovation diffusion is the compatibility with pre-
more attention and appreciation on the platform by others thereby
existing systems (Rogers, 2003). Accordingly, the costs of describing the
leading to higher innovation diffusion success.
recombinant innovation and documenting its development process
Third, we expect that high-quality platform activity will lead to
decrease depending on the degree of relatedness since prior, presumably
higher investments in innovation documentation and in this way, promote
compatible, information can be reused. It is thus, less effort to provide a
the diffusion success of the focal innovation. An innovation of high
detailed documentation of the innovation. A detailed innovation docu­
quality on a platform entails the important aspect that it creates a good
mentation, in turn, reduces complexity – a important factor of diffusion
developer experience when others try to replicate or reuse it. When
(Rogers, 2003) - and thereby, increases the ease of use for others. Given
consumers thus engage with such high-quality objects before innovating
the relatively lower costs associated with proximity, we expect that
themselves, they will follow their example. Further, given the elaborate
consumers will invest time and effort in elaborating on documenting the
documentation of the innovation, complexity will be decreased and the
related innovation and its development process and this in turn, will
ease of use of those high-quality objects will substantially reduce their
result in innovations that are valued and reused by others. Taking these
innovation costs. It will free up time to invest in documenting their own
points together, we hypothesize:
innovation, thereby leading to higher innovation success. Thus, we
hypothesize: H2c. The relatedness of pre-innovation platform activity increases inno­
vation diffusion success.
H2b. The quality of pre-innovation platform activity increases innovation
diffusion success.

5
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

Table 2
Variable definitions and summary statistics.
Variable Definition Dummy Logarithm Mean SD Min Max

Innovation Diffusion Success

Success Index Mean of the standardized values of six success variables − 0.06 0.81 − 1.64 6.74
Pre-Innovation Platform Activity
Pre-Innovation Activity At least one pre-innovation activity before release of first thing X 0.71 0.46 0 1
Frequency of Pre-Innovation Activity Number of pre-innovation activities X (+1) 2.16 1.98 0 9.61
Quality of Pre-Innovation Activity Average quality of pre-innovation activities 0.02 0.02 0 0.26
Relatedness of Pre-Innovation Activity Average relatedness of pre-innovation activities 0.39 0.28 0 1
Innovation Characteristics
Innovation Quality Quality of thing measured as proportion of likes to views 0.02 0.02 0 0.4
Recombinant Innovation Thing derived from at least one prior existing thing X 0.16 0.37 0 1
Description Length Length of the thing’s description in characters X (+1) 5.15 1.15 0 9.75
Controls
Intermediate Skills User reports an intermediate 3D design skill level X 0.34 0.47 0 1
Advanced Skills User reports an advanced 3D design skill level X 0.12 0.32 0 1
Biography Reported User reports information on their biography X 0.27 0.44 0 1
Days to Innovation Days from user registration until release of first thing X (+1) 3.78 2.01 0 8.15

N = 79,186.

4. Empirical setting particularly beneficial for our investigation because individuals join on a
voluntary basis. Since members decide how much time they want to
4.1. The 3D printing community Thingiverse invest on the various activities (e.g. commenting, collecting, copying or
modifying designs), the platform contains members of varying scale and
We test our hypotheses in the context of the Thingiverse platform, a scope of activity. This means they vary both in the extent and depth in
website for user generated content on open source hardware designs. which they interact with other members and in their range of
Members of the platform create and share digital design files related to innovations.
3D printing for free. Three-dimensional printing is a computer Further, this practice produces variety among platform users,
controlled process in which digital files are transformed into physical allowing us to observe those individuals who are “just” members, those
objects (Kyriakou et al., 2017). Because the platform provides free ac­ who directly innovate without performing other activities on the plat­
cess to designs, ideas and information on member needs and techno­ form, and those who first perform activities on the platform and then
logical trends as well as promotes support and assistance among become innovators. The core advantage of this setting is that it allows
members via discussion forums and posts, it is widely accepted by observation of activities in chronological order, that is, we observe in­
makers and the wider “do-it-yourself” community of household sector dividuals’ activities in a line of sequence, thereby allowing us to
innovators. Although the motivations of Thingiverse users are not investigate the effect of pre-innovation platform activity on innovation
explicitly observed, it is reasonable to assume that they are driven by diffusion success. Although many user innovation scholars have
self-rewards such as personal needs, enjoyment and motivation to learn, demonstrated the importance of communities and platforms in user
and not by commercial intent. This assumption seems to be in line with innovation, most studies do not allow for this chronological investiga­
the insights derived from a sample of interviews conducted with Thin­ tion (Franke and Shah, 2003). Further, they typically focus in their final
giverse users from the United States and Europe prior to the data sample on innovators only (de Jong et al., 2015), thereby not providing
collection in order to get insights into the setting. We introduce sup­ a clear distinction between when and how the activities conducted on
porting quotes for this assumption in the Discussion section of the study. the platform occurred in the innovation process.
Furthermore, prior research on the Thingiverse platform shows that
individuals engage in the community to address personal needs and not
4.2. Data and variable description
for a financial purpose (Shaw, 2018) or, to help others (Buehler et al.,
2015).
We obtained all 1,086,989 3D models that have been uploaded to
Besides its wide usage, the context of the Thingiverse platform is
Thingiverse until June 14th, 2018 and profiles of 702,924 individuals

Table 3
Pairwise correlations.
1 2 3 4 5 6 7 8 9 10 11

Innovation Diffusion Success

Success Index 1 1.00
Pre-Innovation Platform Activity
Pre-Innovation Activity 2 0.07 1.00
Frequency of Pre-Innovation Activity 3 0.07 0.70 1.00
Quality of Pre-Innovation Activity 4 0.02 0.83 0.65 1.00
Relatedness of Pre-Innovation Activity 5 0.06 0.89 0.59 0.71 1.00
Innovation Characteristics
Innovation Quality 6 0.13 0.02 0.03 0.10 0.00 1.00
Recombinant Innovation 7 0.10 0.13 0.17 0.10 0.14 − 0.01 1.00
Description Length 8 0.27 0.09 0.10 0.07 0.10 − 0.02 0.12 1.00
Controls
Intermediate Skills 9 0.05 0.02 0.01 0.02 0.01 0.02 − 0.03 0.05 1.00
Advanced Skills 10 0.13 − 0.04 − 0.06 − 0.06 − 0.04 0.00 − 0.04 0.03 − 0.26 1.00
Biography Reported 11 0.12 0.00 − 0.01 − 0.01 0.00 − 0.01 − 0.03 0.10 0.09 0.15 1.00
Days to Innovation 12 − 0.06 0.25 0.28 0.18 0.20 0.05 0.09 0.04 0.01 − 0.03 − 0.06

N = 79,186.

6
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

that have registered on the platform and for which we observed activity. First, Thingiverse members can create individual collections to
The 3D models uploaded on Thingiverse are called things and we will be which they can add things that are of interest to them. The dummy
using this notation in the following. variable Collect takes the value one if a member added things to col­
As this paper focuses on the role of pre-innovation platform activ­ lections after signing up on Thingiverse and before publishing their first
ities, we track member behavior from the time individuals register their own thing. We expect that exploring other designs can be a key learning
account until the time of the first consumer innovation. We define in­ mechanism to help consumers to create their own things. This learning is
novations as the uploading of unique new things on the platform and do facilitated by the fact that diffusion success measures are prominently
not consider customizations (Kyriakou et al., 2017) as innovations as displayed as part of each thing’s description. This way, members can
these are just minor adaptions of existing designs, such as changes in the infer what kind of design characteristics could increase the diffusion
length of an object. Overall, we observe 146,290 members that have success of their own innovations.
created at least one thing on Thingiverse and of those, 79,186 report Second, we construct the dummy variable Copy that measures when
their 3D design skill level. For these individuals, we observe both their members created a print of an existing thing and shared their results on
pre-innovation platform activity, i.e. all observable activities on Thin­ the platform. Printing objects on 3D printers is far more challenging than
giverse from the time of registration until the upload of the first thing, as printing on a regular printer and many prints initially fail. Today’s 3D
well as diffusion success measures and characteristics of their first printers are far from standardized and need to be calibrated precisely to
innovation. One observation in our dataset therefore captures the achieve satisfactory results. The choice of the printing material greatly
pre-innovation period as well as the first innovation for all consumers influences results as well. By experiencing these challenges themselves,
that eventually innovate. Variable definitions and summary statistics are consumers can gain valuable knowledge for creating their own models.
reported in Table 2 and Table 3 shows pairwise correlations. Important insights could be to experience which kind of geometries
We construct three kinds of variables: a measure of innovation work well and what are minimum size thresholds.
diffusion success, variables measuring pre-innovation activity and A third type of platform activity is to engage in discussions with other
observable characteristics of users and the innovations. platform members. The dummy variable Comment takes the value one if
a member has been writing a comment in the discussion forums that are
4.2.1. Measure of diffusion success attached to each thing and to each copy of a thing. As creating new 3D
Thingiverse reports a wide range of diffusion success measures for models can be a highly complex task, engaging in discussions with other
each thing. Views is the cumulative number of individuals that visited a platform members can be a promising channel to acquire new knowl­
thing’s website on the Thingiverse platform and Downloads measures edge that can then be directly applied in own innovation efforts.
how often the files of the 3D model have been downloaded. In contrast We also create the variable Pre-Innovation Activity as a joint measure
to many other platforms, Thingiverse is relatively open in that viewing of pre-innovation platform activity that takes the value of one if the
and downloading of things does not require registration, but all consumer engaged with at least one of the three types of pre-innovation
following innovation diffusion success measures require registration. platform activities. We use this joint measure for our preferred specifi­
Similar to social media platforms, individuals can like a thing and Likes cation, but report robustness tests for the individual measures in Section
measures the total number of likes a thing received. If members want to 5.1.
bookmark things, they can add them to individual collections and Col­ While we have so far operationalized pre-innovation platform ac­
lects is the cumulative number of thing added to member collections. If tivity as a dummy variable, we now disentangle platform activity into
members decide to make a copy of a thing by printing it on a 3D printer, the three components frequency, quality, and relatedness.
they can share the outcome on the thing’s site and Copies counts the First, we count the number of times a consumer has conducted pre-
number of times a report of a copy has been created. Finally, members innovation platform activities and create the variable Frequency of Pre-
can also create recombinant innovations by using one or more existing Innovation Activity. As the frequency of pre-innovation platform activ­
things as a basis for their own innovations and Derivatives counts the ity is highly dispersed, we measure it as the logarithm of the number of
number of times a thing has been used by others in follow-up in­ pre-innovation activities (we add one to avoid losing observations
novations. If we think about the innovation diffusion process as a funnel, without pre-innovation community experience).
Views and Downloads create awareness about an innovation, Likes and Second, we want to measure the average quality of the 3D objects
Collects proxy for evaluating the innovation, and Copies and Derivatives with which the pre-innovation platform activities have been made. This
measure an innovation’s implementation. is possible as the kinds of platform activities that we introduced above
All diffusion success measures are count measures and as diffusion are always linked to a specific 3D object on the platform. We measure
success is highly dispersed, we take the logarithm. To ensure that things quality by dividing the number of a thing’s likes by the number of views
for which diffusion success measures take the value 0 are not dropped and then derive the variable Quality of Pre-Innovation Activity as the
because of the non-defined logarithm of 0, we always add 1 before average quality measure from all activities prior to the first innovation.
taking the logarithm. We only observe the value of diffusion success Third, we create the variable Relatedness of Pre-Innovation Activity to
measures at the time of the data collection, which favors older things as measure the relatedness of pre-innovation platform activities with the
they had more time to attract attention and generate diffusion success. first innovation. We create a relatedness measure by comparing the
In all regressions for which the dependent variable is diffusion success, category of the focal innovation with the categories of the things for
we therefore control for the linear and squared term of the days since a which the pre-innovation platform activities have been conducted. The
thing has been posted as well as the release year, the release month and most basic approach would be to check if any of the prior activity falls in
the weekday of the release. the same category as the focal category. However, this is problematic in
We generate a joint measure of innovation diffusion success. To do case an individual conducted activities in a lot of categories (in our case
this, we first standardize each of the six success measures discussed 80), as only activities in the focal category would be counted and all
above by subtracting their mean and dividing by their standard devia­ other activities would be treated as unrelated. We thus use the following
tion and then take the mean of these standardized values to generate the approach to create a continuous relatedness measure. We start by
variable Success Index. While we mainly use this joint success measure, generating a vector for each member, where one column represents the
we also run robustness checks with the six different success measures. activity in each of the N categories. We then calculate pairwise corre­
lations between all categories, which results in an NxN matrix. Higher
4.2.2. Measures of pre-innovation platform activity values in this matrix indicate that more members are co-engaging in the
We observe three ways in which individuals engage on the Thingi­ two categories. We then calculate relatedness as the weighted sum of all
verse platform before innovating. prior activities a consumer has gained on the platform.

7
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

4.2.3. Measures for innovation characteristics Table 4

To tease out the mechanisms through which pre-innovation platform The role of elements of pre-innovation activity and the mediating role of inno­
activity works, we want to identify innovation characteristics that would vation choices.
translate into innovation diffusion success. We would find support for (1) (2) (3) (4)
these mechanisms if these characteristics work as mediators, i.e. if pre- Success Success Success Success
innovation platform activity is associated with changes in the use of Index Index Index Index

certain innovation characteristics, if these innovation characteristics in Pre-Innovation 0.0871*** − 0.156*** 0.0296*** − 0.0998***
turn affect innovation diffusion success, and if the association between a Activity
(15.72) (− 9.31) (5.73) (− 6.43)
consumer’s pre-innovation platform activity and innovation diffusion
Frequency of 0.0123*** − 0.000768
success becomes weaker once we control for these innovation charac­ Pre-Innovation (6.73) (− 0.45)
teristics. We identify three possible mechanisms: quality of the innova­ Activity
tion, use of recombinant innovations, and investments into innovation Quality of 2.566*** 1.255***
documentation. Pre-Innovation (8.25) (4.23)
Activity
As described above, we measure Innovation Quality as the proportion
Relatedness of 0.244*** 0.174***
of likes to views for as we expect that things with a higher inherent Pre-Innovation (11.51) (8.86)
quality have a higher likelihood of receiving a like when they are viewed Activity
by a platform user. Innovation 9.884*** 9.845***
Quality
Recombinant innovations are innovations that are built on the base
(50.11) (49.90)
of existing innovations. This kind of innovation might be a powerful way Recombinant 0.157*** 0.156***
to re-use elements from successful innovations instead of starting the Innovation
search process for an innovative design from scratch. We create the (23.09) (22.83)
dummy variable Recombinant Innovation that equals one if the creator of Description 0.170*** 0.170***
Length
the thing lists at least one existing thing as an ancestor.
(74.86) (74.67)
Innovation documentation is operationalized with the variable Intermediate 0.0758*** 0.0781*** 0.0545*** 0.0552***
Description Length, which measures the logarithm of the number of Skills
characters for a thing’s description. This measure is an appropriate (14.08) (14.51) (10.93) (11.06)
Advanced 0.193*** 0.197*** 0.166*** 0.167***
proxy for innovation documentation because generating a longer
Skills
description makes it easier for a potential adopter to evaluate if the (21.27) (21.69) (19.92) (19.96)
innovation is of relevance for them and gives a more exhaustive guide on Biography 0.0449*** 0.0441*** − 0.00772 − 0.00810
how to produce and use the innovation. Reported
(7.51) (7.38) (− 1.40) (− 1.47)
Days to − 0.00514*** − 0.00476*** − 0.00843*** − 0.00730***
4.2.4. Control variables
Innovation
It is important to control for the capabilities of the users of Thingi­ (− 3.95) (− 3.58) (− 7.04) (− 5.98)
verse because members with different capability levels might have N 79,186 79,186 79,186 79,186
different costs of conducting activities on the platform and might also R2 0.277 0.280 0.386 0.387
benefit in different ways. Not controlling for capabilities of users might Notes: OLS point estimates with t-values based on robust standard errors in pa­
therefore lead to results for platform activities being biased downwards rentheses. All specifications control for the linear and squared term of the days
(if low-quality innovators have to conduct more pre-innovation platform since a thing has been posted, for category (80), release year (11), release month
activities before coming up with their low-quality innovations) or biased (12), and release day (7) fixed effects, and a constant but these terms are not
upward (if high-quality innovators have lower costs of conducting pre- reported. Asterisks denote significance levels (*** p < 0.001, ** p < 0.01, * p <
innovation platform activity before coming up with their high-quality 0.05).
innovations). To capture the capabilities of users on Thingiverse, we
use the self-assessment of their 3D design skills in the user profile. We 5. Results
construct the variables Intermediate Skills and Advanced Skills that equal
one if the user reports intermediate or advanced 3D design skills (with We conduct our analysis in two steps. First, we test our hypotheses
lower skill levels being the omitted base category). Additionally, we and find strong support for the effects of pre-innovation platform ac­
create the variable Biography Reported that equals one if the user reports tivity as well as the activity’s frequency, quality, and relatedness on
information on their biography. innovation diffusion success. Second, we examine the mechanisms
We furthermore create the control variable Days to Innovation to driving the effects of pre-innovation platform activity on innovation
capture the logarithm of the number of days between a member’s diffusion success and find that the results are to a large part driven by
registration on Thingiverse and the day when the first innovation was innovation quality, recombinant innovations and innovation
uploaded. This variable should proxy for forms of member activity that documentation.
are unobserved by us but vary with the time spent on the platform. This
could for example be learning from viewing or downloading things or 5.1. The effect of pre-innovation community experience on innovation
direct messages exchanged with creators of things. diffusion success
In addition to the control variables that are explicitly reported, we
also control for fixed effects on the level of the 80 different categories We test H1 on the expected positive effect of pre-innovation platform
available on Thingiverse and on the level of the thing’s release year, activity on innovation diffusion success with model (1) in Table 4.
release month, and the weekday of the release. These fixed effects are We find support for H1 as the coefficient for pre-innovation platform
helpful to account for between-category and between-year differences in activity of 0.0871 is positive and highly significant.
overall innovation diffusion success. Finally, as we already mentioned in When we look at the control variables, higher skill levels are asso­
Section 4.2.1, in all innovation diffusion success regressions, we control ciated with more innovation diffusion success and those users that
for the linear and square term of days since the thing was first posted to report biographical details are also more successful. We find a negative
control for different exposure lengths that things have had. and significant effect for days between the registration of the member
and their first innovation.
For H2a-c, we expect that the frequency, quality, and relatedness of

8
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

Table 5
The relationship between pre-innovation activity and thing characteristics.
(1) (2) (3) (4) (5) (6)
Innovation Quality Innovation Quality Recombinant Innovation Recombinant Innovation Description Length Description Length

Pre-Innovation Activity 0.00117*** − 0.00449*** 0.0973*** − 0.0207* 0.202*** − 0.0245

(7.85) (− 9.69) (38.84) (− 2.39) (22.44) (− 0.91)
Frequency of 0.000160*** 0.0297*** 0.0441***
Pre-Innovation Activity (3.31) (29.18) (15.52)
Quality of 0.174*** − 1.280*** − 1.656***
Pre-Innovation Activity (16.28) (− 7.67) (− 3.45)
Relatedness of 0.000668 0.117*** 0.253***
Pre-Innovation Activity (1.38) (10.15) (7.51)
Intermediate Skills 0.00136*** 0.00138*** − 0.0364*** − 0.0330*** 0.0803*** 0.0857***
(9.62) (9.81) (− 12.98) (− 11.84) (9.29) (9.92)
Advanced Skills 0.00238*** 0.00244*** − 0.0588*** − 0.0517*** 0.0742*** 0.0850***
(12.12) (12.42) (− 15.34) (− 13.56) (5.56) (6.37)
Biography Reported 0.000430** 0.000414** − 0.0120*** − 0.0122*** 0.292*** 0.292***
(2.94) (2.84) (− 4.16) (− 4.24) (32.13) (32.13)
N 79,186 79,186 79,186 79,186 79,186 79,186
R2 0.112 0.118 0.052 0.065 0.088 0.091

Notes: OLS point estimates with t-values based on robust standard errors in parentheses. All specifications control for category (80), release year (11), release month
(12), and release day (7) fixed effects, but coefficients of these fixed effects and of the constant are not reported. Asterisks denote significance levels *** p < 0.001, ** p
< 0.01, * p < 0.05).

pre-innovation platform activity are positively related with innovation collects) are larger than for implementation (copies, derivatives). We
diffusion success. We test these hypotheses in model (2) of Table 4 by finally collapse the different diffusion success measures into one diffu­
adding frequency, quality, and relatedness of the pre-innovation plat­ sion success index in column (7) and observe that the relative effect sizes
form activity to the baseline model from column (1). We find strong are highest for collect activity, lowest for comment activity, and inter­
support for H2a-c as the coefficients for the frequency, quality, and mediate for copy activity.
relatedness of pre-innovation platform activity are all positive and sig­
nificant. While the diffusion success index as the dependent variable 5.2. Mechanisms behind the effects of pre-innovation activity
does not allow a direct interpretation in terms of economic significance,
we can compare the relative magnitude of the effects by multiplying the When deriving our hypotheses, we argued that innovation quality,
coefficients with the standard deviations of the respective variables. the use of recombinant innovations, and innovation documentation are
Doing this, we get a value of 0.068 for relatedness, 0.051 for quality, and the mechanisms through which pre-innovation platform activity in­
0.024 for frequency. Therefore, relatedness and quality of pre- creases innovation diffusion success. These mechanisms should there­
innovation platform activity seem to be the strongest predictors for fore mediate the relationship between pre-innovation platform activity
innovation diffusion success. and innovation diffusion success. Three necessary conditions have to be
To test H1 in more detail, we also split up the six different measures fulfilled: 1) we need a positive relationship effect of pre-innovation
of innovation diffusion success and the three different types of pre- platform activity on the use of the three innovation characteristics, 2)
innovation platform activity instead of using the aggregated measures the innovation characteristics have to positively drive innovation
as before. Summary statistics and pairwise correlations for these more diffusion success, and 3) the relationship between pre-innovation plat­
granular measures are reported in Tables A1 and A2 in the appendix. We form activity and innovation diffusion success has to become weaker
see that the different measures for innovation diffusion success are once we control for the innovation characteristics.
generally highly correlated, while the different measures for pre- We start by examining the first condition, i.e. if pre-innovation
innovation platform activity are much less correlated. Results for the platform activity is positively related with innovation quality, the cre­
robustness checks are then reported in Table A3 in the appendix. In ation of recombinant innovations, and longer description lengths as a
columns (1) to (6), we use the six different measures of innovation proxy for innovation documentation. We test these relationships in
diffusion success as dependent variables and resort to the joint measure Table 5. In models (1), (3), and (5), we look at the base effect of pre-
of innovation diffusion success in columns (7). We measure pre- innovation platform activity, while we add frequency, quality, and
innovation platform activity with the three dummy variables for activ­ relatedness of pre-innovation activity in models (2), (4) and (6).
ity regarding collection of things, creating copies of things, and com­ Our results show that pre-innovation platform activity has indeed a
menting on things. positive and significant relationship with the innovation characteristics.
Our results provide additional support for H1. Nearly all measures of Having pre-innovation platform activity is associated with an increase in
pre-innovation platform activity are positive and significant, with the innovation quality amounting to 7.7% of the standard deviation of
only exemptions being the insignificant coefficient for comment activity innovation quality, to a 9.7% higher probability that the innovation is
in model (5) and copy activity in model (6). As we are no longer using an building on existing innovations, and to a 22.4% longer description.
index variable in columns (1) to (6), we can now also assess the eco­ Frequency, quality, and relatedness of pre-innovation platform activity
nomic significance of our results and we can see that they are also show a generally positive and significant relationship with the
economically meaningful. As the dependent variable in models (1) to (6) innovation characteristics, except one insignificant or negative coeffi­
is the logarithm of the diffusion success measures, and the measures of cient in each of the three models.
pre-innovation platform activity are dummy variables, we can approx­ Next, we have to check if the innovation characteristics are positively
imately interpret the coefficients as percentage changes in the depen­ related with innovation diffusion success. To do this, we go back to
dent variable if the activity dummy turns one. The precise interpretation Table 4, where model (3) and (4) include innovation quality, the use of a
is that the percentage change equals 100 × (exp(coefficient) − 1), so the recombinant innovation, and the description length. All coefficients are
effect of collect activity on views in column (1) would for example be positive and significant, i.e. these innovation characteristics are posi­
10.8%. When we compare the different stages of the diffusion funnel, tively associated with innovation diffusion success. The second neces­
effect sizes for awareness (views, downloads) and evaluation (likes, sary condition for our mediation hypothesis is therefore also confirmed.

9
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

Finally, we check if the relationship between pre-innovation plat­ diffusion on digital platforms, that has not received much attention yet
form activity and innovation diffusion success is weakened once we in prior work but that will become increasingly important in the future.
control for innovation characteristics. Comparing the coefficients for Digital technologies are transforming individual behavior and eco­
pre-innovation platform activity in models (1) and (3) in Table 4, we see systems (Nambisan et al., 2019), and the household sector is no excep­
that the effect size in model (3) becomes 66.0% smaller. Similarly, if we tion to that. Digital platforms will play an increasingly important role for
compare the coefficients for frequency, quality, and relatedness of pre- household sector innovators because digital technologies enable in­
innovation platform activity between model (2) and (4), we see that dividuals to collaborate through online infrastructures and work on
the effect sizes went down by 28.7% (relatedness), 51.1% (quality), and projects related to software and beyond. Powerful design tools allow
106.2% (frequency). The third condition for our mediation hypothesis is locally dispersed individuals to virtually prototype by using advanced
therefore also fulfilled. software technologies (D’Adderio, 2001) and facilitate
As all necessary conditions for the mediating role of description innovation-related information transfers (Vaccaro et al., 2009). In line
length and recombinant innovations are fulfilled, we find strong support with this research, our findings show that consumer innovators active on
for a mediating role of the three mechanisms. This type of mediation is a digital platforms can learn from each other and in this way, are enabled
partial mediation and not a full mediation, as we still find a positive (but to develop innovations that diffuse more successfully than those who do
reduced) and significant relationship between pre-innovation platform not engage with the platform community prior to innovating. Specif­
activity and innovation diffusion success. The only exception is fre­ ically, we find that consumers who engage in platform activities more
quency of pre-innovation activity, for which we observe full mediation. frequently or work with other members’ innovations of high quality or
related to their own subsequent innovation, chances of diffusion success
6. Discussion and conclusion are increased.
As such, our study also highlights the importance of innovation
In this study, we analyze the effects of consumers’ platform activities diffusion success for this type of consumer innovators. To date, research
prior to their first innovation on innovation success, i.e. the extent to has shown that consumers’ innovation efforts are driven by the expec­
which their first innovation diffuses on the platform. Our goal is to tation of using the innovation (and thereby address a problem or need of
investigate under which conditions digital platforms can support the theirs), or, because they enjoy the process of innovating and learning or
development and diffusion of innovation by an increasingly important helping others (for an overview, see von Hippel, 2017). Our study ex­
subgroup of household sector innovators, i.e. consumers that are inter­ tends this conception and highlights that there is an important group of
ested in learning and diffusion. We empirically test our hypotheses and consumers that is self-rewarded by learning, but that is distinct from
underlying mechanisms by using the context of the largest 3D printing regular consumer innovators in the sense that they are motivated to
platform, Thingiverse. share and diffuse their innovations with others because the very utili­
In the following, we first summarize results. Then, we discuss the zation and appreciation of the innovation by others in itself is socially
contributions of our study in light of current research on household gratifying. As one member of the Thingiverse platform states: “A lot of
sector innovation. Third, we consider implications for platform opera­ the enjoyment I get out of creating things is knowing that other people
tors and finally discuss implications for industry and society. find them useful.” Our sample interviews with regular members of the
Thingiverse platform located in the United Sates and Europe, provide
6.1. Summary of results further anecdotal evidence that sharing innovations with others is an
important reason for consumers to use digital platforms for innovation
To provide empirical evidence that being active on digital platforms purposes in the first place. Successful diffusion, i.e. receiving high
can promote the development and diffusion success of a consumer’s first numbers of likes and views on innovation objects that they developed
innovation, we first empirically investigated and find a positive effect of acts as an important self-reward for these consumers because it repre­
consumers’ platform activity prior to innovating on the success of sub­ sents a form of recognition and status or reputation. Another inter­
sequent innovation diffusion. Second, we show that the frequency, the viewee for example mentioned that platforms such as Thingiverse
quality and relatedness of these pre-innovation platform activities are provide extrinsic incentives because other users can sign up as followers
key factors strengthening this said relationship. In doing so, we show and/or like other members’ designs. This interviewee stated that he
that the impact of consumers’ pre-innovation platform activities on finds himself incentivized because the “analysis of downloads, likes etc.
innovation diffusion success depends on the extent to which individuals of my own innovation models can be shown [in the user profile].”
engage in those activities as well as the type of activity. We furthermore Achieving status is important to individuals (Lindenberg, 2013) and
identify and test three mechanisms through which pre-innovation plat­ higher ranks matter to them (Tran and Zeckhauser, 2012). Thus, high
form activities relate to innovation diffusion success, namely through numbers of views, likes or copies are perceived as acknowledgement of
the quality of the innovation, through recombining other platform the focal innovator’s performance on the platform. Recognition is an
members’ work into recombinant innovation and through innovation important driver on platforms and particularly motivating when ach­
documentation. Third, we show that diffusion success on the platform ieved by peers (Lerner and Tirole, 2002). In sum, although our empirical
can be considered as a staged approach. Other platform members may analyses do not allow demonstrating the motivational effect of innova­
initially become aware of the focal consumer’s innovation (awareness tion diffusion success, anecdotal evidence from our interviews suggests
measured via views and downloads), they then evaluate the innovation that recognition and status are important motivators for consumers to
(evaluation measured via likes and collects) and finally implement the share and diffuse their innovations on the platform. Considering that the
innovation themselves (implementation in the form of copies and/or current and upcoming consumer generation is becoming increasingly
derivatives). technologically savvy and has an inherent preference to use social media
and platforms such as YouTube, Instagram and Wikipedia in daily life,
6.2. Contributions to research on household sector innovation we believe this is an important insight on the motivations of innovators
in the household sector more generally.
While empirical evidence exists that collective physical (maker) Further, although this subgroup is interested in diffusing, they still
spaces are important in the development and diffusion of consumer share the same problem with general household sector innovators, i.e.
innovation (Franke and Shah, 2003; Halbinger, 2018), our study con­ that the diffusion of innovation is difficult. Finding that engaging on
tributes to this research by highlighting the role of digital infrastructures digital infrastructures can help mitigate this issue is thus, of general
for innovation in the household sector. We investigate a particular group interest to innovation researchers and practitioners. Consumer innova­
of household sector innovators, i.e. consumers interested in learning and tion, if diffused, has the potential to increase social and economic

10
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

welfare (Gambardella et al., 2016), and many researchers have postu­ industry sectors and society as a whole. Forecasts for 2020 predict $15.8
lated infrastructures to support consumer development and its diffusion billion for the global market of additive manufacturing products and
(de Jong et al., 2015; von Hippel, 2017). services and $35.6 billion in 2024 (Wohlers, 2019). The 3D printing
Finally, by investigating the behavior of consumers on a platform technology experiences worldwide substantial investments and new
prior to their first innovation, we also increase understanding on the market entries; it is widely used across industries such as aerospace,
transition from platform participant to innovator – which is a highly automotive and logistics industry for skill development and learning,
risky and challenging transition. Our study thus, provides new insights material development, cybersecurity and automation (McCue, 2019).
on the genesis of household sector innovation by showing the differ­ Thus, understanding the factors that shape the likelihoods of success of
ences between consumers who successfully develop and diffuse inno­ 3D innovations and how the behavior of individual innovators on digital
vation and those who do not. platforms can shape their diffusion rates is certainly of interest to players
in the 3D printing industry.
6.3. Implications for platform operators Our paper has also implications for society at large. Digital platforms
such as Thingiverse are gaining increasing importance in assisting
Our findings also speak to practitioners and in particular platform consumers to collect innovation-related knowledge to solve their inno­
operators. First, our study has implications as to how to achieve an vation problems, tinker, innovate and diffuse their new developments to
important goal of platform operators, namely increasing the installed the outside world. Platforms are increasingly supporting the commercial
base, i.e. the number of users of platforms. Our study highlights the diffusion of innovation – in particular because they facilitate crowd­
factors that lead to successful innovations on platforms. This is impor­ funding campaigns based on both fiat money (e.g. Kickstarter) and
tant because when a platform produces successful innovations, it will cryptocurrencies (e.g. Initial Coin Offerings), thereby lowering the
retain and attract more users. This is key for platform operators as it may threshold for innovation and start-up activity. Not surprisingly, the 3D
provide a competitive advantage – and often, it is the very reason of their printing technology is perceived as core to the general maker movement
existence. Platform communities can generate complementary assets to that some acknowledge as the next industrial revolution because of the
organizations’ market offerings and this has been shown to be key in very nature of the technology empowers all citizens to innovate
increasing the number of product users (Dahlander and Magnusson, (Anderson, 2012; Hatch, 2013; Lindtner et al., 2014) and supports
2005; Dahlander and Wallin, 2006). Further, when platforms are able to entrepreneurial activity (Browder et al., 2019).
produce innovations of value to many others, platform traffic will most Yet, common-based peer production enabled through digital plat­
likely increase. This in turn is interesting from the perspective of forms goes beyond just facilitating the development and diffusion of
advertising-based markets as this offers new business models for orga­ innovation. The supportive and collaborative practices and norms pre­
nizations to generate profits. vailing on digital platforms also provide the basis for “virtue” in society
Second, our study indicates how producers and users can better and, a “context for positive character formation” (Benkler and Nissen­
complement each other. Researchers have argued that overall economic baum, 2006; p.394–395). The technology of 3D printing also enables
and social benefits could be increased if users and producers would digital fabrication which is said to address sustainability concerns as it
collaborate by complementing one another in their innovation en­ allows for bespoke designs and customized production (Hielscher and
deavors (Gambardella et al., 2016). To date, researchers have high­ Smith, 2014). Finally, because digital platforms enable collaboration
lighted that open source software communities are particularly and access to information for everyone, independent of geographic
appropriate to offer the best for ‘both worlds’ because code can be easily location, societal status and background, they are able to address issues
shared, duplicated, amended and further developed online (Hippel and of inclusiveness (Smith, 2017).
Krogh, 2003). But because communication and design tools are
becoming increasing digitized and powerful, we believe that the range Funding
of collaboration opportunities with individuals are expanded via digital
platforms and allow for new forms of collaboration between consumers This research was funded by the Eugene M. Lang Foundation, New
and organizations. One reason is that the collaborative nature of inno­ York, the Lawrence N. Field Center for Entrepreneurship at Baruch
vation processes in platform communities present a novel and successful College, City University of New York and the Research Foundation of the
alternative to concentrated innovation models in organizations since City University of New York.
they are highly informative as to how organizations should develop
innovations and organize their processes (Hippel and Krogh, 2003). CRediT authorship contribution statement
Another reason is that digital platforms provide access to the in­
novations that their members develop and this can fuel the platform Jörg Claussen: Investigation, Formal analysis. Maria A. Halbinger:
provider’s innovation pipeline. This represents a major benefit in mar­ Conceptualization, Writing - original draft.
kets that have become increasingly competitive and dynamic.
Finally, our study provides relevant insights for platform providers as
to how they can shape behavior on platforms. We highlight innovation Declaration of Competing Interest
diffusion success as an important motivator for individuals to participate
and remain on platforms. Particularly for first time contributors, being The authors declare that they have no known competing financial
awarded via peer recognition can increase self-confidence and interests or personal relationships that could have appeared to influence
strengthen the belief that they can make important contributions to the the work reported in this paper.
platform community (Karau and Williams, 1993; Rashid et al., 2006).
This is an important insight because knowing platform members’ Acknowledgements
motivational factors enables the operators to shape members’ behavior
on platforms by the rewards and platform features that they provide (e. We are grateful to the special issue editors Nikolaus Franke, Jeroen
g. Dasgupta and Hill, 2018). de Jong, and Eric von Hippel as well as Nicolai Foss, Lars Frederiksen,
Romi Kher, Scott Newbert, Jon Pellicoro, three anonymous reviewers,
6.4. Implications for industry and society audiences at the Open and User Innovation Conference 2019, at the
DRUID 2019 conference, at the AOM 2019 conference, and at the ORG
The findings of our study are also important considering the signif­ seminar at LMU Munich for valuable comments and discussions and to
icant growth and emerging importance of 3D printing in various Laura Steinke for research assistance.

11
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

Appendix

Table A1–A3

Table A1
Variable definitions and summary statistics of additional variables .
Variable Definition Dummy Logarithm Mean SD Min Max

Innovation Diffusion Success

Views Number of views of thing X (+1) 5.91 1.56 0.69 13.62
Downloads Number of downloads of thing X (+1) 4.67 1.38 0 12.52
Likes Number of likes of thing X (+1) 2.07 1.42 0 10.05
Collects Number of collects of thing X (+1) 2.30 1.51 0 10.13
Copies Number of copies of thing X (+1) 0.21 0.53 0 7.23
Derivatives Number of derivatives of thing X (+1) 0.10 0.35 0 3.43
Success Index Mean of the standardized values of the six variables above − 0.06 0.81 − 1.64 6.74
Pre-Innovation Platform Activity
Collect At least one thing added to collection before release of first thing X 0.63 0.48 0 1
Copy At least one thing printed before release of first thing X 0.22 0.42 0 1
Comment At least one comment made before release of first thing X 0.23 0.42 0 1
Pre-Innovation Activity At least one pre-innovation activity before release of first thing X 0.71 0.46 0 1

N = 79,186.

Table A2
Pairwise correlations for additional variables.
1 2 3 4 5 6 7 8 9 10

Innovation Diffusion Success

Views 1 1.00
Downloads 2 0.91 1.00
Likes 3 0.86 0.83 1.00
Collects 4 0.87 0.83 0.94 1.00
Copies 5 0.55 0.57 0.58 0.58 1.00
Derivatives 6 0.40 0.41 0.40 0.41 0.47 1.00
Success Index 7 0.92 0.91 0.92 0.92 0.75 0.61 1.00
Pre-Innovation Platform Activity
Collect 8 0.05 0.05 0.06 0.08 0.03 0.03 0.06 1.00
Copy 9 0.05 0.02 0.05 0.06 0.03 0.02 0.05 0.18 1.00
Comment 10 0.03 0.01 0.03 0.04 0.02 0.02 0.03 0.17 0.37 1.00
Pre-Innovation Activity 11 0.07 0.06 0.07 0.09 0.04 0.04 0.07 0.84 0.35 0.36

N = 79,186.

Table A3
The relationship between pre-innovation activity and different measures of innovation diffusion success.
(1) (2) (3) (4) (5) (6) (7)
Awareness Evaluation Implementation
Views Downloads Likes Collects Copies Derivatives Success Index

Collect 0.103*** 0.0734*** 0.113*** 0.165*** 0.0294*** 0.0144*** 0.0652***

(10.61) (9.47) (11.46) (15.78) (7.86) (5.81) (12.41)
Copy 0.0734*** 0.0233* 0.102*** 0.113*** 0.0298*** 0.00517 0.0456***
(6.63) (2.54) (8.74) (9.06) (6.23) (1.57) (7.14)
Comment 0.0775*** 0.0278** 0.0379** 0.0541*** 0.00571 0.00886** 0.0270***
(6.83) (3.01) (3.23) (4.31) (1.22) (2.69) (4.20)
Intermediate Skills 0.171*** 0.134*** 0.170*** 0.174*** 0.0152*** 0.00274 0.0771***
(17.77) (17.15) (17.01) (16.39) (3.90) (1.02) (14.31)
Advanced Skills 0.362*** 0.307*** 0.398*** 0.402*** 0.0888*** 0.0150*** 0.196***
(24.25) (24.49) (24.70) (23.74) (12.34) (3.33) (21.54)
Biography Reported 0.0753*** 0.0584*** 0.0849*** 0.0647*** 0.0249*** 0.00993*** 0.0433***
(7.22) (6.86) (7.77) (5.61) (5.53) (3.30) (7.22)
Days to Innovation − 0.0178*** − 0.0106*** − 0.0103*** − 0.0124*** − 0.00390*** 0.0000185 − 0.00679***
(− 7.56) (− 5.52) (− 4.19) (− 4.78) (− 3.92) (0.03) (− 5.07)
N 79,186 79,186 79,186 79,186 79,186 79,186 79,186
R2 0.376 0.487 0.198 0.208 0.084 0.052 0.278

Notes: OLS point estimates with t-values based on robust standard errors in parentheses. All specifications control for the linear and squared term of the days since a
thing has been posted, for category (80), release year (11), release month (12), and release day (7) fixed effects, and a constant but these terms are not reported.
Asterisks denote significance levels (*** p < 0.001, ** p < 0.01, * p < 0.05).

12
J. Claussen and M.A. Halbinger Research Policy 50 (2021) 103943

References Jenkins, H., Deuze, M., 2008. Convergence Culture. Sage Publications Sage UK, London,
England.
Jenkins, H., Ford, S., Green, J., 2018. Spreadable media: Creating Value and Meaning in a
Aaltonen, A., Seiler, S., 2015. Cumulative growth in user-generated content production:
Networked Culture. NYU Press.
evidence from Wikipedia. Manage. Sci. 62 (7), 2054–2069.
Jeppesen, L.B., Lakhani, K.R., 2010. Marginality and problem-solving effectiveness in
Allen, R.C., 1983. Collective invention. J. Econ. Behav. Organ. 4 (1), 1–24.
broadcast search. Organ. Sci. 21 (5), 1016–1033.
Anderson, C., 2012. Makers: The New Industrial Revolution. Crown Business, New York,
Jeppesen, L.B., Laursen, K., 2009. The role of lead users in knowledge sharing. Res.
NY.
Policy 38 (10), 1582–1589.
Autio, E., Dahlander, L., Frederiksen, L., 2013. Information exposure, opportunity
Karau, S.J., Williams, K.D., 1993. Social loafing: a meta-analytic review and theoretical
evaluation, and entrepreneurial action: an investigation of an online user
integration. J. Pers. Soc. Psychol. 65 (4), 681.
community. Acad. Manage. J. 56 (5), 1348–1371.
Kyriakou, H., Nickerson, J.V., Sabnis, G., 2017. Knowledge reuse for customization:
Baldwin, C., Hienerth, C., Von Hippel, E., 2006. How user innovations become
metamodels in an open design community for 3D printing. Manage. Inf. Syst. Q. 41
commercial products: a theoretical investigation and case study. Res. Policy 35 (9),
(3), 315–332.
1291–1313.
Landau, R., Rosenberg, N., Engineering, N.A., 1986. The Positive Sum Strategy: Harnessing
Baldwin, C., von Hippel, E., 2011. Modeling a paradigm shift: from producer innovation
Technology for Economic Growth. National Academies Press.
to user and open collaborative innovation. Organ. Sci. 22 (6), 1399–1417.
Laursen, K., Salter, A., 2006. Open for innovation: the role of openness in explaining
Benkler, Y., 2006. The Wealth of Networks: How Social Production Transforms Markets and
innovation performance among UK manufacturing firms. Strateg. Manage. J. 27 (2),
Freedom. Yale University Press.
131–150.
Benkler, Y., Nissenbaum, H., 2006. Commons-based peer production and virtue.
Leiponen, A., Helfat, C.E., 2010. Innovation objectives, knowledge sources, and the
J. Political Philos. 14 (4), 394–419.
benefits of breadth. Strateg. Manage. J. 31 (2), 224–236.
Breschi, S., Lenzi, C., 2016. Co-invention networks and inventive productivity in US
Lerner, J., Tirole, J., 2002. Some simple economics of open source. J. Ind. Econ. 50 (2),
cities. J. Urban Econ. 92, 66–75.
197–234.
Browder, R.E., Aldrich, H.E., Bradley, S.W., 2019. The emergence of the maker
Lilien, G.L., Morrison, P.D., Searls, K., Sonnack, M., Hippel, E., 2002. Performance
movement: implications for entrepreneurship research. J. Bus. Ventur. 34 (3),
assessment of the lead user idea-generation process for new product development.
459–476.
Manage. Sci. 48 (8), 1042–1059.
Buehler, E., Branham, S., Ali, A., Chang, J.J., Hofmann, M.K., Hurst, A., Kane, S.K., 2015.
Lindenberg, S., 2013. Social rationality, Seljf-regulation, and l/Vell-Being: the regulatory
Sharing is caring: assistive technology designs on thingiverse. In: Paper presented at
significance of needs, goals, and the self. In: The Handbook of Rational Choice Social
the Proceedings of the 33rd Annual ACM Conference on Human Factors in
Research, 72.
Computing Systems.
Lindtner, S., Hertz, G.D., Dourish, P., 2014. Emerging sites of HCI innovation:
Burtch, G., He, Q., Hong, Y., & Lee, D.D. 2019. Peer awards increase user content
hackerspaces, hardware startups & incubators. In: Paper presented at the
generation but reduce content novelty. Available at SSRN 3465879.
Proceedings of the SIGCHI Conference on Human Factors in Computing Systems.
Burtch, G., Hong, Y., Bapna, R., Griskevicius, V., 2018. Stimulating online reviews by
March, J.G., 2010. The Ambiguities of Experience. Cornell University Press.
combining financial incentives and social norms. Manage. Sci. 64 (5), 2065–2082.
Martin, B., de Jong, J.P., Ben-Menahem, S., von Krogh, G., von Hippel, E., Franke, N., &
Carlile, P.R., 2002. A pragmatic view of knowledge and boundaries: boundary objects in
Füller, J. Free household sector innovation: paradigm and implications special issue
new product development. Organ. Sci. 13 (4), 442–455.
of research policy in 2018–2020 lead editor. 2020.
D’Adderio, L., 2001. Crafting the virtual prototype: how firms integrate knowledge and
McCue, T. 2019. Significant 3D printing forecast surges to $35.6 billion: http://www.
capabilities across organisational boundaries. Res. Policy 30 (9), 1409–1424.
forbes.com.
Dahlander, L., Frederiksen, L., 2012. The core and cosmopolitans: a relational view of
McIntyre, D.P., Srinivasan, A., 2017. Networks, platforms, and strategy: emerging views
innovation in user communities. Organ. Sci. 23 (4), 988–1007.
and next steps. Strateg. Manage. J. 38 (1), 141–160.
Dahlander, L., Magnusson, M.G., 2005. Relationships between open source software
Nambisan, S., Wright, M., Feldman, M., 2019. The digital transformation of innovation
companies and communities: observations from Nordic firms. Res. Policy 34 (4),
and entrepreneurship: progress, challenges and key themes. Res. Policy 48 (8),
481–493.
103773.
Dahlander, L., Wallin, M.W., 2006. A man on the inside: unlocking communities as
Nonaka, I., 1994. A dynamic theory of organizational knowledge creation. Organ. Sci. 5
complementary assets. Res. Policy 35 (8), 1243–1259.
(1), 14–37.
Dasgupta, S., Hill, B.M., 2018. How “wide walls” can increase engagement: evidence
Orr, J.E., 2016. Talking About machines: An Ethnography of a Modern Job. Cornell
from a natural experiment in scratch. In: Paper presented at the Proceedings of the
University Press.
2018 CHI Conference on Human Factors in Computing Systems.
Rashid, A.M., Ling, K., Tassone, R.D., Resnick, P., Kraut, R., Riedl, J., 2006. Motivating
de Jong, J.P., Gillert, N.L., Stock, R.M., 2018. First adoption of consumer innovations:
participation by displaying the value of contribution. In: Paper presented at the
exploring market failure and alleviating factors. Res. Policy 47 (2), 487–497.
Proceedings of the SIGCHI conference on Human Factors in computing systems.
de Jong, J.P., von Hippel, E., Gault, F., Kuusisto, J., Raasch, C., 2015. Market failure in
Rogers, E.M., 2003. The Diffusion of Innovation, 5th edition. Free Press, New York.
the diffusion of consumer-developed innovations: patterns in Finland. Res. Policy 44
Ryall, M.D., Sorenson, O., 2007. Brokers and competitive advantage. Manage. Sci. 53 (4),
(10), 1856–1865.
566–583.
Ellis, H.C., 1965. The Transfer of Learning. The Macmillan Company, New York.
Samila, S., Sorenson, O., 2017. Community and capital in entrepreneurship and
Fagerholm, F., Münch, J., 2012. Developer experience: concept and definition. In: Paper
economic growth. Am. Sociol. Rev. 82 (4), 770–795.
presented at the Proceedings of the International Conference on Software and
Schumpeter, J., 1942. Creative destruction. Capital. Soc. Democr. 825, 82–85.
System Process.
Shaw, C.H., 2018. Why Do Makers Make? Examining Designer Motivations on
Faraj, S., Johnson, S.L., 2011. Network exchange patterns in online communities. Organ.
Thingiverse. com.
Sci. 22 (6), 1464–1480.
Smith, A. 2017. Social Innovation, Democracy and Makerspaces. SWPS 2017-10.
Fogliatto, F.S., Da Silveira, G.J., Borenstein, D., 2012. The mass customization decade: an
Available at SSRN: https://ssrn.com/abstract=2986245.
updated review of the literature. Int. J. Prod. Econ. 138 (1), 14–25.
Srinivasan, A., Venkatraman, N., 2018. Entrepreneurship in digital platforms: a network-
Foss, N.J., Frederiksen, L., Rullani, F., 2016. Problem-formulation and problem-solving
centric view. Strateg. Entrep. J. 12 (1), 54–71.
in self-organized communities: how modes of communication shape project
Stanko, M.A., 2016. Toward a theory of remixing in online innovation communities. Inf.
behaviors in the free open source software community. Strateg. Manage. J. 37 (13),
Syst. Res. 27 (4), 773–791.
2589–2610.
Thomke, S.H., 1998. Managing experimentation in the design of new products. Manage.
Franke, N., Shah, S., 2003. How communities support innovative activities: an
Sci. 44 (6), 743–762.
exploration of assistance and sharing among end-users. Res. Policy 32 (1), 157–178.
Tran, A., Zeckhauser, R., 2012. Rank as an inherent incentive: evidence from a field
Gallus, J., 2016. Fostering public good contributions with symbolic awards: a large-scale
experiment. J. Public Econ. 96 (9–10), 645–650.
natural field experiment at Wikipedia. Manage. Sci. 63 (12), 3999–4015.
Vaccaro, A., Veloso, F., Brusoni, S., 2009. The impact of virtual technologies on
Gambardella, A., Raasch, C., von Hippel, E., 2016. The user innovation paradigm:
knowledge-based processes: an empirical study. Res. Policy 38 (8), 1278–1287.
impacts on markets and welfare. Manage. Sci. 63 (5), 1450–1468.
Van der Vegt, G.S., Bunderson, J.S., Oosterhof, A., 2006. Expertness diversity and
Haefliger, S., Von Krogh, G., Spaeth, S., 2008. Code reuse in open source software.
interpersonal helping in teams: why those who need the most help end up getting the
Manage. Sci. 54 (1), 180–193.
least. Acad. Manage. J. 49 (5), 877–893.
Halbinger, M.A., 2018. The role of makerspaces in supporting consumer innovation and
von Hippel, E., 1986. Lead users: a source of novel product concepts. Manage. Sci. 32 (7),
diffusion: an empirical analysis. Res. Policy 47 (10), 2028–2036.
791–805.
Hargadon, A., Sutton, R.I., 1997. Technology brokering and innovation in a product
von Hippel, E., 1988. The Sources of Innovation. Oxford University Press, New York, NY.
development firm. Adm. Sci. Q. 716–749.
von Hippel, E., 1994. Sticky information” and the locus of problem solving: implications
Harhoff, D., Henkel, J., von Hippel, E., 2003. Profiting from voluntary information
for innovation. Manage. Sci. 40 (4), 429–439.
spillovers: how users benefit by freely revealing their innovations. Res. Policy 32
von Hippel, E., 2017. Free Innovation. MIT Press, Cambridge, MA.
(10), 1753–1769.
von Hippel, E., Katz, R., 2002. Shifting innovation to users via toolkits. Manage. Sci. 48
Hatch, M., 2013. The Maker Movement Manifesto: Rules for Innovation in the New
(7), 821–833.
World of Crafters, Hackers, and Tinkerers. McGraw Hill Professional.
von Krogh, G., von Hippel, E., 2006. The promise of research on open source software.
Henderson, R.M., Clark, K.B., 1990. Architectural innovation: the reconfiguration of
Manage. Sci. 52 (7), 975–983.
existing product technologies and the failure of established firms. Adm. Sci. Q. 9–30.
Wohlers, T., 2019. Wohlers Report 2019 3D Printing and Additive Manufacturing State
Hielscher, S., Smith, A., 2014. Community-Based Digital Fabrication Workshops: A
of the Industry Annual Worldwide Progress Report. Wohlers Associates, Inc., Fort
Review of the Research Literature. Working Paper Series, Brighton. 2014/08, 2014.
Collins, Colorado, USA.
Hippel, E., Krogh, G., 2003. Open source software and the “private-collective” innovation
model: issues for organization science. Organ. Sci. 14 (2), 209–223.

13
DOI: 10.1111/jpim.12565

C AT A LY S T A R T I C L E

Accelerating innovation: Some lessons from the pandemic

Robert G. Cooper 1,2

1
De Groote School of Business,
McMaster University, Hamilton, ON, Abstract
Canada The Pandemic taught us that accelerated new-­product development is more impor-
2
Institute for the Study of Business tant than ever, and provided examples of firms developing breakthrough products
Markets, Smeal College of Business
Administration, Pennsylvania State
in record time. Five approaches to accelerated development are outlined here: The
University, State College, PA, USA first two deal with adequately resourcing new-­product projects, namely the use of
focused teams; and effective portfolio management to prioritize projects and real-
Correspondence
Robert G. Cooper, De Groote School locate resources. Newer digital tools are outlined that speed new-­products devel-
of Business, McMaster University, opments. Finally, two development methods are described that move development
Hamilton, ON, Canada.
projects faster: Lean development and Agile development. Accelerated development
Email: [email protected]
also has hidden costs: undertaking less innovative projects and cutting too many cor-
Editors: Jelena Spanjol and ners. Although important, the topic is under-­researched, and the limited research has
Charles H. Noble.
yielded inconclusive results about acceleration's expected benefits.

Practitioner Points
• The Pandemic has focused the spotlight on the need for accelerated product innova-
tion, and also revealed ways to get to market faster that are generally applicable.
• Focused project teams coupled with effective portfolio management –­fewer pro-
jects but better projects –­increases resources on projects to maximize speed.
• New digital tools –­from Virtual Reality to AI –­can both enhance and accelerate
new-­product projects.
• Lean Development removes waste and inefficiencies in the idea-­to-­launch process
through value stream mapping.
• Managers should also overlap stages in their new-­product process and move key
decision points forward to minimized the impact of long lead-­time tasks.
• Agile methods borrowed from the software world can be built into more traditional
gating processes to yield increased productivity and faster development times.

KEYWORDS
accelerated innovation, accelerated new-­product development, Agile development, Agile-­Stage-­
Gate, portfolio management, Stage-­gate

1  |  H OW TO AC C E L E R AT E N PD over the years on the topic, and some research has been done to
investigate the various acceleration methods and their impacts.
The goal of developing a COVID-­19 vaccine in record time In spite of the research and publications, however, the quest
has shone the spotlight on the need for accelerated product de- for accelerated innovation has proved elusive—­many firms
velopment. A number of articles and books have been written continue to plod along at development much has they have for
Related Commentaries to this catalyst can be found at: https://doi.org/10.1111/jpim.12568, https://doi.org/10.1111/jpim.12569, https://doi.org/10.1111/jpim.12570,
https://doi.org/10.1111/jpim.12571
© 2021 Product Development & Management Association

J Prod Innov Manag. 2021;38:221–232.  |

wileyonlinelibrary.com/journal/jpim     221
 |
222       COOPER

years—­and also many questions about the “how's and why's” resources are spread too thinly across too many projects,
of accelerated innovation remain unanswered. Indeed, a com- so that every project, even the important ones, are under-­
prehensive review of accelerated product development con- resourced (Elonen & Artto, 2003). The lack of needed re-
cludes that “the performance implications of shortening NPD sources is one of the fundamental reasons why projects take
cycle time are still not fully understood.” (Katrin et al., 2013). so long to get to market, and is also blamed for shortcuts
The pandemic heightened interest in accelerated de- taken to save time and effort, that later come back and haunt
velopment, as we watched as pharmaceutical compa- the project team (Cooper et al., 2004).
nies moved quickly to develop vaccines at unheard-­ of Benchmarking studies show that top-­ performing busi-
speeds. But they were not alone: Other firms—­perhaps nesses are considerably more focused than others, with
less spectacularly—­also responded at high speed (Cooper, dedicated resources for product innovation: Half the top
2020). Corning Glass, a long-­time innovator in the field businesses use dedicated teams for projects, and more than
of glass and ceramics, leveraged their glass-­ceramic tech- half have fully dedicated product-­innovation groups that only
nology to create a virus-­fighting anti-­microbial innovation, work on new products (Cooper et al., 2004). Consider this
Guardiant®, glass-­ceramic particles that enable coatings example, Project Lightspeed:
to kill COVID-­19 on painted surfaces. Another is Heron
Gruppe, the Austrian producer of production automation On a Friday in late January 2020, Dr. Ugur
equipment, who used their IT skills to develop SAFEDI Sahin, received an email with news about a
or “safe distance control,” a very precise COVID-­19 de- deadly new coronavirus in China (Pancevski &
tection system for use on the production floor. SnapCab, Hopkins, 2020). The following Monday, the sci-
a manufacturer of portable glass-­and-­aluminum privacy entist and CEO of biotech firm BioNTech SE
pods for use as meeting rooms for the open office, quickly in Mainz, Germany, summoned his board to an-
pivoted, developing and launching a medical testing pod, a nounce that the company would start work on a
mini home-­office pod, and the “God Pod” for private meet- COVID-­19 vaccine. To accomplish this task in
ings in a church. And Marelli, an Italian-­based automotive record speed, BioNTech staff was divided into
components maker, quickly developed its new virus-­killing two dedicated teams working seven-­ days per
cabin air-­filter for shared-­use vehicles, such as taxis and week (separate teams to avoid transmission).
ambulances (Weissler, 2020). BioNTech had later partnered with Pfizer; and
Exemplary stories perhaps, but they prove the point that by early December, just 10.5  months after the
when needed, firms can respond quickly and pivot with inno- project start, the new vaccine, perhaps the most
vations. The trouble is that just doing what is already being significant product development in recent times,
done, but working harder and faster, will not yield the desired was ready for roll-­ out: the first COVID-­ 19
result. A recent Forbes article about pandemic pivots points out vaccine approved in the developed world (on
that: “It takes an accelerated innovation process with a rede- December 9 in Canada and shortly after in the
signed rather than a merely time-­wise compressed innovation USA).
journey.” (Mayer, 2020) Thus, the need is to re-­think the firm's
innovation process and methods—­moving to new approaches While a 100% dedicated team is ideal for maximum speed,
and a re-­designed process. Here, are five practices that can be this is not always practical for many manufacturers. For example,
built into a new or re-­invented innovation process in order to get there are waiting times in such projects (waiting for equipment
new products to market faster. arrival or test results), so team members work on other projects;
and team members often have other important duties. Thus, to
ensure time commitment to the project, an appropriate model
2  |  FO C U S E D P RO JE C T T E A MS for most product-­developers is to deploy a “focused team”—­
not quite 100% dedicated as in the BioNTech or Agile develop-
The most frequently cited challenge in new-­product port- ments, but almost; that is, to use a mostly dedicated team with
folio management is too many development projects in the some core team members devoting the majority (60%–­75%) of
pipeline (Cooper & Sommer, 2020; Dalton, 2016; Edgett, their time to the project. Some firms, such as Tetra Pak, a global
2013; Thomke & Reinertsen, 2012). Portfolio managers are packaging company, limit the number of “other projects” per
“normally concerned and overwhelmed with issues like the team member to one or two, so that for core team members,
prioritization of projects and the continuous distribution of this new-­product project is really their principal job. Thus, the
personnel from the different projects to overcome the urgent core team is protected from outside disruptions that divert their
crises.” (Amaral & Araújo, 2009) The result is that people attention from the project (Cooper & Fürst, 2020a).
COOPER   
|
   223

3   |   F EW ER P RO JE C TS, B E T T ER the line. The goal is to do fewer projects but better projects,
PROJ EC TS ! resource them properly, and get them done!

The development of the Moderna vaccine at warp speed has Example: ABC Company, a manufacturer of
taught us that given unlimited resources, time-­to-­market can baby-­care products near Vienna, had too many
be cut dramatically. (Moderna received $2.5 billion in fund- active projects in their portfolio. At the portfolio
ing from the U.S. government) (Clouse, 2020). But most review, the CEO complained that new-­product
firms do not have access to such unlimited resources, but at projects were taking too long, including his
least they can avoid the opposite situation: under-­resourcing two “executive sponsored” projects. The port-
projects to the point where they move at a crawl. Effective folio manager produced a list of the 16 active
portfolio management is one solution to getting focused on projects—­Table 1—­which we ranked from best
the right development projects, and right number of projects. to worst using both the Project Attractiveness
Here, are some ways (Cooper et al., 2004): Score (based on a scorecard method) and the
Productivity Index, both shown in Table 1. Also
shown is the full-­time equivalent people (FTEs)
3.1  |  Prioritizing projects at regular needed to get the project done on time, and the
portfolio reviews current resource allocation (FTEs).

At these reviews, senior management goes through the list Almost all projects were under-­resourced (actual
of active development projects, checks that there is the right versus needed resources), with some projects
mix and balance, but most important, does a forced ranking having less than one-­half of an FTE assigned;
of projects:1 to N. Resources per project are then added down and the two exec-­sponsored projects were con-
the list. When the resource limit is reached, a line is drawn: suming 66% of the resources. The total resource
Those projects below the line are killed or put on-­hold, and requirements for all 16 projects was 2.3 times
their resources are re-­allocated to the better projects above the 26 FTEs available! After ranking the projects

T A B L E 1   Forced ranking of projects at a quarterly portfolio review

Project
Current attractiveness Productivity New Required Resources Cumulative
Project name resources (FTEs) score/100 Index ranking resources (FTEs) after ranking resources
Ambo 0.5 88.1% 340 1 3.0 3.0 3.0
Mama 0.1 85.2% 318 2 3.0 3.0 6.0
Thermo 1.0 86.0% 300 3 3.0 2.0 8.0
Pump 12.0 79.2% 265 4 12.0 12.0 20.0
Retro 2.5 74.0% 250 5 5.0 3.0 23.0
Soother-­2 0.5 77.6% 223 6 4.0 3.0
Primo 0.2 72.2% 211 7 3.0 0.0 Limit 26.0
Dental-­A 5.2 70.0% 190 8 5.2 0.0 26.0
Cup-­Hold 1.4 72.4% 186 9 2.0 0.0 26.0
Toddler-­BR 0.2 74.5% 199 10 2.0 0.0 26.0
Dental-­B 0.2 68.1% 203 11 3.0 0.0 26.0
Cup-­Warm 0.1 68.0% 200 12 3.5 0.0 26.0
Baby-­Sit 0.5 67.6% 191 13 4.5 0.0 26.0
Inset 1.0 65.5% 187 14 4.0 0.0 26.0
Baby-­BR 0.2 61.0% 192 15 2.0 0.0 26.0
Situ 0.4 64.1% 150 16 1.5 0.0 26.0
Total 26.0 60.7 26.0
Notes: “Current resources” are what project is currently receiving, that is, FTEs (technical people) now working on the project. “Required sources” is what the Project
Leader had requested at the previous gate in order to get the project done reasonably on time, FTEs. “Resources After Ranking” are the resources the project ended
up with as a result of the current Portfolio Review decisions. Projects below the line (in italics) were put on hold: 10 of the 16 original projects. There are 26 FTE
technical people available.
 |
224       COOPER

and adding up resources per project (final col- • Strategic fit and importance.
umn), the resource limit was reached after the • Product advantage—­unique superior product with a com-
sixth project. Thus, 10 of the 16 projects were pelling value proposition.
put on hold, and their resources re-­ allocated • Market attractiveness—­ large, growing, low competitive
to the top six projects. As a result, besides one intensity.
exec-­sponsored project, the other five top prior- • Ability to leverage the firm's core competencies in the new
ity projects saw their resources increased by a project.
factor of 3.0, and could now be accelerated. A • Likelihood of technical success—­size of technical gap,
simple procedure, but very effective. technical complexity.
• Potential for financial reward.
• Risk level (negative)—­investment versus reward, uncer-
tainty of key assumptions.
3.2  |  Productivity index
At the gate meeting, following presentation of the project
Ranking projects by their NPVs or some similar profit metric by the project team, senior management (the gatekeepers),
is not the way to prioritize development projects. The result independently of each other, score the project on the criteria,
is a sub-­optimal list of projects. Instead, the Productivity 0–­10, using a scorecard. Gatekeepers’ scores are displayed on
Index is an effective way to prioritize projects when there are the screen, and major differences are debated. A decision is
constrained resources, which is usually the case (the method then reached, and if “Go,” the resources are committed.
is based on the theory of constraints) (Matheson et al., 1994). Such models are often better predictors of new-­product
The Productivity Index is simply the value of the project di- outcomes than are financial models, which are traditionally
vided by the constraining resource, usually person-­days or unreliable.1 And the transparent decision process that scoring
dollars. It gauges what value is added for every additional models enable produces a fruitful discussion, and usually re-
unit of scarce resource—­ person-­days or dollars—­ that is sults in more thoughtful Go/Kill decisions.
spent on the project:

Output Project NPV Project NPV

Productivity Index = = or
Input Dollars remaining to be spent Person-days needed to complete the project

In practice, one simply takes the NPV from the project's 4  |   DIGITAL TOOL S TO
Business Case, and divides by the person-­days or dollars that ACCELERATE KNOWLEDGE
must be spent in order to complete the project; and then one GENERATION
rank-­orders the projects by this index until out of resources, as in
the example above. Note that sunk costs are not relevant to this Numerous digital tools are available to accelerate the
prioritization model, only the “go forward costs” are counted. product-­development process (Cooper & Fürst, 2020b). A
Ranking projects by this Productivity Index is one way recent survey of 200 firms reveals that they expect their
to cull out the lower “bang for buck” projects—­they fall to investments in digital product development to increase
the bottom of the ranking list. Done correctly, this method their efficiency and performance by 19% over the next 5
maximizes the value of the development portfolio for a given years; and also expect to reduce time-­to-­market by 17%
resource level. (Geissbauer et al., 2019).
Newer technologies have made prototype development
easier, faster, and less expensive. Rapid prototyping based
3.3  |  Qualitative scoring models on 3D printing was the precursor of these new prototyping
tools; a rapid prototype could be used not only to test techni-
Numerous factors have been shown to correlate strongly with cal facets of the product's design, but also to seek customer
new-­product success and profitability (Cooper, 2013, 2019). feedback and validation for the proposed new product. Rapid
As a result, research-­based scoring models have been devel- prototyping, now much cheaper and more ubiquitous, par-
oped that predict new-­product outcomes reasonably well, as tially solves the number one reason for new-­product failure,
high as 83% accurately, pre-­Development (Bronnenberg & namely the lack of understanding of customers’ needs. As
Engelen, 1988). Many of the models were initially developed Steve Jobs, never a proponent of traditional market research,
privately within corporations, but in recent years, more are famously said “People do not know what they want until you
found in the public domain (Cooper, 1985). Seven proven show it to them.” (Isaacson, 2011) Rapid prototyping makes
criteria for such a scoring model include: this possible. And today, 3D printing has gone beyond just
COOPER   
|
   225

prototypes: Finished products are now being 3D printed. For which of its components would provoke an immune response
example, more than a third of the components in GE’s new (Waltz, 2020).
advanced turboprop engine are made by 3D printing (Sheynin
& Bovalino, 2017).
Other digital testing technologies are being used in a sim- 5  |  LEAN DEVELOPM ENT
ilar way. Simulations not only test new products technically,
but also allow customer-­testing of a product that does not yet Many firms’ new-­product processes have become bureau-
exist. For example, Volvo Construction evaluates new truck cratic and slow. Value stream analysis is a well-­ known
designs before a working prototype is built using a real-­time Lean-­ Six-­
Sigma methodology, designed to remove waste
simulator, not unlike a flight simulator.2 And Alphabets’ and inefficiencies from business processes; it has seen wide-
(Google's) self-­driving car, Waymo, logged 10 billion miles, spread success in factory-­floor settings. But the Lean method
all done by simulation (Etherington, 2019). In a similar vein, can also be applied to new-­product development, specifically
developers create a digital twin (or virtual protype) of the to remove waste and to make the idea-­to-­launch system more
new product—­a virtual or digital replica. Sensors collect efficient (Fiore, 2005). Removing unneeded tasks, cutting
real-­time data from the physical item, which is used to op- the preparation of too-­many PPT presentations, and delet-
erate the digital duplicate, allowing it to be understood, ana- ing work that adds no value all save time. And re-­designing
lyzed, or optimized. the process to overlap tasks, namely parallel processing—­
More recently, Virtual and Augmented Realities are being starting one task or stage before the preceding one is 100%
used to test early versions of products, long before the real complete—­can save time too (as the COVID-­19 vaccine de-
product is developed. VR and AR can be used to simulate the velopers did, with a “rolling approval process” in the United
user's environment so that test-­customers can try the product, States; Figure 1 shows this accelerated five-­stage process)
not just in a lab, but in its intended setting. For example, in (U.S. Government Accountability Office, 2020). Other ways
the development of FedEx's new drop-­box, an early prototype to cut cycle time that a value stream analysis may identify are
was first developed from cardboard (Termini, 2018). With to move some of the key decision points forward—­for exam-
potential users fitted with VR goggles, developers were able ple, the decision to purchase production equipment (acquir-
to allow users to see and use the prototype in a variety of ing equipment is often a long lead-­time item). Making this
environments, for example, at their own homes. decision before the field trials are completed is risky, but a
AI is also beginning to be used to predict the outcomes of cost-­benefit analysis often reveals that the payoffs of a faster
different technical solutions—­for example, molecules—­thus launch through rapid manufacturing are worth the risk.
accelerating the choice of the right technical solution. For ex- In practice, in value stream analysis, a task force of
ample, medical chemists must guess which compounds might knowledgeable product-­developers maps the current idea-­to-­
make good medicines; thus they synthesize and test countless launch process for a typical project in the business—­every
variants, and most are failures. The same is true for mate- step, activity, procedure, and decision point. Then they walk
rial science—­coming up with a new molecule for a material. through the process and identify all work that adds no value,
Much of this innovation thus involves making predictions and also tasks that take too long.3 Problem-­solving methods
based on data (Rotman, 2018). But AI and machine learn- are employed to eliminate the sources of delay or to shorten
ing uses real-­world data and analysis to predict outcomes, lengthy tasks or steps. For example, Danfoss, the Danish
faster, and cheaper than traditional lab-­testing methods. For controls company, undertook such a Lean exercise, and over
example, the design or “invention” of a new product can be a 3-­year period, and cut the cycle time for major projects
accelerated by machine learning to select the right molecule from development-­approval through to launch by half—­a re-
and hence help to create new materials for solar panels or markable acceleration of the process at relatively little cost.
batteries. (Jørgensen, 2018) Perhaps not coincidentally, this Lean ap-
The pharmaceutical industry has recognized the power of proach is consistent with Agile principles, which value sim-
AI: “The step-­wise serial process of R&D will change to be plicity, defined as “the art of maximizing the amount of work
hyper-­iterative and integrated, so that real-­world information not done.”
coming back from development will, in real time, change
the research being conducted.” (Mishcon & Robinson, Example: AK produces a key component used
2019) Companies such as Amgen, Pfizer, Novartis, Sanofi, in a large piece of production equipment in
GlaxoSmithKline, and Merck now have partnerships with AI the paper industry. Although the development
firms to discover new drug candidates for a range of diseases. process seemed fairly straightforward, projects
For example, machine-­learning systems and computational were taking much longer than they should have.
analyses played an important role in the COVID-­19 vaccine A small task force of project leaders undertook
quest, helping researchers understand the virus, and predict a value stream analysis exercise: They mapped
 |
226       COOPER

F I G U R E 1   Accelerated model for COVID vaccine development from the U.S. GAO. [Colour figure can be viewed at wileyonlinelibrary.com]

out the entire process, idea to launch, for a typi- and changing information. While a number of different
cal major project on a long sheet of paper along Agile methods were proposed, the Agile field finally came
a wall, about 10 meters in length: every activity, together in 2001 as a set of principles outlined in the Agile
step, and decision point. An abbreviated version Manifesto, (Beck et al., 2001) complete with development
of their 10-­meter map is in Figure 2. Then the methodologies with clear rules (Schwaber & Sutherland,
task force went through the map, step by step, 2017).
and eliminated tasks that were not needed and Agile is incremental and iterative, a series of build-­test-­
worked to reduce the time of lengthy tasks. and-­revise iterations; it is adaptive and evolutionary—­the
product definition and project plan change as the project
The longest task in Figure 2 is the field trials, moves forward; it emphasizes frequent and fast delivery of
8-­14  months, sometimes with several recycles, results (e.g., product versions), in rhythmic takt time; and it
thus up to three years for field trials! The task is based on self-­managed project teams.
force then used root cause analysis (a fishbone Agile (the Scrum method, the most popular) works like
diagram) to identity the main causes of this. Over this for software development: Agile breaks the development
a dozen causes were identified, including that the process into a series of short, iterative, and incremental time-­
customer had to shut down their production in boxed sprints, each typically about 2 weeks long. At the be-
order to do a field trial, at some cost; and often the ginning of each sprint, the development team holds a sprint
field trial failed due to lack of prior technical test- planning meeting to agree on sprint goals and create a task
ing, resulting in recycles. For each of the dozen plan for the sprint. Once the sprint is underway, the team
causes, the task force, with other invited experts, meets every morning—­their daily stand-­ups or scrums—­to
brainstormed and arrived at numerous solutions. ensure that work is on course, share information, and solve
The result was a significant reduction in the field problems. At the end of each sprint, software increments, po-
trial times, and the elimination of recycles. tentially releasable, are demonstrated to stakeholders, both
management and customers. Finally, the team meets in a ret-
rospective meeting to review how they can improve the way
they work and execute the next sprint better. The team then
6  |  AG ILE DE V E LO P ME N T plans the next sprint based on customer and management
feedback.
Agile Development methods were developed in the 1990s In Agile development, the team is dedicated 100% to the
in the software world to deal with projects facing uncertain one project and co-­located physically. Agile defines new
 COOPER

ƵƐƚŽŵĞƌĞĮŶĞĚ EĞǁͲ
WƌŽũĞĐƚ W^d
Ͳ ^ƉĞĐŝĂůŝƐƚƐ ŽƌWWDd
&ŽƌŵĂůůLJ WƌŽĚƵĐƚ
Ͳ WƌŽĚƵĐƚDŐƌ 'Žͬ<ŝůů
KƉĞŶĞĚ Ͳ EWŐƌŽƵƉ /ĚĞĂ
Ͳ ƌŽŽŵ W^d
Ͳ ůĞĂĚĞƌ ŽƌWWDd
'Žͬ<ŝůů
Ͳ ƐĐŽƉĞ
Ͳ ĚĞůŝǀĞƌĂďůĞƐ /ŶƚĞƌŶĂůůLJĞĮŶĞĚ <ŝůů
ZĞĮŶĞĚ
Ͳ ƌĞƐŽƵƌĐĞƐ Ͳ W
/ŶƚĞƌŶĂůĮŶĞƚƵŶŝŶŐĂŶĚŝƚĞƌĂƟŽŶ
Ͳ ŵŝůĞƐƚŽŶĞƐ Ͳ EW'ƌŽƵƉ
ƚŽŐĞůŝĚĞĂ
<ŝůů Ͳ WƌŽĚƵĐƚDŐƌ

ϮǁĞĞŬƐ ϮŵŽŶƚŚƐ ϮǁĞĞŬƐ

ZĞǁŽƌŬ
ZĞǁŽƌŬ ZĞǁŽƌŬ ZĞǁŽƌŬ ZĞǁŽƌŬ

ĞǀĞůŽƉŵĞŶƚ WƌŽƚŽƚLJƉŝŶŐ
Ͳ ĚĞƐŝŐŶĮŶĂů Ͳ ĨƵůůƐĐĂůĞ
Ͳ ǁŽƌŬŝŶŐŵŽĚĞů W^d Ͳ ƐĞĂŵƉƌŽŽĮŶŐ W^d W^d
Ͳ ůĂďƚĞƐƟŶŐ ŽƌWWDd Ͳ ǀĂůŝĚĂƚĞĞĸĐŝĞŶĐLJ ŽƌWWDd
&ŝĞůĚdƌŝĂůƐ ŽƌWWDd
ͲƵƐƚŽŵĞƌƐŝƚĞ
>ĂƵŶĐŚ
Ͳ ƚĞƐƚĚĞǀĞůŽƉ 'Žͬ<ŝůů Ͳ ǀĂůŝĚĂƚĞĐŽƐƟŶŐ 'Žͬ<ŝůů 'Žͬ<ŝůů
Ͳ ŵĂƚĞƌŝĂůƐ Ͳ ƌĞƚŽŽůŝŶŐ Ͳ &ĞĞĚďĂĐŬ
Ͳ ƉƌŽŐƌĞƐƐƌĞǀŝĞǁ Ͳ ƐƉĞĐƐĮŶĂůŝnjĞ
Ͳ ƉƌĞůŝŵĐŽƐƟŶŐ Ͳ ƚƌĂŝŶŝŶŐ

<ŝůů <ŝůů <ŝůů

ϰͲϲŵŽŶƚŚƐ ϯͲϱŵŽŶƚŚƐ ϴͲϭϰŵŽŶƚŚƐ ϮŵŽŶƚŚƐ

F I G U R E 2   The value stream map for a typical major project in the example company. [Colour figure can be viewed at wileyonlinelibrary.com]
  
|   227
 |
228       COOPER

h^dKDZŽƌh^Z
^ZhD ^ZhD ^ZhD ^ZhD ^ZhD
^ZhD

h/> s>KWͲ
/ KEWd h^/E^^ DEdΘ s>/d/KE >hE,
^WZ/Ed ^WZ/Ed ^WZ/Ed ^WZ/Ed ^WZ/Ed ^WZ/Ed
^WZ/Ed^ ^WZ/Ed ^ d^d/E' ^WZ/Ed^ ^WZ/Ed^
^WZ/Ed^ ^WZ/Ed^

^ƚĂŐĞϭ ^ƚĂŐĞϮ

'ĂƚĞ 'ĂƚĞ 'ĂƚĞ 'ĂƚĞ 'ĂƚĞ

ϭ Ϯ ϯ ϰ ϱ

/ƚĞƌĂƟŽŶƐŽƌƐƉƌŝŶƚƐ;ĂďŽƵƚϯǁĞĞŬƐůŽŶŐͿŽĐĐƵƌǁŝƚŚŝŶƚŚĞƐƚĂŐĞƐŽĨƚŚĞ^ƚĂŐĞͲ'ĂƚĞƉƌŽĐĞƐƐ
;ŵĂŶLJŝƚĞƌĂƟŽŶƐǁŝƚŚŝŶůŽŶŐĞƌƐƚĂŐĞƐƐƵĐŚĂƐĞǀĞůŽƉŵĞŶƚͿ͘ĂĐŚŝƚĞƌĂƟŽŶƐƚĂƌƚƐǁŝƚŚĂ
WůĂŶŶŝŶŐDĞĞƟŶŐ͕ΘĞŶĚƐǁŝƚŚĂĞŵŽƚŽƐƚĂŬĞŚŽůĚĞƌƐ͕ĨŽůůŽǁĞĚďLJĂZĞƚƌŽƐƉĞĐƚ͘

F I G U R E 3   The Agile-­Stage-­Gate model with Agile project management built Into the stages. [Colour figure can be viewed at wileyonlinelibrary.com]

roles—­the Scrum Master and Product Owner, but no tradi- 2016). What leading manufacturers have done is to build
tional project leader or project manager. The method is also Agile project management methods into the stages of their
very visual, with tools such as the burndown chart, Kanban traditional gating process, replacing the more rigid project
chart, and sprint backlog visibly displayed in the team room. management methods such as Gantt charts, timelines, and
Although Agile Development has its roots in the soft- milestones. (Cooper & Fürst, 2020a) A typical gated model
ware industry where it has become widely accepted, firms with Agile built in is shown in Figure 3. By integrating Agile
that produce physical products have benefited from Agile methods with traditional gating models, leading firms such
Development methods in recent years (Cooper & Sommer, as Honeywell, GE, and LEGO have dramatically reduced

/ŶĐƌĞĂƐĞĚŇĞdžŝďŝůŝƚLJƚŽƌĞĂĐƚƚŽĐŚĂŶŐĞƐ ϰ͘Ϯϰ

/ŵƉƌŽǀĞĚƚĞĂŵŵŽƌĂůĞͬŵŽƟǀĂƟŽŶ ϯ͘ϴϬ

/ŶĐƌĞĂƐĞĚĐƵƐƚŽŵĞƌƐĂƟƐĨĂĐƟŽŶ ϯ͘ϳϵ

/ŶĐƌĞĂƐĞĚĐŽŵŵŝƚŵĞŶƚƚŽƉƌŽũĞĐƚďLJŝŶǀŽůǀĞĚƉĞŽƉůĞ ϯ͘ϳϵ

/ŶĐƌĞĂƐĞĚƉƌŽĚƵĐƟǀŝƚLJŽĨĚĞǀĞůŽƉŵĞŶƚƉƌŽũĞĐƚ ϯ͘ϲϬ

/ŵƉƌŽǀĞĚĂĚŚĞƌĞŶĐĞƚŽƟŵĞƐĐŚĞĚƵůĞ ϯ͘ϱϬ

ZĞĚƵĐĞĚƉƌŽũĞĐƚƌŝƐŬ;ƚĞĐŚĨĞĂƐŝďŝůŝƚLJ͖ƉƌŽũĞĐƚĨĂŝůƵƌĞͿ ϯ͘ϰϬ

/ŵƉƌŽǀĞĚƉƌŽĚƵĐƚƋƵĂůŝƚLJ ϯ͘ϭϳ

^ŚŽƌƚĞƌƉƌŽĚƵĐƚĚĞǀĞůŽƉŵĞŶƚ;ƟŵĞƚŽŵĂƌŬĞƚͿ ϯ͘ϭϭ

Hardly Any Ϭ ϭ Ϯ ϯ ϰ ϱ
Value Very Large
Value of Agile (1-5) Value

F I G U R E 4   Performance Improvements realized using Agile for physical new products. [Colour figure can be viewed at wileyonlinelibrary.com]
COOPER   
|
   229

time to market as well as getting the product right. (Cooper the time schedule are also benefits (Atzberger et al., 2020;
& Sommer, 2018). Invention Center, 2018; Schmidt et al., 2018).

Example: Chamberlain, a manufacturer of

remote-­controlled devices (garage door and com- 7  |   ACCELERATION IS NOT A
mercial gate openers) moved to its hybrid model, PANACEA
Agile within Stage-­ Gate, in 2013. They were
one of the first in North America. The company Not all the promised benefits of accelerated development
had used a traditional stage-­and-­gate process for exist. The theory is clear: Faster to market means higher
product development for years, but an ever-­larger profits (sales and profits are realized sooner; and money has
percentage of each new-­product project entailed a time value). Being in the market first results in higher mar-
software development for Smartphone connec- ket share, hence more sales and profits—­“first mover advan-
tivity. Thus, the conflict: Stage-­Gate or Agile? tage”; and being late to market means missing the window
of opportunity, or incurring other penalties (e.g., annoying
The solution was to introduce the concept of valued customers dependent on the product).
“Agile within Stage-­Gate”, whereby the firm uses The logic appears sound, but the research evidence is
Agile sprints and scrums for both physical and lacking. For example, there has never been a consistent set of
IT development, and in doing so, improved de- research results to support the argument that “first in wins.”
velopment efforts considerably across all groups Often the first to get to market does better, but not always;
and all project types (Cooper & Sommer, 2018). sometimes the “second in” learns from the mistakes of the
Agile is employed in the development and testing pioneer, and does better. Indeed, a review of empirical studies
stages of their Stage-­Gate system: Each stage is reveals that research has produced inconsistent, even conflict-
divided into a series of sprints, each lasting ex- ing, results on the relationship between accelerated develop-
actly three weeks. Project teams have 100% ded- ment and project success (Chen et al., 2005, Langerak, 2010).
icated team members for each project; but given One major problem with trying to prove that accelerated
that dedicated teams are not feasible for every development has economic payoffs is the challenge of putting
project, the firm only uses this Agile-­Stage-­Gate an economic value on hypothesized benefits. For example,
approach for about 20% of their projects, specifi- how does one measure the economic gains from “a possible
cally the larger, major revenue generator projects. increase in market share because the product was launched
After seven years using the system, the firm es- sooner”; or the economic losses of “missing the window of
timates “a 20% to 30% cycle time reduction be- opportunity” because the product was slow to market (oppor-
cause there is much less ‘redo’ in projects now”, tunity costs are always tricky to estimate)? Such economic
as well as improvements in productivity. metrics are difficult to gauge, simply because an assumption
must be made about “what would have been” if the product
This new Agile-­Stage Gate approach yields three important were launched sooner or perhaps later….a hypothetical situa-
positive results for manufacturers (Cooper & Fürst, 2020a): tion. Some companies have compared the success of projects
done before versus those done after implementing acceler-
1. Development is faster: Sprints are timeboxed—­ a hard ated development (profits, sales etc.) (Jørgensen, 2018); but
stop; the team is focused, and partially or fully dedi- that requires a large sample of projects.
cated; and frequent scums resolve problems immediately. Some research on accelerated development looked at
The result is higher productivity, closer adherence to more easily measured dependent variables, such as the im-
the time schedule, and shorter development cycles. pact of speed on development costs; these studies often yield
2. Gets the product right: Product designs (features, function- negative results. While “development cost” is more reliably
ality, etc.) are validated by customers (and management) as measured, it may not be a valid metric, however: One usu-
the project moves along—­early, often and cheaply. ally moves quickly not to save money, but in order to gain
3. Team morale is higher: The team is self-­managed, self-­ competitive advantage and market share. Indeed, some of
organizing, ideally co-­located, and has decision authority. the actions taken done in order to move quickly, such as
those prescribed by Agile—­dedicated teams, daily scrums
Specific gains when Agile is applied to manufacturers’ meetings, multiple iterations, demos to customers and
projects are in Figure 4 from several large-­sample studies. management—­are resource intensive and may actually cost
The greatest impacts are increased flexibility to react to more. Other metrics also produce mixed results (Langerak,
changes, improved team morale, and higher customer satis- 2010): For example, the relationship between cycle time
faction; shorter development times and more adherence to and product quality is also unclear: One study found that
 |
230       COOPER

higher product quality is related to decreases in cycle time 1. Does accelerated innovation really work—­does it yield
(Langerak, 2010), while others suggest the opposite effect benefits, specifically which benefits and how much?
(Griffin, 2002). Yet, another study found shorter cycle time 2. What are the hidden costs of accelerated innovation?
was correlated with an increase in the new product's sales 3. Which proposed methods work the best, why, and under
when it is managed across all of the stages of the process which conditions?
(Katrin et al., 2013). Conflicting outcomes regarding the
importance of accelerated development for new-­ product One reason for the inconclusive results regarding the ben-
success may also occur because the impact of cycle time re- efits of accelerated innovation is the unreliable and often in-
duction may pale when compared to other key success driv- valid metrics that try to capture the benefits of speed to market.
ers; thus its effect is swamped by the noise (Cooper, 2013, A fourth research question focuses on the implementa-
2019). tion of acceleration methods. Agile methods were first im-
Cycle time reduction may also yield major negatives plemented in the software world; but developing a physical
(Crawford, 1992). In order to cut cycle time, a firm may avoid product is much different: it's difficult to do demos on-­line
bolder innovations, which often involve learning and exper- to customers; and it's expensive to pivot and change a prod-
imentation and thus take longer; the firm may instead focus uct's design near the end of development. Thus manufactur-
on smaller, less challenging projects. There is evidence to ers have been forced to make significant modifications to the
support this fear: NPD cycle time was found to be cut signifi- Agile approach. But more research is required here, not just
cantly over a 14-­year period, based on data from two PDMA for Agile, but for the other acceleration methods too. Only
best practices studies (Adams, 2004); but in that same period, then will we have the knowledge, confidence, and motivation
innovative products dropped almost in half as a percent of the to move forward with accelerated innovation.
total portfolio of projects—­from 20.4% to 11.5% (Cooper,
2005). CONFLICT OF INTEREST
A second major negative is cutting corners and mak- The author's work herein was not funded by any corporation
ing mistakes—­cutting short key activities such as VoC or or government agency.
front-­end homework. But these same activities have been
shown to have a strong impact on new-­product success, and ENDNOTES
hence are vital. One reason for insufficient VoC and front-­ 1 Based on analyses of actual versus forecasted sales or profits of
end homework is limited resources; but the desire to reduce NPD projects within firms. One study in the public domain is from
time-­to-­market impacts negatively on resources available Procter & Gamble; the “50% success rate” is defined as “Actual NPV/
Forecasted NPV”: Mills. 2007. Implementing a Stage-­Gate Process at
(people x time) (Cooper & Edgett, 2002). Time demands
Procter & Gamble. Stage-­Gate Leadership Summit, Conference, St.
may also truncate the process, where important stages in
Petersburg Beach, FL
a well-­designed system maybe be skipped—­for example,
2 Volvo Construction Equipment streamlines product development with
skipping the validation stage in the process (production
a real time, human-­in-­the-­loop simulator. MathWorks® User Stories:
trials or extended field trials), simply because that stage is https://www.mathw​orks.com/compa​ny/user_stori​es/volvo-­const​ructi​
deemed unnecessary given the tight deadline. on-­equip​ment-­strea​mline​s-­produ​ct-­devel​opmen​t-­with-­a-­real-­time-­
Project teams, driven by time, may also become too com- human​-­in-­the-­loop-­simul​ator.html (accessed February 15, 2021).
mitted to their project and its plan, and fail to pivot when 3 Lean value stream analysis method, see: Cooper, R.G. 2017. Winning
needed. Finally, high priority projects tend to chew up a lot at New Products: Creating Value Through Innovation 5th ed. New
of resources—­pulling team members from other important York, NY: Basic Books, Perseus Books Group: 171-­180.
duties to focus on the one project.

R E F E R E NC E S
8  |   CO NC LUS ION Adams, Marjorie. 2004. PDMA Foundation New Product Development
Report Initial Findings: Summary of Response From 2004 CPAS.
Accelerated innovation is a worthy goal in product develop- Chicago, IL: PDMA Foundation.
Amaral, António, and Madalena Araújo. 2009. “Project Portfolio
ment, and has the potential for delivering many benefits, as
Management Phases: A Technique for Strategy Alignment.” World
witnessed in developments during the pandemic. Five ap-
Academy of Science, Engineering and Technology International
proaches to accelerated innovation have been o­ utlined  —­from Journal of Economics and Management Engineering 3 (10): 1919–­27.
better focus and portfolio management through to Agile and Atzberger, Alexander, Simon Jakob Nicklas, Johan Schrof, Stefan
Lean development. Weiss, and Kristi Paetzold. 2020. Agile Development of Physical
Many unanswered questions remain however, and thus ac- Products: A Study on the Current State of Industrial Practice.
celerated innovation is a fertile field for academics to investi- Neubiberg, Germany: Universitat der Bundeswehr Munchen.
gate. Three closely related research questions are: https://doi.org/10.18726/​2020_10 (accessed February 15, 2021).
COOPER   
|
   231

Beck, Kent, Mike Beedle, Arie van Bennekum, Alistair Cockburn, Crawford, C. Merle. 1992. “The Hidden Costs of Accelerated Product
Ward Cunningham, Martin Fowler, James Grenning, et al. 2001. Development.” Journal of Product Innovation Management 9 (3):
Manifesto for Agile Software Development. http://agile​manif​esto. 188–­92.
org/ (accessed February 15, 2021). Dalton, Mike. 2016. “Manage Pipeline Bandwidth to Avoid Derailing
Bronnenberg, J. J. A. M., and M. L. van Engelen. 1988. “A Dutch New Products.” Industry Week, August 23. https://www.indus​
Test With the NewProd Model.” R&D Management 18 (4): 321–­32. trywe​ek.com/innov​ation/​proce​ss-­impro​vemen​t/artic​le/22007​299/
Chen, Jiyao, Richard R. Reilly, and Gary S. Lynn. 2005. “The Impacts manag​e-­pipel​ine-­bandw​idth-­to-­avoid​-­derai​ling-­new-­products (ac-
of Speed-­to-­Market on New Product Success: The Moderating cessed February 15, 2021).
Effects of Uncertainty.” IEEE Transactions on Engineering Edgett, Scott J. 2013. “Portfolio Management for Product Innovation.”
Management 52 (2): 199–­212. In PDMA Handbook of New Product Development, 3rd ed., edited
Clouse, Allie 2020. “Moderna Vaccine Funded by Government by Kenneth B. Kahn, Chapter 9: 154–­66.Hoboken, NJ: Wiley.
Spending, with Notable Private Donation.” USA Today, November Elonen, Suvi, and Karlos A. Artto. 2003. “Problems in Managing
11. https://www.msn.com/en-­us/news/us/fact-­check​-­moder​na-­ Internal Development Projects in Multi-­Project Environments.”
vacci​ne-­funde​d-­by-­gover​nment​-­spend​ing-­priva​te-­donat​ions-­and-­ International Journal of Project Management 21 (6): 395–­402.
resea​rch-­grant​s/ar-­BB1bku1h (accessed February 15, 2021). Etherington, Darrell. 2019. “Waymo has now Driven 10 Billion
Cooper, Robert G. 1985. “Selecting Winning New Products: Using the Autonomous Miles in Simulation.” TechCrunch, July 10. https://
NewProd System.” Journal of Product Innovation Management 2 techc​r unch.com/2019/07/10/waymo​-­has-­now-­drive​n-­10-­billi​on-­
(1): 34–­44. auton​omous​-­miles​-­in-­simul​ation/ (accessed February 15, 2021).
Cooper, Robert G. 2005. “Your NPD Portfolio May Be Harmful to Your Fiore, Clifford. 2005. Accelerated Product Development: Combining
Business’s Health.” PDMA Visions XXIX (2) Apr-­May: 22–­26. Lean and Six Sigma for Peak Performance. New York: Productivity
Cooper, Robert G. 2013. “New Products—­What Separates the Winners Press.
from the Losers and What Drives Success.” In PDMA Handbook Geissbauer, Reinhard, Schrauf, Stefan, and Morr, Jochen-­ Thomas.
of New Product Development, 3rd ed., edited by Kenneth B. Kahn, 2019. Digital Product Development 2025: Agile, Collaborative,
Chapter 1: 25–­33. Hoboken, NJ: Wiley. AI Driven and Customer Centric. https://www.pwc.de/de/digit​ale-­
Cooper, Robert G. 2019. “The Drivers of Success in New-­Product trans​forma​tion/pwc-­studi​e-­digit​al-­produ​ct-­devel​opmen​t-­2025.pdf
Development.” Industrial Marketing Management 76: 36–­47. (accessed February 15, 2021).
Cooper, Robert G. 2020. “The Pandemic Pivot: The Need for Product, Griffin, Abbie. 2002. “Product Development Cycle Time for Business-­
Service and Business Model Innovation.” InnovationManagement. To-­Business Products.” Industrial Marketing Management 31 (4):
SE, December 2. https://innov​ation​manag​ement.se/2020/12/02/ 291–­304.
the-­pande​mic-­pivot​-­the-­need-­for-­produ​ct-­servi​ce-­and-­busin​ess-­ Invention Center. 2018. Consortium Benchmarking 2018 “Agile
model​-­innov​ation/ (accessed February 15, 2021). Invention”: Evaluation of the Study Results. Rheinisch-­
Cooper, Robert G, and Scott J Edgett. 2002. “The Dark Side of Time Westfälische Technische Hochschule Aachen (RWTH). https://
and Time Metrics in Product Innovation.” PDMA Visions XXVI www.konso​r tial​-b­ ench​marki​ng.de/en/2018a​gilei​nvent​ion.html
(22) Apr-­May: 14–­16. (accessed February 15, 2021).
Cooper, Robert G., Scott J. Edgett, and Elko J. Kleinschmidt. 2004. Isaacson, Walter. 2011. Steve Jobs: The Exclusive Biography. New
“Benchmarking Best NPD Practices-­ 2: Strategy, Resources York, NY: Simon & Schuster.
and Portfolio Management Practices.” Research-­Technology Jørgensen, Bo Bay 2018. “Agile Stage Gate at Danfoss.” GEMBA
Management 47 (3) May-­June: 50–­60. Innovation Conference, Copenhagen, April.
Cooper, Robert G., and Peter Fürst. 2020a. “Agile Development Katrin, Eling, F. Langerak, and Abbie Griffin. 2013. “A Stage-­Wise
for Manufacturers: The Emergent Gating Model.” Approach to Exploring Performance Effects of Cycle Time
InnovationManagement.SE, November 10. https://innov​ ation​ Reduction.” Journal of Product Innovation Management 30 (4)
manag​ement.se/2020/11/10/agile​-d­ evel​opmen​t-­for-­manuf​actur​ July: 626–­41.
ers-­the-­emerg​ent-­gatin​g-­model/ (accessed February 15, 2021). Langerak, Fred. 2010. Accelerated Product Development. Product
Cooper, Robert G., and Peter Fürst. 2020b. “Digital Transformation Innovation and Management, Part 5. Hoboken, NJ: Wiley. https://
and Its Impact on New-­Product Development for Manufacturers.” doi.org/10.1002/97814​44316​568.wiem0​5004
InnovationManagement. SE, March. https://innov​ation​manag​ement. Matheson, David, James E. Matheson, and Michael M. Menke. 1994.
se/2020/03/11/digit​al-­trans​forma​tion-­and-­its-­impac​t-­on-­new-­produ​ “Making Excellent R&D Decisions.” Research Technology
ct-­manag​ement​-­for-­manuf​actur​ers/ (accessed February 15, 2021). Management 37 (6) Nov-­Dec: 21–­4.
Cooper, Robert G., and Anita F. Sommer. 2016. “The Agile–­Stage-­Gate Mayer, Hannah M. 2020. “Innovation Due to Covid: Yes, But How?”
Hybrid Model: A Promising New Approach and a New Research Forbes, August. https://www.forbes.com/sites/​hanna​hmaye​
Opportunity.” Journal of Product Innovation Management 33 (5) r/2020/08/26/innov​ation​-­due-­to-­covid​-­yes-­but-­how/?sh=23c19​
Sept: 513–­526. b7b4b7e (accessed February 15, 2021).
Cooper, Robert G., and Anita F. Sommer. 2018. “Agile-­Stage-­Gate For Mishcon, Denise, and Robin Robinson. 2019. “Artificial Intelligence:
Manufacturers—­Changing the Way New Products are Developed.” Molecule to Market.” PharmaVOICE, January. https://www.
Research-­Technology Management 61 (2): Mar-­Apr: 17–­26. pharm​avoice.com/artic​le/2019-­01-­pharm​a-­ai/ (accessed February
Cooper, Robert G., and Anita F. Sommer. 2020. “New-­ Product 15, 2021).
Portfolio Management With Agile: Challenges and Solutions for Pancevski, Bojan, and Jared S. Hopkins. 2020. “How Pfizer Partner
Manufacturers Using Agile Development.” Research-­Technology BioNTech Became a Leader in Coronavirus Vaccine Race.”
Management 63 (1): 29–­36. Morningstar News—­Dow Jones, October 22. https://www.morni​
 |
232       COOPER

ngstar.com/news/dow-­jones/​20201​02245​42/how-­pfize​r-­partn​er-­ Weissler, Paul. 2020. “Marelli’s Covid-­Killing Cabin Purifier.” SAE

biont​ech-­becam​e-­a-­leade​r-­in-­coron​aviru​s-­vacci​ne-­race (accessed Mobilus, August 21. https://www.sae.org/news/2020/08/marel​ li-­
February 15, 2021). covid​-­cabin​-­filter (accessed February 15, 2021).
Rotman, David. 2018. “AI is Reinventing the Way We Invent.” MIT
Technology Review. https://www.techn​ology​review.com/s/61289​8/
ai-­is-­reinv​entin​g-­the-­way-­we-­inven​t/ (accessed February 15, 2021). AUTHOR BIOGRAPHY
Schmidt, Tobias Sebastian, Stefan Weiss, and Kristin Paetzold.
2018. Agile Development of Physical Products: An Empirical
Study About Motivations, Potentials and Applicability. Munich,
Germany: University of the German Federal Armed Forces. www.
Robert G. Cooper is a Fellow of the PDMA, ISBM
unibw.de/produ​ktent​wickl​ung-­en (accessed February 15, 2021). Distinguished Research Fellow at Pennsylvania
Schwaber, Ken, and Sutherland, Jeff. 2017. The Scrum Guide. http://www. State University's Smeal College of Business
scrum​guides.org/scrum​-g­ uide.html (accessed February 15, 2021). Administration, and Professor Emeritus, DeGroote
Sheynin, Dmitry, and Yari M. Bovalino. 2017. “A Treat for the AvGeeks: School of Business, McMaster University, Hamilton,
An Inside Look at GE’s 3D-­Printed Aircraft Engine.” GE Reports, Ontario, Canada. He pioneered the original research
July 24. https://www.ge.com/news/repor​ts/treat​-­avgee​ks-­insid​e-­ that led to many discoveries, including the key
look-­ges-­3d-­print​ed-­aircr​aft-­engine (accessed February 15, 2021).
success drivers in new-­ product development and
Termini, Brandon Scott. 2018. “How Handsome Used VR to Test
Products for FedEx.” Inside Design, August 21. https://www.invis​
the Stage-­Gate® Idea-­
to-­Launch process. He has
ionapp.com/insid​e-­desig​n/vr-­user-­testi​ng/ (accessed February 15, published more than 140 academic articles and 11
2021). books, including the best-­selling “Winning at New
Thomke, Stefan, and Donald Reinertsen. 2012. Six myths of product Products.”
development. Harvard Business Review 90 (5): 84–­94.
U.S. Government Accountability Office. 2020. Covid-­19 Vaccine
Development. May. https://www.gao.gov/asset​ s/710/707152.pdf
(accessed February 15, 2021).
How to cite this article: Cooper RG. Accelerating
Waltz, Emily. 2020. “AI Takes Its Best Shot: What AI Can—­And
Can’t—­ Do in the Race for a Coronavirus Vaccine.” IEEE
innovation: Some lessons from the pandemic. J Prod
Spectrum 57 (10), Oct: 24–­67. https://ieeex​plore.ieee.org/docum​ Innov Manag. 2021;38:221–232. https://doi.
ent/9205545 (accessed February 15, 2021). org/10.1111/jpim.12565
 Skills and Education for Additive
Manufacturing: A Review of Emerging Issues

Mélanie Despeisse1 ✉ and Tim Minshall2

( )

1
Industrial and Materials Science, Chalmers University of Technology, Gothenburg, Sweden
[email protected]
2
Institute for Manufacturing, University of Cambridge, Cambridge, UK
[email protected]

Abstract. The recent advances in digital technologies and in additive manufac‐

turing (AM) in particular are revolutionising our industrial landscape. These
changes require new engineering and management skills to exploit fully and
sustainably the benefits offered by these advanced technologies. The current talent
shortage calls for new education programmes to deliver a skilled, capable and
adaptable workforce. Existing courses on design, engineering and management
related to production and manufacturing do not systematically deliver the neces‐
sary skills and knowledge for an effective deployment of AM technologies. Based
on a literature review and evidence collected from multi-stakeholder workshops,
this paper presents the key themes for education programmes to address the
current skill gap and barriers to AM adoption and exploitation.

Keywords: Additive manufacturing · 3D printing · Education · Skills

1 Introduction

1.1 Background
Additive Manufacturing (AM), also commonly referred to as 3D printing, was first
developed in the 1980s [1, 2]. Early AM technologies were mainly used for rapid proto‐
typing and tooling [1]. Recent advances in AM have enabled its use for direct manu‐
facture of components and end-products for a broad range of applications.
AM builds artefacts from CAD models by adding material layer by layer [3], as
opposed to conventional machining techniques which are ‘subtractive’ as they cut mate‐
rial away to obtain the shape desired. The direct use of digital files for manufacturing
simplifies the process from design to production, enables the potential for mass custom‐
ization and can significantly reduce lead times [4, 5]. This ‘digitalisation’ trend in the
manufacturing industry is a game changer as (i) current limitations are progressively
being removed (e.g. design constraints, centralised manufacturing to achieve economies
of scale) and (ii) new limitations emerge (e.g. resource scarcity, cap on waste and emis‐
sions).

© IFIP International Federation for Information Processing 2017

Published by Springer International Publishing AG 2017. All Rights Reserved
H. Lödding et al. (Eds.): APMS 2017, Part I, IFIP AICT 513, pp. 289–297, 2017.
DOI: 10.1007/978-3-319-66923-6_34
290 M. Despeisse and T. Minshall

A new mindset is required for product and process design, manufacturing system
configurations and business models more broadly. These new rules of how, where and
when products are manufactured, distributed and used are still evolving. Accordingly,
education programmes need to be adapted [6] and shift towards systems-oriented solu‐
tions integrating science, engineering and management principles [7].

1.2 Current Additive Manufacturing Applications

The growth of AM utilisation has impacted all sectors of the economy:

Primary sector – raw and intermediate materials for AM. Currently, there are
limited options for standard AM materials. In metal powder manufacturing, the powders
used for AM are the same standard powders as for other manufacturing techniques such
as powder forging, spray deposition, metal injection moulding or hot isostatic pressing.
Although it is difficult to tailor materials for AM [8], it is possible and in fact required
for some high-value AM produced parts. In addition, powder reclamation for reuse and
recycling is a growing sector with particular promises in contributing to a more efficient
use of material at systems level.

Secondary sector – additively manufactured products. The most direct impact of

AM adoption is on the manufacturing sector, and especially for:
Aerospace. AM is particularly appropriate in aerospace applications as parts are
complex, need to be lightweight, high performance, produced in relatively small quan‐
tities, and made from advanced, expensive materials. New applications and level of
performance present challenges for current (and future) designers and engineers, e.g.
new design rules, new materials, new technology, etc.
Automotive. The automotive industry has used AM technologies since the late 80s
for rapid prototyping, tooling and custom parts in small production volumes [1], with a
recent focus on lightweight materials to improve vehicle fuel efficiency [9].
Medical. AM has been widely used for hearing aids, dental implants and prosthetics.
Further applications include with bio-printing to additively manufacture skin tissue and
human organs [9]. In the pharmaceutical industry, the technology shows potential to
produce custom-made pills to adapt to the patient specific needs, simplifying medication,
improving compliance to prescription and reducing risk of medication errors.

Tertiary and quaternary sector – services for AM or services using AM. The
service sector has also been impacted by the adoption of AM. Example of AM services
include AM machinery sales, leasing, maintenance, upgrade and training; design, manu‐
facturing and delivery as services [4]; development of design libraries providing CAD
templates, customized designs; certification, etc.
 Skills and Education for Additive Manufacturing 291

2 Research Method

This research addresses the general need to explore how to make the best out of AM
technologies, while avoiding lock-in or unsustainable and irresponsible use of the tech‐
nology. The work was carried out as part of the ‘Bit by Bit’ project1 and the activities
leading to the development of the UK National Strategy for AM2. As part of the AM
national strategy process, the main evidence was collected over the period March-
October 2015 [10] with additional data gathered during a series of working group
sessions in the period November 2015-December 2016, supplemented by desk-based
research drawing on secondary sources. The data revealed seven key themes (barriers
and opportunities) for AM adoption from a UK perspective [11]. The seven themes
identified are highly interconnected and overlap, with a repeated emphasis on the need
for skills, education and professional training across almost all the other themes. Thus
the skills and education theme was identified as a core issue for the development of the
national strategy. Prior to the finalisation of the UK National Strategy itself, an event
on ‘Identifying and Developing Additive Manufacturing Skills for UK Industry’ was
held in March 2017 to specifically focus on skills and education, drawing together 80
representatives of a wide range of AM relevant stakeholder organisations.
The next section presents the education and training issues identified to enable the
effective deployment of AM technologies in the short to long-term. It does so in three
steps: (1) first the barriers and industry needs (skills) for AM were reviewed from the
literature and the industry data available, (2) then a review of existing AM courses was
conducted, (3) finally, a gap analysis was performed to highlight the key topics that need
to be integrated into new education and training programmes.

3 Results

3.1 Review of Skills and Industry Needs for AM

There is a growing body of literature reviewing the barriers and enablers for AM adop‐
tion and exploitation [e.g. 8–17]. Some of the work particularly focuses on skills [6, 12–
16]. In addition, a synthesis of the findings from various workshops and surveys [6, 10,
11] is included to provide a broader view of the issues from industry, government and
academic perspectives. A summary of current industry needs is provided in Table 1. It
is important to note that the themes often overlap.

1
‘Bit by Bit: Capturing the Value from the Digital Fabrication Revolution’, UK Engineering
and Physical Sciences Research Council (EPSRC) award number EP/K039598/1.
2
http://www.amnationalstrategy.uk/.
292 M. Despeisse and T. Minshall

Table 1. Themes and perceived barriers to the adoption and exploitation of AM.
Issues/themes Summary of industry needs and common perceived barriers
Design & modelling Need for design guidelines, modelling software, 3D scanning, education
software programmes on design for AM, better understanding of new freedoms
and constraints, new focus on functionality, availability of AM-skilled
designers, security of design data
Materials, processes & Understanding properties of different processes, machines,
machines applications, size, costs, availability (e.g. intellectual property (IP),
independent suppliers). Use of mixed materials with implications for
recyclability and biocompatibility. AM-skilled operators, and
importance of post-processing
Skills & education Lack of appropriate skills preventing adoption, up-skilling current
workforce vs. training next generation. Public awareness and education.
3DP in schools, “teaching the teacher”. Balance the provision of
generalist and specialist skills through “mix-and-match” modular
programmes. Industry mentoring schemes, more hands-on teaching
methods. Access to open-source learning material
Quality, standards & Metrology, in-process inspection and controls. Need for data libraries
testing (designs). Lack of standards for processes, materials, software,
products, tests, etc. General vs. sector-specific standards. Tests for
higher volumes, non-destructive testing. Quality assurance, certification
and liability
Partnerships & Regional and national hubs or platforms, collaborative and community-
networks oriented networks (such as maker spaces) to raise awareness and provide
alternative educational pathways to learn about AM. Ensure that
knowledge and emerging industry needs are captured to teach industry-
relevant and up-to-date topics
Creativity & Cross-functional teamwork, ideation techniques and new teaching
innovation methods to foster creativity. Use of 3DP in schools to engage and inspire
the next generation of students
Energy & Increase the use of AM for green solutions, i.e. renewable and clean
sustainability energy such as AM-based materials and components to enhance fuel
cell performance. Need to reduce dependency on fossil resources and
the environmental burden of human activities. Need for life-cycle
considerations to assess social and environmental impact
Cost & financing Realistic estimate of costs compared to scale of opportunity to allow for
viable business cases. Funding to increase awareness and reduce risk of
adoption (e.g. scale-up, machine purchase) especially for SMEs
Understanding of full costs, including materials, post-processing and
testing
IP & security Balance need for openness to share knowledge with need for
commercial protection to capture value from investments, enforcement
of IP rights. Current security, IP and legal systems not appropriate for
digital networks. Global IP leakage. Cyber security concerns preventing
rapid AM adoption
 Skills and Education for Additive Manufacturing 293

3.2 Review of Education and Training Programmes

A non-exhaustive, exploratory review of existing education programmes – both Contin‐
uous Professional Development (CPD) and Higher Education (HE) – was conducted. A
total of 45 sources were identified. These include university undergraduate and post‐
graduate courses, from full programmes to short modules, professional training courses,
massive open online courses and downloadable packages. They allowed the key topics
being taught in existing programmes to be identified. Table 2 lists these topics along
with a description of typical course content and activities.

Table 2. Topics and activities in current AM education and training programmes.

Topic Description of typical courses and activities
AM fundamentals Lectures about advances in AM and production engineering.
Learning about a range of technologies, compatibilities between
AM and traditional techniques (pros & cons)
AM industry applications Case studies on current industrial applications. Examples of best
practices. Prototyping, tooling, direct manufacturing, mass
customization
Advanced process Physical sciences, applied mathematical, advanced engineering
engineering design, material science. Hands-on experience with a variety of
AM equipment. More in-depth understanding of material
characteristics, process parameters, and machine designs.
Advanced controls and parts testing. Process optimization.
Experiments to explore technology limitations
Design and modelling Reverse engineering and 3D laser scanning for direct digital
fabrication purposes. Hands-on labs to design parts on CAD.
Design for manufacture, material-specific design principles and
qualification approaches. Design software options for designers.
Ergonomics in design
Production economics & Method to improve productivity and reduce manufacturing costs.
business management Ensure products and services delivered to industry at the quality
and time required. Inventory management
Management and Teamwork, problem-solving skills, critical and analytical skills,
communication communication and management skills
Research & development Lab-based R&D in measurement methods, models,
instrumentation, sensors and data. Improve scientific
understanding of AM. Advances in process controls, equipment
and material properties. Industry-focused research project with
novel processes, materials and simulation

3.3 Summary of Skills for AM and Suggestions for Education Programmes

The list of topics taught in programmes available (Sect. 3.2) was compared to needs
collected from the literature and industry (Sect. 3.1) to identify the gaps in current
programmes available. The first observation was that there is still a relatively low
number of AM programmes available. It shows that (1) AM is not systematically taught
294 M. Despeisse and T. Minshall

in design and engineering curricula within universities, and (2) few institutions are
proposing specialised courses for AM. The second observation was that, while most
courses focused on technical aspect, some topics around softer issues are poorly covered
(if at all); e.g. IP, liability, quality assurance, sustainability and business models specially
as they relate to AM.
In summary, the following recommendations are made to enhance the AM skills for
designers, engineers and managers:
• Recognise AM as a family of technologies. Appreciate the value of AM as tool for
manufacturing along more traditional techniques, and as an enabler of innovation
(see design considerations below).
– Know when to use AM (and when not) by considering other manufacturing tech‐
niques and their respective strengths and limitations. AM may not be the best
solution: other traditional or mixed/hybrid methods could work better, be cheaper,
more efficient, result in better quality products, etc.
– Need to understand the appropriateness of various manufacturing techniques for
different materials and applications.
– Need for new process controls and quality assurance methods.
• Effectively design products for AM by shifting focus and using new tools.
– New levels of performance achieved through AM with innovative solutions based
on new design thinking (focus on functionality, performance, quality).
– Designing for AM requires new design approaches and modelling software, e.g.
3D scanning, new design freedoms and structures.
– New design guidelines should include considerations for emerging constraints,
e.g. limited availability of AM materials, resource scarcity for rare metals, design
for modularity, upgradability, disassembly, recycling, etc. (principles of design
for the environment).
• Understand AM processes and materials and their broader implications on the value
chain and business model.
– Economics of AM go beyond basic machine and material cost.
– Need for basic understanding of legal (e.g. IP, cyber security) and liability issues
(e.g. quality assurance and insurance) surrounding AM utilisation.
– Need for more responsible models of production and consumption (sustainable
products and services) through innovative business models.
– Need to consider the full impacts of AM from a life cycle perspective. It is
currently challenging to assess the sustainability of AM technology due to the lack
of tools and methods to do so.
• Acknowledge that AM technology is evolving quickly so the limitations we see today
may very well be removed the next year.
– AM creates new opportunities, new business models, new jobs, etc.
– Need to prepare current and next generation to respond effectively and keep an
open-mind as the rate of technological change is accelerating.
– Need to quickly adapt and update programmes to technology progress.
– Need to “teach the teachers” to avoid spreading misunderstanding and misper‐
ception of AM technologies and capabilities.
 Skills and Education for Additive Manufacturing 295

• Balance the provision of generalist skills and knowledge as well as more advanced
courses for specialists.
– Promote distributed and modular models for all educational levels.
– Need for open-access, free, online resources.
– Need for regional and national platforms and community-oriented networks (such
as maker spaces) to raise awareness and provide alternative educational pathways
to learn about AM.
• Prepare students for jobs in industry and support incumbent workers by ensuring that
AM courses and programmes are up-to-date and easy access.
– Raise awareness among the public and provide easy access to training programs
for workers and students by establishing collaborative and community-oriented
networks for AM awareness and education (i.e. above-mentioned alternative ways
to learn about AM).
– Support educational programs in STEAM (STEM + arts) and across a wide range
of learning environments to provide industry-relevant knowledge and experience,
e.g. tutoring, industry-based projects, lab-based activities.
– Need for new partnerships and innovation centres bringing industry and academic
institutions to keep education and research relevant and closely linked to industry
needs.

4 Conclusions

Research is still on-going to better understand the business opportunities created by AM

as well as barriers to realise the full benefits of AM. The challenges encountered by
industry in effectively using the technology are not fully understood. This paper attempts
to capture the current state of these industry challenges and barriers to AM adoption
which should be addressed by education programmes. Industry needs and challenges
were categorised and linked to specific sets of skills for AM. We then reviewed existing
education programmes and compared their content to the industry needs to characterise
the skills gap for AM adoption and exploitation. Finally, we made some initial recom‐
mendations for developing new education and training programmes to support the
provision of AM-skilled designers, engineers and managers.
The review of education programmes revealed that there are some courses available
providing AM content covering a broad range of issues as identified in the review of
industry needs. However the number of programmes available is low compared to
today’s high demand for AM skills. In addition, there are increasing concerns about the
content of these courses: Do they provide the necessary skills and knowledge to realise
the full potential of the technology? There is an urgent need to develop university and
other training programmes to increase awareness and educate the next generation of
workers and students in productive utilisation of AM technologies and generate some
societal “pull” for AM technologies [17].
Some of the most commonly mentioned barriers are lack of AM-specific design
skills, limited working materials, uncertainties in part qualification, lack of metrology
and production standards. Besides these technical barriers to AM adoption reported
296 M. Despeisse and T. Minshall

mainly in the literature, there is a clear need to better educate future engineers as unfa‐
miliarity with AM technologies is recognised as a major barrier for industrial adoption.
Finally, some of the issues poorly covered as highlighted in bold in Fig. 1. Further work
is still needed to consolidate the findings and recommendations made in this paper.

Fig. 1. Mindmap of recommendations for AM education and training.

Acknowledgments. This research was carried out as part of the UK Engineering and Physical
Sciences (EPSRC) funded project entitled ‘Bit by Bit: Capturing the Value from the Digital
Fabrication Revolution’, EP/K039598/1.

References

1. Langdon, R.: A decade of rapid prototyping. Automot. Eng. 22(4), 44–59 (1997)
2. Kruth, J.-P., Leu, M.C., Nakagawa, T.: Progress in additive manufacturing and rapid
prototyping. CIRP Ann. Manuf. Technol. 47(2), 525–540 (1998)
3. ASTM: ASTM F2792e12a: Standard Terminology for Additive Manufacturing Technologies.
ASTM International, West Conshohocken (2012)
 Skills and Education for Additive Manufacturing 297

4. Weller, C., Kleer, R., Piller, F.T.: Economic implications of 3D printing: market structure
models in light of additive manufacturing revisited. Int. J. Prod. Econ. 164, 43–56 (2015)
5. Atzeni, E., Salmi, A.: Economics of additive manufacturing for end-usable metal parts. Int.
J. Adv. Manuf. Technol. 62(9–12), 1147–1155 (2012)
6. Simpson, T.W., Williams, C.B., Hripko, M.: Preparing industry for additive manufacturing
and its applications: summary & recommendations from a National Science Foundation
workshop. Addit. Manuf. 13, 166–178 (2017)
7. Swearengen, J.C., Barnes, S., Coe, S., Reinhardt, C., Subramanian, B.: Globalization and the
undergraduate manufacturing engineering curriculum. J. Eng. Educ. 91(2), 255–261 (2003)
8. Gu, D.D., Meiners, W., Wissenbach, K., Poprawe, R.: Laser additive manufacturing of
metallic components: materials, processes and mechanisms. Int. Mater. Rev. 57(3), 133–164
(2012)
9. Guo, N., Leu, M.C.: Additive manufacturing: technology, applications and research needs.
Front. Mech. Eng. 8(3), 215–243 (2013)
10. Dickens, P., Minshall, T.: Update Report 1: How Was The Evidence Collected? (2015)
11. Dickens, P., Minshall, T.: Update Report 2: What Did The Initial Evidence Reveal? (2015)
12. Gao, W., Zhang, Y., Ramanujan, D., Ramani, K., Chen, Y., Williams, C.B., Wang, C.C.L.,
Shin, Y.C., Zhang, S., Zavattieri, P.D.: The status, challenges, and future of additive
manufacturing in engineering. Comput. Aided Des. 69, 65–89 (2015)
13. Ford, P., Dean, L.: Additive manufacturing in product design education: out with the old in
and in with the new? In: Proceedings of the 15th International Conference on Engineering
and Product Design Education (EPDE 2013), pp. 611–616 (2013)
14. Junk, S., Schrock, S.: New developments in design education for additive manufacturing. In:
Proceedings of the 18th International Conference on Engineering and Product Design
Education (EPDE 2016), pp. 290–295 (2016)
15. Williams, C.B., Seepersad, C.C.: Design for additive manufacturing curriculum: a problem-
and project-based approach. In: Proceedings of the 23rd Annual International Solid Freeform
Fabrication Symposium (SFF 2012), pp. 81–92 (2012)
16. Minetola, P., Iuliano, L., Bassoli, E., Gatto, A.: Impact of additive manufacturing on
engineering education – evidence from Italy. Rapid Prototyp. J. 21(5), 535–555 (2015)
17. Bourell, D.L., Leu, M.C., Rosen, D.R.: Roadmap for additive manufacturing: identifying the
future of freeform processing. The University of Texas at Austin (2009)
 Technovation 31 (2011) 627–637

Contents lists available at ScienceDirect

Technovation
journal homepage: www.elsevier.com/locate/technovation

Measuring industry clockspeed in the systemic industry context

Ozgur Dedehayir n, Saku J. Mäkinen
CITER, Center for Technology and Innovation Research, Department of Industrial Management, Tampere University of Technology, PO Box 541, FI 33101 Tampere, Finland

a r t i c l e i n f o abstract

Available online 10 August 2011 Industry clockspeed has been used in earlier literature to assess the rate of change of industries but this
Keywords: measure remains limited in its application in longitudinal analyses as well as in systemic industry
Systemic industry contexts. Nevertheless, there is a growing need for such a measure as business ecosystems replace
Technological system standalone products and organisations are required to manage their innovation process in increasingly
Reverse salient systemic contexts. In this paper, we firstly derive a temporal measure of technological industry
Industry clockspeed clockspeed, which evaluates the time between successively higher levels of performance in the
industry’s product technology, over time. We secondly derive a systemic technological industry
clockspeed for systemic industry contexts, which measures the time required for a particular sub-
industry to utilise the level of technological performance that is provisioned by another, interdepen-
dent sub-industry. In turn, we illustrate the use of these measures in an empirical study of the systemic
personal computer industry. The results of our empirical illustration show that the proposed clock-
speeds together provide informative measures of the pace of change for sub-industries and systemic
industry. We subsequently discuss the organisational considerations and theoretical implications of the
proposed measures.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction adapt to their dynamic, competitive environments (Bourgeois and

Eisenhardt, 1988; Hannan and Freeman, 1977; Van de Ven and
Industries change over time, thereby altering the competitive Poole, 1995; Subramanian et al., 2011).
environment of organisations (McGahan, 2004; Eisenhardt and Industry change has been measured using different para-
Brown, 1998; Porter, 1985). The performance as well as the meters, such as the number of organisations in the industry
survival of organisations is dependent on how they negotiate the (Chesbrough, 2003; Suarez and Utterback, 1995; Carroll and
changing industry through appropriate technological choices Hannan, 1989), the number of industry patents (Agarwal, 1998;
along the time path of industry evolution (Gort and Wall, 1986; Brockhoff et al., 1999), and the industry’s annual turnover
Strebel, 1987; Rice and Galvin, 2008). In making these technolo- (Gooroochurn and Hanley, 2007). In this paper we focus on the
gical choices, organisations additionally face the need to manage clockspeed measure of industry change, which differentiates
their innovation process in an intertwined environment where between industries by comparing their rates of change in product
actors form business ecosystems (Tiwana et al., 2010; Adner and technology, process technology, and organisational capability
Kapoor, 2010; Adner, 2006). However, the pace of industry change (Fine, 1998). While the clockspeed measure has so far been used
itself varies over time (Klepper and Graddy, 1990; Suarez and to evaluate the rate of change of specific industries, it has not
Lanzolla, 2005), emphasising the need for organisations to deploy been used to measure the rate of change of the industry in its
technological strategies in a timely fashion and engage in time- systemic industry context. Nevertheless, there is a growing need
based competition (Stalk, 1988; Bessant et al., 1996). In this for such a measure as business ecosystems replace standalone
dynamic environment, organisations face increasing time pressure products (Tiwana et al., 2010) and organisations are required to
to innovate (Gupta and Sounder, 1998), which calls for flexible manage their innovation process in a systemic context.
new product development processes that facilitate both rapid Here, by systemic industry we refer to a hierarchical network of
and anticipative management of the timing of innovations sub-industries (Ethiraj and Puranam, 2004) that specialise in
(MacCormack et al., 2001). Consequently, knowledge concerning producing the interdependent technical sub-systems of a hier-
the speed of change in the industry can benefit organisations to archically structured technological system (Murmann and
Frenken, 2006; Tushman and Murmann, 1998; Clark, 1985). In
this manner, sub-industries are interdependently connected to
n
Corresponding author. Tel.: þ358 44 586 8514; fax: þ 358 3 3115 2207.
one another due to the interrelatedness of their output technol-
E-mail address: ozgur.dedehayir@tut.fi (O. Dedehayir). ogies. Systemic industries may therefore produce complex product

0166-4972/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.technovation.2011.07.008
628 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637

systems (CoPS) such as aircraft engines, flight simulators, and that are witnessed throughout the industry’s development. From an
offshore oil platforms, (Hansen and Rush, 1998; Hobday, 1998; organisational standpoint, the evolution of the industry is important,
Miller et al., 1995), and modular systems such as PCs (personal since different phases pose varying conditions that influence the
computers) and automobiles (Baldwin and Clark, 1997; Langlois performance as well as the survival of the organisation (Suarez and
and Robertson, 1992; Persson and Åhlström, 2006) that integrate Utterback, 1995; Jovanovic, 1982; Jovanovic and MacDonald, 1994;
functionally interdependent sub-systems, produced by their own Agarwal et al., 2002; Agarwal and Gort, 1996). Moreover, organisa-
specialised sub-industries, into holistic systemic products. tions occupy ‘macro-niches’ and ‘micro-niches’ that pose differing
In systemic industries, a clockspeed measure that compares competitive landscapes (Baum and Singh, 1994), which also have
the pace of change of a particular sub-industry with respect to the differing dynamics in their pace of change.
pace of change of other, interdependent sub-industries can Industries not only change due to competitive dynamics but
greatly aid organisations to cope with their dynamic environ- do so at different rates, subsequent, for example, to the differing
ment. This is because the new product development (NPD) pace of introduction and utilisation of technological innovations
processes of organisations in systemic industries are intertwined in the industry (Klepper and Graddy, 1990; Suarez and Lanzolla,
with performance improvements in the sub-system technologies 2005; Audretsch, 1995). Consequently, organisations must imple-
that are central to their own sub-industry, as well as with the ment and dynamically alter strategies to cope with the different
interfaces between other, interdependent sub-industries that rates of change in the industry’s evolution (Eisenhardt and
produce interdependent technological sub-systems (Ethiraj and Martin, 2000; Eisenhardt and Tabrizi, 1995; Teece et al., 1997;
Puranam, 2004). Moreover, an organisation inside a sub-industry Helfat et al., 2007), which are manifest, for instance, in the
may remain more competitive by utilising information about innovations and technological changes (e.g. Romanelli and
the rate of change of interdependent sub-industries because Tushman, 1994). The changing pace of industry development
performance deficiencies in complementary technologies may thus requires the timely modification of the organisation’s activ-
jeopardise the ability of the organisation to create its own value ities and assets in order to remain competitive in the industry
for end-users (Adner and Kapoor, 2010). A clockspeed measure for (McGahan, 2004; Eisenhardt and Brown, 1998; McGahan, 2000).
the systemic context could therefore aid organisations to recog- Under rapidly changing circumstances, organisations may
nise whether their systemic industry and its sub-industries can be engage in time-based competition that centres on improving
considered to lie within moderately changing or high-velocity manufacturing, reducing sales and distribution time, and gener-
markets (Eisenhardt and Martin, 2000), and act accordingly ally improving innovation capability (Stalk, 1988; Bessant et al.,
through NPD processes. 1996), as readily witnessed in fast-cycle industries (Bourgeois and
In this paper we aim to extend the current understanding of Eisenhardt, 1988; Williams, 1992) such as the computer hard-
the rate of industry change by firstly developing an additional ware and software industry (Eisenhardt and Tabrizi, 1995; Rosas-
measure of industry clockspeed, namely, technological industry Vega and Vokurka, 2000). Such transformations of organisational
clockspeed. This clockspeed measure is premised on the product activities have been shown to be effective if they do not purely
technology measure proposed by Fine (1998), and measures the concentrate only on productivity gains but at the same time
time between successively higher levels of performance in the consider innovativeness, timing, and flexibility (Benner and
industry’s product technology across time. Secondly, we develop Tushman, 2003; Brown and Eisenhardt, 1997). Especially in
a clockspeed measure for systemic contexts, namely, systemic dynamic environments, organisations engage in time-based com-
technological industry clockspeed, by considering the perfor- petition and exploration to focus on reducing the time to market
mance discrepancies that emerge between interrelated sub-sys- of their innovations (Bayus, 1997; Cohen et al., 1996), in other
tems over time. In these states of technological disequilibrium words, the duration of time required for the conception, devel-
(Rosenberg, 1976), we focus on the reverse salient sub-system opment, production, and delivery of the innovation (Datar et al.,
(Hughes, 1983), which utilises the lowest level of performance 1997; Meyer and Utterback, 1995).
provisioned by the system’s technological performance frontier Different metrics have consequently been developed to evaluate
and thus hinders the performance delivery of the systemic and explain the rate of change of industries. However, in this paper
product. In turn, the systemic technological industry clockspeed we specifically focus on the clockspeed measure of industry change
measures the duration of time required for the reverse salient introduced by Fine (1998). The clockspeed of industries has been
sub-system to utilise the provisioned level of performance. We analysed in the literature with various methods, such as the
demonstrate the use of these clockspeed measures in an illus- changes in product technology using the proxy of the rate of new
trative, longitudinal study of the systemic PC (personal computer) product introductions, the changes in process technology using the
industry, focusing on specific sub-industries that produce the proxy of the rate of obsolescence of process technologies, and the
sub-systems, which are central for computer gaming on the changes in organisational capability using the proxy of the rate of
holistic PC system. change in organisational structure (Fine, 1998). A number of
quantitative studies have subsequently illustrated these methods
in the analysis of endogenous and exogenous factors that lead to
2. Theoretical background different rates of change in industries (Nadkarni and Narayanan,
2007; Souza et al., 2004; Mendelson and Pillai, 1999; Guimaraes
2.1. Industry dynamics and clockspeed et al., 2002). As demonstrated in these studies, the clockspeed
measure differentiates between the high and low clockspeeds of
The evolution of industries is marked by changing competitive specific industries. However, organisations positioned in dynami-
environments (Audretsch, 1995; Malerba et al., 1999; Wezel, 2005). cally changing competitive environments additionally require a
Organisations subsequently make technological and competitive temporal clockspeed measure of industries that can detect possible
choices along the time path of industry evolution (Gort and Wall, life cycle effects (Fine, 1998), thereby providing a guideline for the
1986; Strebel, 1987; Rice and Galvin, 2008), which determine the implementation of NPD processes (Mu et al., 2009; Brown and
rates of product and process innovation (Abernathy and Utterback, Eisenhardt, 1995; Bessant and Francis, 1997).
1978), the entry and exit of organisations (Klepper and Graddy, 1990; In this paper we aim to develop an additional measure of
Gort and Klepper, 1982; Utterback and Suarez, 1993; Klepper, 1996), industry clockspeed, which is premised on the product technology
and the changing basis of product competition (Christensen, 1997) measure proposed by Fine (1998). Rather than assessing the rate of
 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637 629

new product introductions, our clockspeed measure, namely tech- systemic relationship between sub-industries additionally sug-
nological industry clockspeed, measures the time between succes- gests that changes in the technological output of one sub-industry
sively higher levels of performance in the industry’s product may affect the technological output of other, interdependent sub-
technology evolution, traditionally observed as S-curves (Fisher industries and also the output of the systemic industry as a
and Pry, 1971; Foster, 1986; Ehrnberg, 1995; Sood and Tellis, collective whole.
2005; Nieto et al., 1998; Roussel, 1984; Lee and Nakicenovic, However, the evolution of sub-industries leads to the emer-
1988; Becker and Speltz, 1983). Therefore, shorter spans of time gence of discrepancies in the level of technological development
between successively higher levels of performance indicate faster and the level of utilisation of these technological developments
industry clockspeeds. As a consequence, we take an evolutionary between interdependent sub-industries. These discrepancies are
view (Van de Ven and Poole, 1995), such that we assign clockspeed results of different levels of technological performance that are
measures to assess the changes in activities of the population of delivered by the sub-systems, and they, at the same time, curb the
entities (i.e. a population of organisations or those inhabiting level of performance that is delivered by the systemic product to
niches), where the population consists of the organisations that the end-user. In this state of disequilibrium, the technical sub-
share common form with respect to inputs and outputs and which system that provides the lowest level of technological perfor-
lie within a boundary defined by their markets (Hannan and mance relative to other sub-systems, and subsequently limits the
Freeman, 1977). The clockspeed measure indicates changes at the output performance of the technological system, is termed the
population or industry level, which arises from the search for reverse salient (Hughes, 1983). Its low level of performance
‘‘technical and evolutionary fitness’’ at the organisational level means that the reverse salient strays behind the system’s tech-
(Helfat et al., 2007). nological performance frontier that denotes the highest level of
performance (Dosi, 1982) established by the relatively advanced,
2.2. Systemic industries and the time gap measure of clockspeed or salient, sub-system. Furthermore, the reverse salient inhibits
the performance delivery of the technological system because
Systemic industries produce technological systems (Bonaccorsi technological improvements in other sub-systems cannot materi-
and Giuri, 2000; Barnett, 1990) that are hierarchically structured alise at the holistic system level performance delivery until the
networks of interdependent technical sub-systems (Murmann and reverse salient advances sufficiently to capitalise on its own
Frenken, 2006; Tushman and Murmann, 1998; Clark, 1985). Sys- potential gains (Rosenkopf and Nerkar, 1999). Hence, examples
temic industries therefore comprise sub-industries (Ethiraj and of reverse salients include the motor and capacitor sub-systems of
Puranam, 2004), which are specialised in producing the different the early direct-current electric system that inhibited the efficient
technical sub-systems that, due to their functional relatedness, work distribution of electricity (Hughes, 1983), the gyroscope sub-
together to provide the performance of the systemic industry’s system of the ballistic missile technological system that limited
product to the end-user (Garud and Kumaraswamy, 1995; Ulrich, the accuracy of missiles (MacKenzie, 1987), and the computing
1995). Therefore, the utility of the systemic products is provided by hardware sub-system in the flight simulator technological system
the integration of functionally interdependent sub-systems that are that prevented application capabilities in simulations (Rosenkopf
produced by their own specialised sub-industries, for example and Nerkar, 1999).
engines and airframes in the case of aircraft (Bonaccorsi and Giuri, This state of technological disequilibrium, at a particular point
2000; Schilling, 2000), and microprocessors, graphics processors, in time, reflects the lower level of performance that is utilised by
and software in the case of PCs (Ethiraj and Puranam, 2004; Macher the reverse salient sub-system with respect to the level of
and Mowery, 2004). performance that is provisioned by the interdependent, salient
The networked structure of systemic industries may arise over sub-system. Here, the provisioning and utilisation of performance
time from the vertical specialisation of industries (Macher and is not limited to merely complementary sub-systems of the
Mowery, 2004; Stigler, 1951; Hyvättinen, 2006) and the mod- technological system (Teece, 1986), such as hardware and soft-
ularisation of products and processes (Baldwin and Clark, 1997; ware sub-systems of the PC, but rather stems from the inter-
Langlois and Robertson, 1992; Baldwin and Clark, 2000; Brusoni dependence between sub-systems in general. Subsequently, the
and Prencipe, 2001). Further, systemic industries are depicted in technological advancement of a sub-system at the technological
complex product systems (CoPS) such as nuclear plants, weapon performance frontier provisions, directly or indirectly, a level of
systems, and turnkey industrial equipment (Bonaccorsi et al., performance potential to other sub-systems that, driven by the
1996), which are high cost capital goods that are customised and evolution of the technological system, is utilised by these sub-
therefore manufactured on a project basis rather than mass systems over time. The length of time required for the reverse
produced (Hansen and Rush, 1998; Hobday, 1998; Miller et al., salient to utilise the level of performance that is provisioned by
1995; Prencipe, 2000), and large technical systems (LTS) such as the interdependent, salient sub-system importantly informs of
electric power and railway networks (Hughes, 1983; Mayntz and the rate of development of the reverse salient sub-system in the
Hughes, 1988; Geels, 2006). In this manner, modular systems, technological system context.
CoPS, and LTS are representatives of systemic industries because It therefore follows that the duration of time required for the
they are technological systems that integrate a large number of sub-industry that produces the reverse salient technical sub-
different, interdependent technical sub-systems, which are devel- system (hereafter referred to as the reverse salient sub-industry)
oped by specialised producers. to produce the higher level of technological performance that is
The evolution of these technological systems is the outcome of embedded in the technical sub-system produced by the salient
underlying change mechanisms that are supply-side, in other sub-industry, denotes the rate of change of the reverse salient
words technology push mechanisms, on the one hand and/or sub-industry in its systemic industry context. We subsequently
demand-side, in other words market pull mechanisms, on the use this duration of time for the reverse salient sub-industry to
other (Dosi, 1982; Adner and Levinthal, 2001; Mowery and utilise the level of technological performance provisioned by the
Rosenberg, 1979). Consequently, systemic industries and techno- salient sub-industry to measure the systemic technological indus-
logical systems undergo change subject to technological evolu- try clockspeed. Similar to the technological industry clockspeed,
tions including both intra- and inter-industry innovations taking shorter spans of time between the provisioning and utilisation of
place within the hierarchy of sub-industries (Henderson and technological performances indicate faster clockspeeds in the
Clark, 1990), driven by endogenous and exogenous factors. The systemic industry context. For organisations positioned in
630 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637

dynamically changing systemic industries, the systemic techno- the solid line. Thus, the solid line represents the technological
logical industry clockspeed offers a central measure in consider- performance frontier, and at the same time, the potential level of
ing the time-to-market decisions in the NPD process. technological performance that can be utilised by the reverse
salient across time. The magnitude of systemic technological
industry clockspeed at any point along the time axis is measured
by calculating the horizontal separation of the reverse salient
3. Methodology
from the salient line at that point. This separation reflects the
length of time that is required for the reverse salient sub-system
In this paper we develop two clockspeed measures. Firstly, we
to utilise the level of technological performance that had been
propose a technological industry clockspeed, which evaluates the
provisioned by the salient sub-system at an earlier point in time.
rate of performance accumulation in an industry with respect to a
In this manner, we develop both proposed clockspeed mea-
given technology over time, following the traditional S-curve
sures from the relational perspective of time (Hood, 1948; Zaheer
model of technological performance evolution (Fisher and Pry,
et al., 1999), whereby time derives from the relations of occasions
1971; Foster, 1986; Ehrnberg, 1995; Sood and Tellis, 2005; Nieto
(Rhodes, 1885). Here, we consider, for example, occasion A as the
et al., 1998; Roussel, 1984; Lee and Nakicenovic, 1988; Becker and
salient sub-industry’s launch of a new product with a new, higher
Speltz, 1983; Ayres, 1988). In this manner, we derive the evolu-
level of technological performance, and a later occasion B as the
tion of the technological sub-system by quantitatively measuring
reverse salient sub-industry’s launch of a new product utilising
the change in a particular technological performance indicator
the same level of performance that was launched at occasion A
over time (Sahal, 1981).
by the salient sub-industry. Subsequently, with respect to the
Secondly, we propose a clockspeed measure for the systemic
systemic industry, the duration of time required for the reverse
industry context, which compares the rate of technological
salient sub-industry to produce the technical sub-system with the
performance development of a sub-industry with the rate of
same level of technological performance that is embedded in the
development of other, interdependent sub-industries, namely,
technical sub-system produced by the interdependent, salient
the systemic technological industry clockspeed. To derive this
sub-industry, determines the systemic technological industry
clockspeed measure, we consider the provisioning and utilisation
clockspeed of the reverse salient sub-industry. Hence, Fig. 1
of technological performance in the technical sub-systems pro-
illustrates two technological industry clockspeed measurements
duced by interdependent sub-industries. We begin by considering
for the reverse salient sub-industry, denoted by DtA1 and DtA2, at
the technological evolution of two interdependent sub-systems
time t3 and t4, and two systemic technological industry clock-
within a given technological system. We assume that one sub-
speed measurements for the reverse salient sub-industry in its
system provisions a certain level of technological performance
systemic context, denoted by DtB1 and DtB2, at time t3 and t4,
(i.e. the salient sub-system), while the other sub-system utilises
respectively.
the provisioned technological performance (i.e. the reverse salient
sub-system). To measure the systemic technological industry
3.1. The PC technological system
clockspeed we then evaluate the time gaps of the technology
provisioning and utilisation from the superimposition and com-
In our empirical illustration, we have considered the PC
parison of the technological evolution curves of the two sub-
industry as a systemic industry that produces the PC technologi-
systems on a common set of axes (see Fig. 1).
cal system, whereby the utility of the holistic PC system to the
Our schematic representation of reverse salience in Fig. 1
consumer is provided by the integration of functionally inter-
compares the technological performance evolution of the reverse
dependent sub-systems that are produced by their own specia-
salient, represented by the dashed line, with the technological
lised sub-industries (Ethiraj and Puranam, 2004; Macher and
performance evolution of the salient sub-system, represented by
Mowery, 2004). Further, we have analysed the PC technological
system by focusing specifically on its function as a computer
gaming platform. The PC technological system provided a suitable
empirical setting for our illustration due to its significance as a
platform for computer gaming since the 1990s (Hayes and Dinsey,
1995; Poole, 2004), its systemic nature (Baldwin and Clark, 1997;
Langlois and Robertson, 1992; Schilling, 2000), and its highly
dynamic nature of technological evolution (Eisenhardt and
Tabrizi, 1995; Rosas-Vega and Vokurka, 2000). In particular, the
evolution of the PC technological system has witnessed and
embraced Moore’s law (Moore, 1965), which has established the
pace of change in the PC industry and its sub-industries. There-
fore, actors in the systemic PC industry, like hardware and
software developers, have become accustomed to expect this
type of dynamics in utilising the evolving performance of tech-
nology in NPD processes. All of the above lend us a rich setting
with continuously changing performance disparities, thereby
making the PC technological system informative for illustration
purposes.
Our empirical illustration concentrates on the technical sub-
systems that are central for the delivery of gaming performance in
the PC. In our study we therefore analyse the PC game software as
one technical sub-system, focusing on its co-evolution with two
important and interdependent hardware sub-systems: the CPU
Fig. 1. Schematic of clockspeed measurements from superimposed S-curves of the (central processing unit) and GPU (graphics processing unit),
salient and reverse salient sub-systems. respectively. The interdependence of these sub-systems is apparent,
 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637 631

since PC game software is intended to function together with the hardware requirement stipulated by the PC game software as its
CPU and GPU sub-systems, utilising the performance, which they technological performance indicator. In this manner, we identify a
make available. Moreover, the hardware inherently acts as a plat- PC game to have a higher level of technological performance than
form upon which game software can be developed (Tiwana et al., another when its stipulated minimum hardware requirements are
2010; Kanev and Sugiyama, 1998). As part of this process, many higher than those of the latter. This is because a PC game that
hardware manufacturers provide support for software develop- requires a higher level of hardware performance has more
ment, for example, in the form of software development kits capacity to deliver a higher level of functional performance on
(SDK) following the launch of their new hardware, which game the holistic PC system.
developers anticipate and strategically select and utilise in the While the performance of the CPU can be measured along
development of their new products. This naturally results in a time different dimensions (for example MIPS—millions of instructions
delay for the software to fully utilise the performance of available per second, number of transistors, bus bandwidth, etc.), we have
hardware. This delay is further exacerbated by the developers’ need selected the performance indicator of processing speed, measured
to test the compatibility of the software and hardware, for example, in Hertz (Hz), since this is the parameter that is most important
by testing each new game’s capability to function with individual for PC game developers and publishers for game functionality.
video cards (Ripolles and Chover, 2008). Furthermore, the search for The processor speed indicates the CPU’s speed of operation,
technology-market opportunities, realisation of concepts, and the governing the computational performance of the PC through its
desire to avoid risks of project delays and cost over-runs in a highly interaction with software programs such as PC games. Higher
dynamic technological ecosystem motivates software developers to speeds mean faster data manipulation, which increases the
delay their technology selection as much as possible (MacCormack holistic PC system’s potential to deliver performance such as
et al., 2001). during gaming. Similarly, while the performance of the GPU can
Due to the resulting delay in utilising the technological be measured along different dimensions, we have selected the
performance provisioned by the CPU and GPU sub-systems, the performance indicator of graphics memory, measured in mega-
PC game sub-system inevitably emerges as the reverse salient bytes (MB), since this is the parameter that is most important for
that curbs the holistic PC system performance. Our empirical PC game developers and publishers for game functionality. The
study therefore illustrates the systemic technological industry graphics memory denotes the GPU’s ability to store and make
clockspeed of the PC game software sub-industry, which, in available graphics data to the PC. Larger memory increases the
producing game software, requires some duration of time to PC’s potential to deliver graphics performance during, for exam-
utilise the provisioned level of technological performance ple, game play.
embedded in the hardware products of the CPU and GPU sub- Our empirical illustration comprised two parts. First, we
industries. Here, we make an important distinction between the studied the technological evolutions of the CPU and PC game
co-evolution of technical sub-systems (e.g. that of hardware and sub-systems. We measured the technological performance of the
software sub-systems), and the technical performance levels that CPU using the parameter of processor speed and the PC game
can be observed for these sub-systems. While co-evolution using the parameter of minimum processor speed requirements,
processes influence sub-system performances, in this study we both measured in Hertz. Second, we studied the technological
focus only on the resultant technical performance levels. As a evolutions of the GPU and PC game sub-systems. We used the
result, although we acknowledge the commonly held view that parameter of graphics memory to measure the technological
game software drives the development of computer hardware performance of the GPU, and the parameter of minimum graphics
(e.g. Shirley et al., 2008) we focus merely on the performance memory requirement to measure the technological performance
utilisation delays that inherently exist in software development, of the PC game, both measured in megabytes. The technological
rather than seeking the mechanisms, which may bring this about. evolutions of the CPU and PC game sub-systems, as well as the
GPU and PC game sub-systems, were in turn superimposed onto
3.2. The data respective sets of axes. As a result, we were firstly able to measure
the technological industry clockspeeds of the sub-industries using
For each of the studied technical sub-systems we selected the processor speed and graphics memory technological perfor-
technological performance indicators that would allow the ana- mance indicators, and secondly, to evaluate the measures of the
lysis of performance evolution and the measurement of clock- systemic technological industry clockspeed for the PC game
speeds. For the PC game software sub-system we have considered software sub-industry in its systemic context with respect to
that the software is designed to utilise a certain level of hardware the CPU and GPU sub-industries, respectively.
performance such that the intended game qualities can materi- We collected data on CPU processor speeds from processor
alise on the holistic PC system. For this purpose, the game performance databases found on Intel and AMD (Advanced Micro
software stipulates a set of minimum hardware performance Devices) corporate web sites, the two primary manufacturers of
requirements with which the software will function as designed. CPUs. In the generation of this data set we omitted CPUs that
However, the game software can additionally be designed to were designed for applications other than desktop PCs because
deliver a second, higher level of gaming experience (i.e. recom- our study focuses on games that are played by consumers on
mended hardware performance requirements), for example by personal computers. Additionally, we omitted multi-core CPUs
making additional game features accessible during game play. from the data set because this new technological platform
This higher level of gaming performance can be experienced if the represents a significant shift in the technological paradigm of
specified level of hardware performance is available, which is the CPU, whereby processing speed may no longer be the most
discretely higher than the performance specified by the minimum prevalent indicator of CPU performance, and we deemed this
requirements. Nevertheless, in our study we have focused only on additional dimension unnecessary in our illustration of the clock-
the functional performance of the game software and not on the speed measurement. Similarly, we accessed GPU graphics mem-
additional gaming experience that is designed into the software. ory data from the corporate web sites of NVIDIA and ATI (acquired
Therefore, we use the minimum requirements as a proxy for by AMD on October 2006), the two dominant players in the
the intended level of functional performance that the PC system graphics processor industry. The data on PC game minimum
as a whole should be delivering from the software developers’ processor speed requirements as well as minimum graphics
point of view. Consequently, we have used the minimum memory requirements were, in turn, collected from the web sites
632 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637

of game publishers and game developers, as well as the gaming

community Gamespot.com and a major online PC game vendor,
Amazon.com. The list of PC games used in our empirical analysis
was generated as a result of filtering steps that ensured the
reliability of the data. Firstly, we limited the list of PC games to
only those that had been reviewed and rated by either one
reputable online source, Gamespot.com, or one reputable printed
source, PC Gamer magazine (the reputation of these sources was
identified by experts from the game development industry). The
reputability of these sources was deemed to provide the required
list of all the PC games that influenced the industries involved in a
meaningful way. A game, which had therefore not been reviewed
by either of these sources, was discarded on the basis that it did
not meet the quality standards of the gaming industry at large
and that its inclusion could therefore reduce the reliability of our
data set. Further, we limited the list of PC games to those, which
had been launched in the United States. Fig. 2. The technological performance S-curves for CPU and PC game sub-systems
In our data gathering method we selected appropriate time with respect to processor speed.
scales, as outlined by Zaheer et al. (1999), which fit the objective
of our research. We firstly considered the existence interval, in
other words the length of time over which the phenomenon
occurs, to be one day, because the launch of new CPU and GPU
hardware, and launch of new PC game software is associated with
a particular day of the year. Secondly, in recording the data, we
did not select a specific recording interval because our study has
been designed to gather and analyse data, ex-post. We thirdly
used varying aggregation intervals to report the gathered data,
implementing an event-driven rather than an interval-driven
recording strategy. Finally, we selected an observation interval
that stretches from August 1995 until the end of 2008. We
selected the beginning of the observation interval to correspond
to the launch of the Windows 95 operating system that estab-
lished a new technological platform upon which hardware and PC
game software could be developed and therefore produced a
significant change in the gaming industry (Hayes and Dinsey,
1995). Altogether, 600 CPU and 2900 PC game related data points
on processor speed, as well as 120 GPU and 2050 PC game related Fig. 3. The technological performance S-curves for GPU and PC game sub-systems
data points on graphics memory, were collected and evaluated to with respect to the amount of graphics memory.

reveal the technological evolution curves of the sub-systems and

therefore the technological industry clockspeeds of the individual the lowest level of technological performance and concurrently
sub-industries, and in turn, the systemic technological industry hinders the meta-system level performance delivery.
clockspeed of the PC game software industry in its systemic From the technological development curves shown in
context with the CPU and GPU sub-industries, respectively. Figs. 2 and 3, we measure the technological industry clockspeeds
of the CPU, GPU, and PC game sub-industries from the durations
of time between the launch of products with successively higher
4. Results and discussion levels of technological performance. We present the evolution of
the technological industry clockspeed measures of the CPU sub-
In line with our schematic framework of reverse salience in industry in Fig. 4, and the PC game software sub-industry, using
Fig. 1, we first plot the technological development curves of the the minimum processor speed requirement performance indica-
CPU and PC game sub-systems with respect to the technological tor, in Fig. 5.
performance indicator of processing speed, and the technological Fig. 4 indicates that after an early period of low technological
development curves of the GPU and PC game sub-systems with industry clockspeed in the CPU sub-industry, an increase in the
respect to the technological performance indicator of graphics clockspeed is witnessed from approximately 1999 until 2003, as
memory. We present these plots in Figs. 2 and 3. the time between successive launches in technological perfor-
In the above figures, the development curves depict the succes- mance levels reduces. We posit that the latter period, in which a
sively higher levels of technological performance that are intro- CPU with higher level of technological performance is launched
duced by the CPU, GPU, and PC game sub-industries, over time. every 200 days or less, indicates the high tempo of time pacing in
Figs. 2 and 3 also show that the technological performance devel- the sub-industry (Eisenhardt and Brown, 1998) and the intensify-
opment of the PC game software sub-industry trails the technolo- ing focus of the CPU sub-industry on the processor speed
gical performance development of the CPU and GPU sub-industries, performance. Subsequently, organisations in the CPU sub-indus-
respectively, at all times within the timeframe of analysis. By not try engage in time-based competition in response to the rapidly
fully utilising the level of technical performance that is provisioned changing technological environment to maintain competitiveness
by the hardware sub-systems, the PC game sub-system curbs the (Stalk, 1988), thereby continuing the period of frequent launch
performance delivery of the PC system. We subsequently ascribe of products with higher performance. However, recent years,
the PC game sub-system as a reverse salient, according to the following the period of intense time-based competition, show a
definition of a reverse salient, which is the sub-system that delivers decrease in the technological industry clockspeed. Interestingly,
 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637 633

Fig. 6. The technological industry clockspeed of the GPU sub-industry measured

Fig. 4. The technological industry clockspeed of the CPU sub-industry measured as the number of days between product launches with successively higher
as the number of days between product launches with successively higher graphics memory.
processor speed.

Fig. 7. The technological industry clockspeed of the PC game sub-industry

Fig. 5. The technological industry clockspeed of the PC game sub-industry measured as the number of days between product launches with successively
measured as the number of days between product launches with successively higher minimum graphics memory requirement.
higher minimum processor speed requirement.

life cycles, a phase, which was also witnessed in the technological

this period of time coincides with the technical limitations that industry clockspeed evolutions of the CPU and PC game software sub-
materialised in CPU development, leading to diminishing gains in industries using minimum processor speed requirement. We there-
processor speeds, which may have led to the increasing duration fore suspect a gradual intensification of time-based competition
of time between successively higher performing CPUs. However, based on graphics performance among organisations in the GPU
these technical reverse salients in the CPU were circumvented by and PC game software sub-industries, leading to increased clock-
the CPU sub-industry in 2005 through the introduction of a new speeds, as the sub-industries progress through their life cycles
technological paradigm reflected in the multi-core CPU, while the over time.
processor speed performance of the single core paradigm stag- Our analyses, overall, show that the technological industry clock-
nated at the end of 2004, as illustrated in Fig. 2. speeds of the sub-industries slow down to as low as two to three
The technological industry clockspeed of the PC game software years in the rate of launch of new, higher levels of technological
sub-industry using the minimum processor speed requirement performance. This outcome is commensurate with the findings of
performance indicator, shown in Fig. 5, similarly displays a period Mendelson and Pillai (1999), who have also shown the rhythm of
of increasing clockspeed from the beginning of the analysed development activity in the PC and minicomputer industries to
timeframe until approximately the end of 2004, after which a stretch to as much as three years. In our study, we additionally
brief period of high variation in the clockspeed is witnessed. analysed whether the technological industry clockspeed differenti-
We in turn present the evolution of the technological industry ates the sub-industries from one another in their systemic contexts.
clockspeed measures for the GPU sub-industry in Fig. 6, and for As a result, the mean for the technological industry clockspeed
the PC game software sub-industry, using the minimum graphics measure for the CPU sub-industry was 263 days (std. dev. 245), and
memory requirement performance indicator, in Fig. 7. for the PC game software sub-industry, using minimum processor
Both Figs. 6 and 7 show significant variation in the GPU and PC speed requirement as performance indicator, 372 days (std. dev.
game software sub-industry technological industry clockspeeds 267). Further, the mean for the technological industry clockspeed
across the timeframe of analysis. Reflecting on the evolution of these measure for the GPU sub-industry resulted in 546 days (std. dev.
sub-industries, represented by their technological performance evo- 248) and for the PC games software sub-industry, using minimum
lution curves plotted in Fig. 3, we believe that the sub-industries are graphics memory requirement as performance indicator, in 541 days
progressing through the early phases of their respective technology (std. dev. 247). At face value, we therefore observe that the dynamics
634 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637

of the technological industry clockspeed measures for the CPU and of delivered in the holistic PC system, in other words the greater
PC game software sub-industries, using minimum processor speed utility the consumer receives in playing computer games on the
requirement as performance indicator, are comparable to one PC, subsequent to the interdependence of the CPU and PC game
another, and similarly the GPU and PC game software sub-industries, software sub-systems, requires longer lengths of time to materi-
using minimum graphics memory requirement as performance alise as the systemic industry evolves. Interestingly, this outcome
indicator, exhibit dynamics of the technological industry clockspeed contrasts with the evolution of the sub-industry’s own technolo-
measure that are akin to each other. Furthermore, the observed gical industry clockspeed measure presented earlier in Fig. 5,
clockspeed measures show that the variances of the two sub- which showed increasing clockspeed and intensifying time-based
industry pairs are different overall (Levene’s test, po0.0001), and competition among organisations within the sub-industry. Hence,
further, that the means of the pairs are also different (two-tailed t- from an organisational standpoint, the combination of these
test, po0.000). These results support the systemic relationship results suggests that while PC game software developing organi-
between the sub-industry pairs, whereby the technological perfor- sations do see value in engaging in time-based competition to
mance that is provisioned by one sub-industry influences the remain competitive in the context of their own sub-industry, they
technological performance that is utilised by the other, interdepen- do not necessarily derive competitive advantage by engaging in
dent sub-industry. time-based competition in the systemic context. It therefore
Finally, to derive the systemic technological industry clock- appears that PC game software developers engage in time-based
speed of PC game software, we consider the level of technological competition that centres on satisfying needs other than those of
performance that is provisioned by the CPU and GPU sub- strictly utilising the level of processing performance that is
industries, and evaluate the time delay for the PC game software provisioned by the CPU sub-industry.
sub-industry’s utilisation of the processor and graphics perfor- In contrast, the longitudinal analysis of the PC game software
mance, respectively. Figs. 8 and 9 display the evolutions of the sub-industry’s systemic technological industry clockspeed with
systemic technological industry clockspeeds across the timeframe respect to the GPU sub-industry development shows three
of our analysis. sequential eras marked by increasing, constant, and finally
The systemic technological industry clockspeed of the PC game decreasing clockspeed. This evolution of the systemic technolo-
software sub-industry with respect to the CPU sub-industry gical industry clockspeed indicates the diminishing competitive
development shows a continuously increasing time delay of advantage that PC game software developers derive from time-
technological performance utilisation. At first glance, this result based competition centring on graphics performance, in the
suggests that the higher levels of computer gaming performance systemic context. Interestingly, the changes in the systemic
technological industry clockspeed reflect the changing basis of
competition that is expected to take place throughout the life
cycle of an industry (Christensen, 1997). In this manner, we
interpret that in the first era of increasing clockspeed, the basis
of competition in the PC game software sub-industry has centred
on utilising, with reducing delay, the level of graphics perfor-
mance provisioned by the GPU sub-industry. This era therefore
emphasises the importance that organisations assign to reacting,
in a timely manner, to changes in the systemic industry that are
manifest in the rate of change in interdependent sub-industries.
Nevertheless, we posit that the following eras of constant and
decreasing systemic technological industry clockspeed denote a
shift away from the basis of competition that utilises the provi-
sioned graphics memory capacity by the GPU sub-industry. The
decrease in the systemic technological industry clockspeed may,
for example, reflect the shift in the basis of competition to the
timely utilisation of technological performance provisioned by
Fig. 8. The systemic technological industry clockspeed of the PC game sub- other sub-industries.
industry with respect to the CPU sub-industry.

5. Conclusion and implications

In this paper we have contributed to the study of systemic

industries that produce technological systems, by developing
means through which the rate of change of sub-industries can
be measured. To this end, we have extended the current under-
standing of industry evolution by developing a temporal measure
of the sub-industry’s own technological industry clockspeed, and
further, by developing a temporal measure of the sub-industry’s
systemic technological industry clockspeed. First, we have
developed the measure of the sub-industry’s own technological
industry clockspeed by evaluating the durations of time between
the launch of products with successively higher technological
performance. Second, building on the theory of reverse salience,
we have developed the systemic technological industry clockspeed
measure by evaluating the length of time that is required for the
Fig. 9. The systemic technological industry clockspeed of the PC game sub- reverse salient sub-industry (i.e. the sub-industry producing the
industry with respect to the GPU sub-industry. sub-system with the lowest level of technological performance) to
 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637 635

utilise the level of technological performance provisioned by the away from utilising the technological performance provisioned by
advanced, salient sub-industry. In turn, we have demonstrated the the salient sub-industry, toward following the technological
use of these temporal clockspeed measures in an illustrative, long- requirements of consumers. Nevertheless, the limitation of our
itudinal analysis of the systemic PC industry that is commonly paper necessitates further research that considers consumer
perceived as being a technology-intensive industry (Eisenhardt and demand and its evolution in analysing such transition between
Tabrizi, 1995). In our illustration we have focused specifically on the competitive domains of the reverse salient sub-industry.
sub-industries that produce the technological sub-systems, which Fourthly, the observed shift may result from the change in the
are central for computer gaming in the holistic PC system. basis of competition from utilising the provisioned performance
The results of our empirical illustration show that the techno- of one sub-industry, to utilising the provisioned performance of
logical industry clockspeeds of the studied CPU, GPU, and PC game another. In our illustrative study, such a shift may, for example,
software sub-industries evolve across time, which verifies Fine’s take place when the PC game software sub-industry reorients its
(1998) notion of some deviation in the industry’s rate of change NPD focus from utilising the processor speed performance provi-
over time. In these observations, we suggest that increasing sioned by the CPU sub-industry, to utilising the graphics memory
technological industry clockspeeds signify intensifying time-based performance provisioned by the GPU sub-industry. The temporal
competition among organisations in the respective sub-industries limitation imposed by our observation interval does not, however,
that focus on quicker development of new products with higher allow us to determine the point of transition between these
technological performance. Additionally, the observed patterns of competitive domains, thereby requiring supplementary research.
technological industry clockspeed evolution suggest that there Furthermore, future studies can additionally extend the scope of
may be different eras of change that emerge across the evolution analysis, limited in this paper for illustrative purposes, to study
of these sub-industries. Moreover, these technological industry the evolutions of other interdependent sub-industries linked to
clockspeed eras, in a similar manner to the number of organisa- computer gaming on the PC system. This extension will allow the
tions, product sales, and innovations (Abernathy and Utterback, study of a sequence of competitive domain transitions for the PC
1978; Gort and Klepper, 1982; McGahan, 2000) may be linked to game software sub-industry in its systemic industry context, as
the different phases of industry evolution. The limited observation the sub-industry successively shifts its competitive focus in
interval of our empirical study, however, prevents us from compar- utilising technological performance that is provisioned by differ-
ing the different life cycle phases of the studied sub-industries and ent sub-industries.
the different technological industry clockspeed eras. The proposed measures of industry clockspeed can be integral
Our empirical results further suggest that in the systemic parts of the decision making process concerning the designed
industry context, the systemic technological industry clockspeed performance levels of new products and the timing of their
of the reverse salient sub-industry may progress through three introductions, and hence offer both managerial and theoretical
sequential eras marked by increasing, constant, and finally implications. The systemic technological industry clockspeed
decreasing clockspeed. We suspect that this pattern of change measure firstly pinpoints the rate of technological performance
indicates the diminishing competitive advantage that organisa- gap closure between different sub-industries and therefore the
tions in the sub-industry derive from time-based competition development of technological performance that can be delivered
centring on the utilisation of the level of technological perfor- by the systemic product, over time. This is valuable information
mance that is provisioned by the salient sub-industry. Moreover, for organisations, which are embedded in systemic industries that
organisations within the reverse salient sub-industry may over produce technological systems, such as CoPS, LTS, and modular
time engage in time-based competition that centres on satisfying systems, because these organisations derive competitive advan-
other needs within the systemic context, thereby reflecting the tage by understanding the value of performance improvements in
shifting basis of competition that is expected to take place during the sub-system technology that is central to their own sub-
the development of the sub-industry (Christensen, 1997). industry, as well as through the management of the interface
We believe that this shift may materialise in the systemic between other, interdependent sub-industries that produce the
context for several reasons. Firstly, the NPD efforts of organisa- interdependent technological sub-systems (Adner and Kapoor,
tions in the reverse salient sub-industry may be delayed in 2010; Ethiraj and Puranam, 2004). As a result, firms are able to
response to the technological progress in interdependent sub- formulate strategies by not only monitoring the rate of growth
industries due to the increased learning demands imposed by and the level of R&D activity in their own industries (Griffiths and
increasing complexities. In our empirical illustration, despite their Webster, 2010), but also by observing the changes taking place in
capacity to anticipate future demand for hardware performance, their immediate environments (Subramanian et al., 2011).
game developers may not be able to fully utilise the potential Furthermore, the systemic technological industry clockspeed
performance levels due to numerous other game attributes, measure employed in longitudinal studies reveals the shifts in
which require attention. These include artistic, graphical, char- technological performance utilisation of sub-industries and there-
acter-role, and plot related considerations, which are interlinked fore informs of change in the technological development focus of
to one another as well as to the hardware performance through the systemic industry. Organisations in systemic industries can
which they materialise. benefit from this external information by reducing uncertainties
Secondly, the decreasing clockspeed in the systemic industry and risks associated with their R&D activities (Wang et al., 2010),
context may be attributed to the growing discrepancy between and at the same time deploying appropriate competitive strate-
the NPD processes in the salient and reverse salient sub-indus- gies and firm structures with respect to NPD processes (de Visser
tries. In this manner, the results of our empirical illustration may et al., 2010). In this manner, organisations may set their own pace
be caused by the uncoupling of the hardware manufacturers’ and of change to match the environmental pace of change (Eisenhardt
game software developers’ NPD activities, which may inform of and Martin, 2000), in other words concurrently utilise the
the degree of interdependence between these sub-industries. information on the technological industry as well as systemic
Thirdly, technological developments in the salient sub-indus- technological industry clockspeeds. From an organisational
try (i.e. hardware industries) may substantially overshoot the standpoint, understanding the changing basis of competition
performance demands of consumers (Christensen, 1997). As a (Christensen, 1997) within the specific sub-industry, as well as
result, organisations in the reverse salient sub-industry (i.e. soft- within the systemic industry context, can assist organisations in
ware developers) may shift their new product development focus coping more effectively with their dynamic and competitive
636 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637

environment, thereby positively influencing performance and Cohen, M.A., Eliashberg, J., Ho, T., 1996. New product development: the perfor-
survival. mance and time-to-market tradeoff. Management Science 42 (2), 173–186.
Datar, S., Jordan, C., Kekre, S., Rajiv, S., Srinivasan, K., 1997. New product development
structures and time-to-market. Management Science 43 (4), 452–464.
de Visser, M., de Weerd-Nederhof, P., Faems, D., Song, M., van Looy, B., Visscher, K.,
2010. Structural ambidexterity in NPD processes: a firm-level assessment of
Acknowledgements the impact of differentiated structures on innovation performance. Technova-
tion 30, 291–299.
We are grateful for the valuable comments and suggestions of Dosi, G., 1982. Technological paradigms and technological trajectories. Research
Policy 11, 147–162.
two anonymous referees, who have helped us further develop Ehrnberg, E., 1995. On the definition and measurement of technological disconti-
this paper. nuities. Technovation 15 (7), 437–452.
Eisenhardt, K.M., Brown, S.L., 1998. Time pacing: competing in markets that won’t
stand still. Harvard Business Review March–April, 59–69.
Eisenhardt, K.M., Tabrizi, B.N., 1995. Accelerating adaptive processes: product
References innovation in the global computer industry. Administrative Science Quarterly
40 (1), 84–110.
Abernathy, W.J., Utterback, J.M., 1978. Patterns of industrial innovation. Technol- Eisenhardt, K.M., Martin, J.A., 2000. Dynamic capabilities: what are they? Strategic
ogy Review 80 (7), 40–47. Management Journal 21, 1105–1121.
Adner, R., 2006. Match your innovation strategy to your innovation ecosystem. Ethiraj, S., Puranam, P., 2004. The distribution of R&D effort in systemic industries:
Harvard Business Review April, 98–107. implications for competitive advantage. In: Baum, J.A.C., McGahan, A.M. (Eds.),
Adner, R., Kapoor, R., 2010. Value creation in innovation ecosystems: how the Business Strategy Over the Industry Life Cycle. Elsevier Ltd., Oxford,
structure of technological interdependence affects firm performance in new pp. 225–253.
technology generations. Strategic Management Journal 31, 306–333. Fine, C.H., 1998. Clockspeed: Winning Industry Control in the Age of Temporary
Adner, R., Levinthal, D., 2001. Demand heterogeneity and technology evolution: Advantage. Perseus Books, Reading (MA).
implications for product and process innovation. Management Science 47 (5), Fisher, J.C., Pry, R.H., 1971. A simple substitution model of technological change.
611–628. Technological Forecasting and Social Change 3 (1), 75–88.
Agarwal, R., Gort, M., 1996. The evolution of markets and entry, exit and survival Foster, R.N., 1986. Innovation: The Attacker’s Advantage. Summit Books, New
of firms. The Review of Economics and Statistics 78 (3), 489–498. York, USA.
Agarwal, R., 1998. Evolutionary trends of industry variables. Research Policy 16, Garud, R., Kumaraswamy, A., 1995. Technological and organizational designs for
511–525. realizing economies of substitution. Strategic Management Journal 16,
Agarwal, R., Sarkar, M.B., Echambadi, R., 2002. The conditioning effect of time on 93–101.
firm survival: an industry life cycle approach. The Academy of Management Geels, F.W., 2006. Co-evolutionary and multi-level dynamics in transition: the
Journal 45 (5), 971–994. transformation of aviation systems and the shift from propeller to turbojet
Audretsch, D.B., 1995. Innovation and Industry Evolution. The MIT Press, Cam- (1930–1970). Technovation 26 (9), 999–1016.
bridge, MA. Gooroochurn, N., Hanley, A., 2007. A tale of two literatures: transaction costs and
Ayres, R.U., 1988. Barriers and breakthroughs: an ‘‘expanding frontiers’’ model of property rights in innovation outsourcing. Research Policy 36, 1483–1495.
the technology-industry life cycle. Technovation 7 (2), 87–115. Gort, M., Wall, R.A., 1986. The evolution of technologies and investment in
Baldwin, C.Y., Clark, K.B., 2000. Design Rules: The Power of Modularity. MIT Press, innovation. The Economic Journal 96 (383), 741–757.
Cambridge, MA. Gort, M., Klepper, S., 1982. Time paths in the diffusion of product innovations. The
Baldwin, C.Y., Clark, K.B., 1997. Managing in an age of modularity. Harvard Economic Journal 92 (367), 630–653.
Business Review 75 (5), 84–93. Griffiths, W., Webster, E., 2010. What governs firm-level R&D: internal or external
Barnett, W.P., 1990. The organizational ecology of a technological system. factors? Technovation 30, 471–481.
Administrative Science Quarterly 35 (1), 31–60. Guimaraes, T., Cook, D., Natarajan, N., 2002. Exploring the importance of business
Baum, J.A.C., Singh, J.V., 1994. Organizational niches and the dynamics of clockspeed as a moderator for determinants of supplier network performance.
organizational mortality. The American Journal of Sociology 100 (2), 346–380. Decision Sciences 33 (4), 629–644.
Bayus, B.L., 1997. Speed-to-market and new product performance trade-offs. Gupta, A.K., Sounder, W.E., 1998. Key drivers of reduced cycle time. Research
Journal of Product Innovation Management 14, 485–497. Technology Management July–August, 38–43.
Becker, R.H., Speltz, L.M., 1983. Putting the S-curve to work. Research Management Hannan, M.T., Freeman, J., 1977. The population ecology of organizations. The
September–October, 31–33. American Journal of Sociology 82 (5), 929–964.
Benner, M.J., Tushman, M.L., 2003. Exploitation, exploration, and process manage- Hansen, K.L., Rush, H., 1998. Hotspots in complex product systems: emerging
ment: The productivity dilemma revisited. The Academy of Management issues in innovation management. Technovation 18 (8/9), 555–561.
Review 28 (2), 238–256. Hayes, M., Dinsey, S., 1995. Games War. Bowerdean Publishing Company Ltd.,
Bessant, J., Caffyn, S., Gilbert, J., 1996. Learning to manage Innovation. Technology London.
Analysis and Strategic Management 8 (1), 59–70. Helfat, C.E., Finkelstein, S., Mitchell, W., Peteraf, M.A., Singh, H., Teece, D.J., Winter,
Bessant, J., Francis, D., 1997. Implementing the new product development process. S.G., 2007. Dynamic Capabilities. Understanding Strategic Change in Organiza-
Technovation 17, 189–197. tions. Blackwell Publishing.
Bonaccorsi, A., Giuri, P., 2000. When shakeout doesn’t occur: the evolution of the Henderson, R.M., Clark, K.B., 1990. Architectural innovation: the reconfiguration of
turboprop engine industry. Research Policy 29, 847–870. existing product technologies and the failure of established firms. Adminis-
Bonaccorsi, A., Pammolli, F., Tani, S., 1996. The changing boundaries of system trative Science Quarterly 35 (1), 9–30.
companies. International Business Review 5 (6), 539–560. Hobday, M., 1998. Product complexity, innovation and industrial organisation.
Bourgeois III, L.J., Eisenhardt, K.M., 1988. Strategic decision processes in high Research Policy 26, 689–710.
velocity environments: four cases in the microcomputer industry. Manage- Hood, W.C., 1948. Some aspects of the treatment of time in economic theory. The
ment Science 34 (7), 816–835. Canadian Journal of Economics and Political Science 14 (4), 453–468.
Brockhoff, K.K., Ernst, H., Hundhausen, E., 1999. Gains and pains from Hughes, T.P., 1983. Networks of Power: Electrification in Western Society, 1880–
licensing—patent-portfolios as strategic weapons in the cardiac rhythm 1930. The John Hopkins University Press, USA.
management industry. Technovation 19, 605–614. Hyvättinen, H., 2006. Interface standards and creating innovation markets—
Brown, S.L., Eisenhardt, K.M., 1997. The art of continuous change: linking implications on SMEs in a technology programme. Technovation 26, 262–273.
complexity theory and time-based evolution in relentlessly shifting organiza- Jovanovic, B., MacDonald, G.M., 1994. The life cycle of a competitive industry. The
tions. Administrative Science Quarterly 42, 1–34. Journal of Political Economy 102 (2), 322–347.
Brown, S.L., Eisenhardt, K.M., 1995. Product development: past research, present Jovanovic, B., 1982. Selection and the evolution of industry. Econometrica 50 (3),
findings, and future directions. Academy of Management Review 20 (2), 649–670.
343–378. Kanev, K., Sugiyama, T., 1998. Design and simulation of interactive 3D computer
Brusoni, S., Prencipe, A., 2001. Unpacking the black box of modularity: technol- games. Computer & Graphics 22 (2-3), 281–300.
ogies, products and organizations. Industrial and Corporate Change 10 (1), Klepper, S., Graddy, E., 1990. The evolution of new industries and the determinants
179–205. of market structure. The RAND Journal of Economics 21 (1), 27–44.
Carroll, G.R., Hannan, M.T., 1989. Density delay in the evolution of organizational Klepper, S., 1996. Entry, exit, growth, and innovation over the product life cycle.
populations: a model and five empirical tests. Administrative Science The Amercian Economic Review 86 (3), 562–583.
Quarterly 34 (3), 411–430. Langlois, R.N., Robertson, P.L., 1992. Networks and innovation in a modular
Chesbrough, H., 2003. Environmental influences upon firm entry into new sub- system: lessons from the microcomputer and stereo component industries.
markets evidence from the worldwide hard disk drive industry conditionally. Research Policy 21, 297–313.
Research Policy 32, 659–678. Lee, T.H., Nakicenovic, N., 1988. Technology life-cycles and business decisions.
Christensen, C.M., 1997. The Innovator’s Dilemma: When New Technologies Cause International Journal of Technology Management 3, 411–426.
Great Firms to Fail. Harvard Business School Press, Boston, Massachusetts. MacCormack, A., Verganti, R., Iansiti, M., 2001. Developing products on internet
Clark, K.B., 1985. The interaction of design hierarchies and market concepts in time: the anatomy of a flexible development process. Management Science 47
technological evolution. Research Policy 14 (5), 235–251. (1), 133–150.
 O. Dedehayir, S.J. Mäkinen / Technovation 31 (2011) 627–637 637

Macher, J.T., Mowery, D.C., 2004. Vertical specialization and industry structure Suarez, F.F., Utterback, J.M., 1995. Dominant designs and the survival of firms.
in high technology industries. In: Baum, J.A.C., McGahan, A.M. (Eds.), Business Strategic Management Journal 16 (6), 415–430.
Strategy Over the Industry Life Cycle. Elsevier Ltd., Oxford, pp. 317–355. Subramanian, A.M., Kah-Hin, C., Shifeng, M., 2011. Capability reconfiguration of
MacKenzie, D., 1987. Missile accuracy: a case study in the social processes of incumbent firms: nintendo in the video game industry. Technovation 31, 228–239.
technological change. In: Bijker, W.E., Hughes, T.P., Pinch, T.J. (Eds.), The Social Teece, D.J., Pisano, G., Shuen, A., 1997. Dynamic capabilities and strategic manage-
Construction of Technological Systems. The MIT Press, USA, pp. 195–222. ment. Strategic Management Journal 18 (7), 509–533.
Malerba, F., Nelson, R.R., Orsenigo, L., Winter, S.G., 1999. ’History friendly’ models of Teece, D.J., 1986. Profiting from technological innovation: implications for integra-
industry evolution: the computer industry. Industrial and Corporate Change 8 (1), tion, collaboration, licensing and public policy. Research Policy 15 (6), 285–305.
3–40. Tiwana, A., Konsynski, B., Bush, A.A., 2010. Platform evolution: coevolution of
Mayntz, R., Hughes, T.P., 1988. The Development of Large Technical Systems. platform architecture, governance, and environmental dynamics. Information
Campus Verlag, Frankfurt am Main. Systems Research 21 (4), 675–687.
McGahan, A.M., 2004. How industries change. Harvard Business Review 82 (10), 87–94. Tushman, M.L., Murmann, J.P., 1998. Dominant designs, technology cycles, and
McGahan, A.M., 2000. How industries evolve. Business Strategy Review 11 (3), organizational outcomes. Research in Organizational Behavior 20, 231–266.
1–16. Ulrich, K., 1995. The role of product architecture in the manufacturing firm.
Mendelson, H., Pillai, R.R., 1999. Industry clockspeed: measurement and operational Research Policy 24, 419–440.
implications. Manufacturing & Service Operations Management 1 (1), 1–20. Utterback, J.M., Suarez, F.F., 1993. Innovation, competition, and industry structure.
Meyer, M.H., Utterback, J.M., 1995. Product development cycle time and commer- Research Policy 22 (1), 1–21.
cial success. IEEE Transactions on Engineering Management 42 (4), 297–304. Van de Ven, A., Poole, M.S., 1995. Explaining development and change in
Miller, R., Hobday, M., Leroux-Demers, T., Olleros, X., 1995. Innovation in complex organizations. Academy of Management Review 20 (3), 510–540.
system industries: the case of flight simulators. Industrial and Corporate Change Wang, J., Lin, W., Huang, Y., 2010. A performance-oriented risk management
4, 363–400. framework for innovative R&D projects. Technovation 30, 601–611.
Moore, G.E., 1965. Cramming more components onto integrated circuits. Electro- Wezel, F.C., 2005. Location dependence and industry evolution: founding rates in
nics 38, 8. the United Kingdom motorcycle industry, 1895–1993. Organization Science 26
Mowery, D.C., Rosenberg, N., 1979. The influence of market demand on innova- (5), 729–754.
tion: a critical review of some recent empirical studies. Research Policy 8, Williams, J.R., 1992. How sustainable is your competitive advantage? California
103–153. Management Review 34 (3), 29–51.
Mu, J., Peng, G., MacLachlan, D.L., 2009. Effect of risk management strategy on NPD Zaheer, S., Albert, S., Zaheer, A., 1999. Time scales and organizational theory.
performance. Technovation 29, 170–180. Academy of Management Review 24 (4), 725–741.
Murmann, J.P., Frenken, K., 2006. Toward a systematic framework for research on
dominant designs, technological innovations, and industrial change. Research
Policy 35, 925–952.
Nadkarni, S., Narayanan, V.K., 2007. Strategic schemas, strategic flexibility, and
firm performance: the moderating role of industry clockspeed. Strategic
Management Journal 28, 243–270. Ozgur Dedehayir, Ph.D., is a
Nieto, M., Lopéz, F., Cruz, F., 1998. Performance analysis of technology using the S Research Fellow at the Tampere
curve model: the case of digital signal processing (DSP) technologies. Tech-
novation 18 (6/7), 439–457.
University of Technology (TUT), Fin-
Persson, M., Åhlström, P., 2006. Managerial issues in modularising complex land. He is a member of CITER
products. Technovation 26, 1201–1209. (Center for Innovation and Technol-
Poole, S., 2004. Trigger Happy: Video Games and the Entertainment Revolution. ogy Research), conducting his
Arcade Publishing, New York.
Porter, M., 1985. Competitive Advantage: Creating and Sustaining Superior
research on the evolution of the PC
Performance. Free Press, New York. technological system, focusing on
Prencipe, A., 2000. Breadth and depth of technological capabilities in CoPS: the processor and graphics related tech-
case of the aircraft engine control system. Research Policy 29, 895–911. nical performance in computer
Rhodes, E.H., 1885. The scientific conception of the measurement of time. Mind 10
(39), 347–362.
gaming. The central theme in his research is the concept
Rice, J., Galvin, P., 2008. Alliance patterns during industry life cycle emergence: the of reverse salience, which depicts the disequilibrium
case of Ericsson and Nokia. Technovation 26 (3), 384–395. between different sub-systems in a continuously evolving
Ripolles, O., Chover, M., 2008. Optimizing the management of continuous level of socio-technical system. He received a B.Sc. in Mechanical
detail on GPU. Computers & Graphics 32, 307–319.
Romanelli, E., Tushman, M.L., 1994. Organizational transformation as punctuated
Engineering from the University of Melbourne, Australia,
equilibrium: an empirical test. The Academy of Management Journal 37 (5), and an M.Sc. in Industrial Management from TUT, Finland.
1141–1166.
Rosas-Vega, R., Vokurka, R.J., 2000. New product introduction delays in the computer
industry. Industrial Management and Data Systems 100 (4), 157–163.
Rosenberg, N., 1976. Perspectives on Technology. Cambridge University Press,
Cambridge. Saku J. Mäkinen, Ph.D., is Professor
Rosenkopf, L., Nerkar, A., 1999. On the complexity of technological evolution: of Technology and Innovation Man-
exploring coevolution within and across hierarchical levels in optical disc
agement and the Head of the
technology. In: Baum, J.A.C., McKelvey, B. (Eds.), Variations in Organization
Science. Sage Publications, Inc., USA. Department of Industrial Manage-
Roussel, P.A., 1984. Technological maturity proves a valid and important concept. ment at Tampere University of Tech-
Research Management Januray—February, 29–34. nology (TUT), Finland. Prof. Mäkinen
Sahal, D., 1981. Patterns of Technological Evolution. Addison-Wesley, London.
was formerly with the Department
Schilling, M.A., 2000. Toward a general modular systems theorys application to
interfirm product modularity. The Academy of Management Review 25 (2), of Marketing, Faculty of Business
312–334. Administration, National University
Shirley, P., Sung, K., Brunvand, E., Davis, A., Parker, S., Boulos, S., 2008. Fast ray of Singapore, NUS. He received an
tracing and the potential effects on graphics and gaming courses. Computers &
Graphics 32, 260–267.
M.Sc. in Electrical Engineering and a Ph.D. in Technology
Sood, A., Tellis, G.J., 2005. Technological evolution and radical innovation. Journal Strategy, both from TUT, Finland. He has widely published
of Marketing 69 (July), 152–168. in international forums like Technological Forecasting and
Souza, G.C., Bayus, B.L., Wagner, H.M., 2004. New-product strategy and industry Social Change, and IEEE Transactions on Engineering Man-
clockspeed. Management Science 50 (4), 537–549.
Stalk Jr, G., 1988. Time—the next source of competitive advantage. Harvard
agement in the areas of his research interests, which
Business Review 66 (4), 41–51. include technology and innovation management, strategy,
Stigler, G.J., 1951. The division of labor is limited by the extent of the market. The and technology and industry co-evolution.
Journal of Political Economy 59 (3), 185–193.
Strebel, P., 1987. Organizing for innovation over an industry cycle. Strategic
Management Journal 8 (2), 117–124.
Suarez, F.F., Lanzolla, G., 2005. The half-truth of first-mover advantage. Harvard
Business Review 83 (4), 121–127.
TFS-18292; No of Pages 13
Technological Forecasting & Social Change xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Technological Forecasting & Social Change

The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing

Joel West a,⁎, George Kuk b
a
KGI — Keck Graduate Institute, United States
b
Nottingham Business School, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: Selective openness allows a firm to sell a systemic innovation that combines both open and proprietary technol-
Received 30 January 2014 ogies. Such firm strategies are now common for open source software and other information goods. However,
Received in revised form 15 May 2015 they pose conceptual and practical uncertainties for hardware-focused companies, particularly as research on
Accepted 28 July 2015
open hardware has emphasized community rather than firm success. Here we study firm openness in 3D
Available online xxxx
printing, with a case study of how MakerBot Industries leveraged external communities and selective openness
Keywords:
become the consumer market leader. After reviewing the literature on systemic innovation and selective openness,
3D printing we document the proprietary strategies of a dozen startup companies during the first two decades of the 3D print-
Online communities ing industry. We contrast this to the open hardware, software and content strategy that MakerBot's founders used
Open innovation to enter and grow the consumer market from 2009 onward. We show how MakerBot shifted to a selectively open,
Open design systemic innovation strategy that complemented proprietary hardware and software with open user-generated
Complementary assets content from its Thingiverse online community. From this, we suggest the inherent complementarity of selective
openness strategies between open and proprietary components, and conclude with predictions as to when and
how a startup or incumbent firm will combine open and proprietary elements.
© 2015 Published by Elsevier Inc.

1. Introduction communities has suggested that their economics and organization are
quite different (e.g., Raasch et al., 2009), and thus a goal of this paper
For decades, the incentives to create and diffuse innovations have is to extend our understanding of these communities to explain how
been associated with the strength of appropriability regimes and the firm strategies leverage these communities.
ability of a private inventor to appropriate private returns from their Here we examine how openness influenced the success of 3D
economic investment. Traditionally such incentives are subject to printers, an industrial technology from the 1980s that faced slow adop-
strong IP protection mechanisms such as patents, copyright and trade tion until the RepRap open hardware project stimulated the subsequent
secrets (Nordhaus, 1969; Teece, 1986). entry of consumer-focused startup producers. We focus on the case of
Over the past decade, some previously proprietary incumbent firms MakerBot Industries, the leading maker of consumer 3D printers, with
have experimented with opening parts of their complex offerings to win its unique value creation strategy. The firm was founded based on an
cooperation from adopters, complementors and even rivals (West, open source hardware and software strategy — complemented by its
2003; Henkel, 2006; Shah, 2006; West and Gallagher, 2006; West and Thingiverse open content community — and kept its user-generated con-
O'Mahony, 2008; Spaeth et al., 2010; Henkel et al., 2014). The availabil- tent open even after it switched to proprietary hardware and software
ity of shared IP from an open source community has also enabled the (See Fig. 1).
formation and entry of new firms that build upon these collective efforts From this, we extend the research on selective openness from soft-
(Gruber and Henkel, 2006; Dahlander, 2007). ware to hardware business models. Specifically, we propose an exten-
Prior studies have emphasized openness in software, endowed sion of the open source model to suggest a more general pattern of
with the characteristics of information good. Therefore, we would ex- what we term the complementarity of openness. In such cases, openness
pect the findings in software to representative of the larger class of increases the value of the non-open part of a firm's value proposition,
information goods, which are nonrivalrous in consumption and have lit- and thus a firm can only be open if it has something that is closed
tle or no marginal cost, making them nearly costless to disseminate (cf. which allows it to generate revenues and profits to support the busi-
Varian, 1998; Kogut and Metiu, 2001; Cusumano, 2004; Von Hippel and ness. We also highlight the unique selective openness constraints of
von Krogh, 2003). However, limited research on open hardware new firms, that leverage open technologies to create value and enter
markets while seeking proprietary imitation barriers to protect value
capture. We offer specific predictions as to how resource limits will in-
⁎ Corresponding author. fluence a new firm's openness strategies, and also suggest which parts
E-mail addresses: [email protected] (J. West), [email protected] (G. Kuk). of a complex offering will be selectively opened (or closed).

http://dx.doi.org/10.1016/j.techfore.2015.07.025
0040-1625/© 2015 Published by Elsevier Inc.

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
2 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx

Fig. 1. MakerBot and Thingiverse success.

The paper begins by reviewing the relevant theoretical literature, A particular type of systems innovation is a platform, which has two
and then summarizes the technical and market context of late 20th cen- key attributes. The platform provides well-defined interfaces that en-
tury industrial 3D printing that enabled a rash of consumer-focused able third parties to provide complementary goods (such as “apps”)
startups in the early 21st century. We review the evolution of the for the platform, and the platform sponsor nurtures an ecosystem of
MakerBot product and content strategies, and from this suggest testable firms (or individuals) who supply such complements (Bresnahan and
propositions, other theoretical implications and opportunities for future Greenstein, 1999; Gawer, 2009). Sponsors selectively provide
research. openness — in some parts of their technology but not others — to attract
both adopters and a supply of complements while retaining enough
proprietary control to extract profits from the platform (West, 2003;
2. Open strategies for systems, communities
Boudreau, 2010; Eisenmann et al., 2011).
External user communities have been increasingly important to
The value proposition for consumer 3D printing requires two com-
firms in industries organized around digital goods. Firms can tap exter-
plementary offerings: systems and content. Consumer value is realized
nal communities as part of an open innovation strategy (Spaeth et al.,
when digital designs are transformed into physical objects using a
2010; West, 2014). Such communities provide a source of complemen-
system that includes graphics software, a computer and a 3D printer.
tary goods, particularly for information goods such as online informa-
Content means the creation, modification and sharing of these digital
tion (Nov, 2007), musical sounds (Jeppesen and Frederiksen, 2006)
representations of tangible objects — often shared through online
and video games (Jeppesen and Molin, 2003). Online communities can
communities. Digital designs are complementary goods that make the
also provide support (Lakhani and von Hippel, 2003), bug reports
adoption of 3D printing systems more valuable.
(Dahlander and Magnusson, 2008) and marketing feedback (Schau
Here we review research on the role of complementary goods in sys-
et al., 2009). For would-be entrepreneurs, the firm-sponsored commu-
tems adoption, with a focus on community-sourced complements. We
nities of today's “app economy” provide an opportunity to develop
also consider how firms interact with external communities, and selec-
and diffuse new add-on products (MacMillan et al., 2009).
tively provide outbound openness (and accept inbound openness) to
advance firm goals, particularly in the context of open source software
2.2. Selective openness
and hardware.
Firms often choose to be selectively open in dealing with external ac-
2.1. Complementary goods and systems adoption tors (Alexy et al., 2013). Because too much revealing may erode compet-
itive advantage and enable competitors, they open part of their overall
Innovations are adopted based on their characteristics, and also the offering (such as a module within a complex system) or provide partial
characteristics of their adopters. In particular, earlier adopters (such as openness for a broader part of that offering (West, 2003). They freely
hobbyists or other enthusiasts) are more eager to try new technologies reveal information that could also induce potential rivals to follow a
without proven benefits and are tolerant of usage difficulties (Rogers, similar openness strategy, particularly for information that is less strate-
1995). For many products, the value of the innovation depends on the gically important (Henkel, 2006; Henkel et al., 2014). At the same time,
provision of certain complementary goods that make the innovation their willingness to be open may be limited by their ability to appropri-
more valuable (Teece, 1986). ate value from the non-open parts (West, 2006).
The creation and adoption of a technological system is different from Over the past 15 years, most of the experimentation in (and research
other kinds of innovation, in that the value received by the adopter on) firm openness has focused on open source software. Firms have
depends on the value of the overall system. Promoters of such systems worked with and created external online communities that create free
innovation face the challenge in combining and coordinating the pro- and open source software (Lakhani and von Hippel, 2003; Shah, 2006;
duction and adoption of different components of the system by external Dahlander and Magnusson, 2008). Such communities are characterized
actors (Maula et al., 2006). Although systems innovators can attract a by a standard form of IP license, a form of cooperative production and a
variety of complements, it is difficult to align external actors and pro- mechanism of governance (West and O'Mahony, 2008). The license al-
cesses. In relation to open innovation, systematic innovation entails locates rights to use and modify IP for community members and non-
boundary management (Jarvenpaa and Lang, 2011; Teigland et al., members alike (Perens, 1999).
2014) within and across the borders of firms. Open innovation research For systems adoption, these communities can provide the core func-
only partially explains how firms may combine open and proprietary tionality of a system, enabling firms to sell higher-value proprietary
components that are subject to different licensing regimes (West, complements. Conversely, firms may choose to offer a proprietary plat-
2006). form and seek open source donations of complements (West and

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx 3

Gallagher, 2006). Overall, such communities provide an important their industrial forbears. The influence of an early 3DP open source
source of open innovation for firms (West and O'Mahony, 2008; Piva hardware community also allows us to study how the openness strate-
et al., 2012). gies of these 3DP vendors are similar to or different from the often
Open source communities have enabled the creation of new firms, studied open source software companies.
by providing the technology necessary for entrepreneurial entrants to To better develop theory and be able answer “how” and “why”
provide new offerings (Gruber and Henkel, 2006; Dahlander, 2007; questions, we employed an exploratory research approach using a qual-
Piva et al., 2012). In many cases, these firms create their offerings by itative case study (cf. Yin, 2003; Eisenhardt and Graebner, 2007). To un-
“forking” off a proprietary version from the community's shared source derstand the consumer segment, we use a case study on MakerBot
code. Entrepreneurial companies have also created communities, Industries, which we chose for three reasons. First, it is one of the
enforcing selective openness through rules that keep the community most successful consumer 3DP firms: the best known in the US, the
activity aligned to support the firm's objectives (Dahlander and global market share leader, and the first to reward its founders with a
Magnusson, 2005; West and O'Mahony, 2008). successful exit. Second, its founders were first to establish a
While technologies from open source communities reduce firm consumer-oriented digital content community to complement the
entry barriers and enable value creation for customers, the open tech- value created by 3D printers. Third, MakerBot used two distinct IP
nologies reduce the opportunity for uniqueness and thus value capture. strategies for its 3DP systems, shifting from mostly open to a mostly
In responses, firms developed complex offerings that combine open proprietary approach. We believe that MakerBot thus fits the guidelines
(commodity) and closed (proprietary) components, while embedding of Yin (2003) for unique cases in single-case research designs, and the
this value proposition in value networks (West and Gallagher, 2006; case selection advice of Eisenhardt and Graebner (2007: 27) for a case
Morgan and Finnegan, 2014). Success requires accessing external that is “unusually revelatory, extreme exemplar” and “particularly suit-
knowledge of new or established communities, while structuring the able for illuminating and extending relationships and logic among
rules inside and outside to constantly realign the community efforts constructs.”
with the firm's goals (Dahlander and Magnusson, 2008; West and We examine MakerBot's first five years of existence (January 2009–
O'Mahony, 2008). January 2014) — which includes its entire period as an independent
Enthusiasts have sought to replicate the success of open source company and its first few months after acquisition — and also the paral-
software communities to produce tangible goods including electronics lel existence of Thingiverse (an online content community with over-
and hardware. These efforts are often organized using processes and lapping founders) during this same period.
principles directly modeled on open source software. This includes We used multiple sources of evidence as recommended by Yin
modularization to minimize complexity, adapting open source IP (2003). The case study utilizes secondary data, and archival primary
licenses, a culture of collaboration, and the selective use of openness data. Information on the origins of the industry and technology is
by project sponsors. Compared to the production of software or other taken from a variety of secondary sources, including technical articles
information goods, however, open source projects producing hardware and books (e.g. Jacobs, 1992), news articles, FundingUniverse.com,
have higher barriers to entry and collaboration inertia, and a greater company websites and patents; particularly helpful were the annual re-
need for an external sponsor (typically a single firm) to provide ports on 3D printing by Wohlers Associates.
essential infrastructure such as manufacturing (Balka et al., 2009, To avoid retrospective biases (Eisenhardt and Graebner, 2007), the
2010; Raasch, Herstatt & Balka, 2009). primary data includes contemporaneous reports of key executive moti-
As with open source software, the initial research on open source vations, as captured by published interviews (Stern, 2010) and first
hardware emphasized community success. Unlike research on open hand accounts (e.g. Smith, 2008; Pettis, 2010). Information on the
source software (e.g., the work of Dahlander, Henkel, Piva, West and RepRap Project, MakerBot and Thingiverse were taken from the
others), little (if any) research has considered how firms achieve their community wikis and discussion group. Key postings of the RepRap Pro-
strategic objectives by working with these hardware communities. As ject from 2005–2011 are compiled in Hodgson (2012). The evolution of
Raasch and her colleagues have shown, the differences between soft- the Thingiverse web site was analyzed using the Internet Archive
ware and hardware participation and replication costs (among other (archive.org).
differences) lead to important differences in strategies for community To provide internal validity, we used pattern matching in the data to
success, and thus we would expect similar (if not exactly parallel) build possible explanations (Yin, 2003). In particularly, we analyzed the
differences in strategies for firm success. secondary and primary data to develop a historical narrative of the rela-
This paper examines this latter topic: firm strategies for leveraging tionship between MarketBot and Thingiverse. This approach sensitized
open source hardware. In particular, we focus on two specific questions. us to the complementary relationship between software and hardware.
First, how do new firms leverage open hardware communities? Second, For MakerBot, the study examined all 3D printer models released during
how does an open source hardware firm combine open and proprietary the study period, as well the associated software offerings during this
elements of their business model to create value while creating same period. In our analysis, we focused on the inflection points —
sufficient barriers to allow the firm to capture value? places where the strategy changed, particular in the degree of open-
ness — and thus sought to understand the strategy before and after
3. Research design each inflection point.
For Thingiverse, there were no corresponding discrete inflection
We examine the history of consumer 3D printing (3DP) within the points. We thus sought data that captured the high-level differences
settings of 3DP industry, focusing on the unfolding of major events among content and participant categories in the community, as well
over time as a key aspect of this process. In doing so, we can illuminate the secular growth of the community during the relevant period. In
the relationships between software and hardware, and between infor- doing so, we drew upon the broader perspective developed by the sec-
mation and tangible goods. As tangible goods, 3D printers play a unique ond author during a three year study of the Thingiverse community.
role in bridging the digital and physical worlds by providing a way to
transform digitally stored representations of objects into a tangible 4. Origins of consumer 3D printing
good.
This fundamental relationship and basic 3DP technology were The term 3D printing refers to a family of technologies that allow the
20 years old before the first consumer 3D printer was sold. We thus translation of digital designs into physical objects via a computer-
sought to explain why the consumer vendors emerged when they did, controlled custom manufacturing device. While “3D printing” once re-
and how their IP and product strategies were different from those of ferred to a specific process invented by MIT, today the term is used

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
4 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx

Table 1 Others adapted 2-D inkjet printing to build up 3D objects. This in-
Comparison of industrial and consumer markets. cluded MIT professor Emanuel (Ely) Sachs, who with his researchers
Source: Estimates as published in Wohlers Associates (2014).
was granted more than a dozen patents from 1993 to 2006, in a process
Industrial Consumer called “3D Printing”4 that MIT licensed to eight spinoff companies in the
First system sold 1987 2008 1990s (Shane, 2000). Other variants included Inkjet Printing, a low-cost
Units sold (through 2013) 66,700 140,425 system from Sanders Prototype for printing wax molds for jewelry and
Unit sales (2013) 9832 72,503 dental applications, and PolyJet, used in printers from Israel's Objet to
Average unit price (2013) $73,800 $1208
print hundreds of rows of photo-sensitive polymer plastics.
Systems revenues (2013) $976 million $88 million
An even simpler approach came from S. Scott Crump, who after tin-
kering with making objects with a glue gun formed Stratasys in 1988,
more generically to include a range of additive manufacturing filing his first of many patents in 1989. The Fused Deposition Modeling
technologies.1 (FDM) combined the mechanical advantages of the nozzle approach of
The industry and technology had two distinct phases and markets inkjet printing with a stable feedstock (a spool of plastic). Finally, Mi-
(Table 1). The first 20 years were limited to industrial and commercial2 chael Feygin cut and layered paper sheets to create wood-like structures
vendors, who used patented technology to sell hundreds (later thou- in a process called Laminated Object Modeling (LOM); Feygin formed
sands) of units every year at prices exceeding $100,000. The consumer Hydronetics (later Helisys) to commercialize the technology and (like
market was created and grew explosively after 2005 in response to DTM and MIT licensee Z Corp) received early SBIR funding from the
three factors: the “maker” movement of do-it-yourself hobbyists, the NSF Strategic Manufacturing Initiative.5
formation of the RepRap open hardware community, and the entry of In the industry's first decade, two companies — 3D Systems and
dozens of startup firms selling low-priced printers (priced less than Stratasys — were able to go public, gaining both one-time and ongoing
$5000) after the 2006 expiration of a key industrial patent. As a result, access to capital, as well as increased public visibility. Helisys went pub-
in 2013, consumers bought more personal printers than were pur- lic in 1996 but lost money until it failed in 2000. The surviving first gen-
chased by industrial buyers in the first 25 years. As with computing eration companies were acquired between 2001–2012 by one of the
and other technologies that shifted from business to consumer buyers, two market leaders.6
the shift required adapting the technology to be simpler and less In its first five years, 3D Systems captured 73% of the global market
expensive. by unit sales (Wohlers, 1992). However, it struggled for years to find a
To understand MakerBot's strategies and growth, we review the large enough market to support its growth and generate consistent
technology and business models of first decades of the 3DP industry, profits (Pendleton, 1994; Brooks, 1997). It took a decade of product
and then discuss how they were adapted by the consumer 3D printing sales, until in 1997 more than 1000 systems units were sold worldwide
firms of the 21st century. in a single year; in that same year four technologies (SLA, IJP, FDM, LOM)
each had between 15–26% market share, with Stratasys the market
leader with a 25% share (Kruth et al., 1998).
4.1. Invention and industrial applications of 3D printing
In 2013, 3D Systems and Stratasys remained by far the largest 3D
printing companies, with respective revenues of $513 and $484 million
A series of additive manufacturing technologies were invented in the
and year-end market capitalization of $9.6 and $6.6 billion. Together
20th century, most by U.S. inventors during the late 1980s (Table 2).
they accounted for 73% of the industrial printers sold that year
During this period, none emerged as a clear dominant design that
(Wohlers Associates, 2014).
displaced the others, with market share fragmented between multiple
technologies. All of these approaches involve creating a 3-D object as a
series of thin two-dimensional layers, one on top of another. These tech- 4.2. Consumer market
nologies were created by academic, corporate or individual inventors,
many of whom went on to found startups to commercialize the technol- For years, futurists and other analysts predicted that 3DP would be-
ogy (Table 3). These 3D printers had three initial applications: rapid come cheap enough to become available to consumers. For example,
prototyping, making molds for high-volume production and direct from 2010–2012 the world's largest printer manufacturer, Hewlett-
short-run production. Packard, sold a $17,000 FDM printer manufactured by Stratasys. By
Two of the major technologies involve having light focused to 2013, the consumer segment had developed to the point that dozens
solidify the input material at a specific location. Stereolithography of hobbyist-oriented models were sold at prices ranging from $300 to
(SLA) involves shining an ultraviolet light on a liquid resin to harden $3000 (Make, 2013). Consumer adoption required overcoming numer-
the liquid into plastic at a particular location. It was invented by Chuck ous barriers (Table 4).
Hill, who left his employer to found 3D Systems. As with personal computers, penetration of the consumer market
In parallel, Carl Deckard (a PhD student at UT Austin) invented Selec- began with the hobbyist market, in this case dedicated aficionados of
tive Laser Sintering (SLS), which uses an infrared laser to fuse a plastic or what during the early 21st century was termed the “maker” movement
metal powder. It was an important as the first technology to directly (Anderson, 2012). Consistent with Rogers (1995), these earliest adopters
produce metal objects. UT Austin was granted a series of patents begin- were (when compared to other consumers) the most motivated and tol-
ning in 1989, and licensed the patents to a number of spinoff firms. erant of performance limitations, complexity and (relatively) high costs.
Deckard cofounded DTM Corporation, which operated a 3DP service However, the eventual consumer market for 3DP was enabled by the
bureau starting in 1988 and sold its first printer products to industrial diffusion of technology (and concomitant cost reduction) provided by a
customers in 1992.3 3DP open design community. The RepRap Project was announced in
March 2005 by Adrian Bowyer, a UK mechanical engineering professor.
1 He used the commercial Stratasys printer in his lab both for
The summary of the key technologies is taken from the technical literature, including
Wohlers (1992), Jacobs (1992, 1996), Kruth et al. (1998), and Gibson et al. (2010).
2 4
While most of the early systems were sold to industrial manufacturers, a few systems According to US Patent & Trademark Office databases, MIT also applied to trademark
(such as those by Solidscape née Saunders) targeted commercial customers such as jew- “3DP” and a 3DP logo in 1992, but abandoned the application in 1994.
5
elers and dentists. In addition, the creation of SLS at U. Texas and 3DP at MIT were funded by NSF re-
3
Other metal fabrication techniques emerged in the 1990s. Independent inventor Ralf search grants (Weber et al., 2013).
6
Larson worked with Chalmers University of Technology to replace the laser of SLS with When Stratasys merged with Objet in December 2012, their respective shareholders
an electron beam and create Electron Beam Melting (EBM), while other researchers mod- split the combined equity 55/45. The successor company kept the Stratasys's name and
ified FDM to use spools of metal filament. NASDAQ stock listing but, like Objet, was incorporated in Israel.

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx 5

Table 2
Additive manufacturing technologies and patents.
Sources: Key additive manufacturing inventors, universities and firms were identified from technical publications. For this list, Google Patents was used to search for all US granted patents
prior to 2000.

Process First granted US patent (priority date) Key inventor (employer) Feedstock

Stereolithography (SLA) 4,575,330 (1984) Chuck Hill (UVP, later 3D Systems) Liquid plastic
Laminated Object Modeling (LOM)a 4,752,352 (1986) Michael Feygin (later Helisys) Paper
Selective Laser Sintering (SLS) 4,863,539 (1986) Carl Deckard (U. Texas) Plastic or metal powder
Solid Ground Curing (SGC) 4,961,154 (1986) Itzchak Pomerantz (SciTex, later Cubital) Liquid plastic
Fused Deposition Modeling (FDM)b 5,121,329 (1989) Scott Crump (Stratasys) Continuous spool of plastic (later metal)
Electron Beam Melting (EBM) 5,786,562 (1993) Ralf Larson (Larson Brothers) Metal powder

Inkjet-based approaches
Three-Dimensional Printing (3DP)c 5,204,055 (1989) Michael Cima, Emanuel Sachs (MIT) Liquid plastic or plastic-metal
Inkjet Printing (IJP) 5,506,607 (1991) Royden Sanders Jr. (later Solidscape) Wax
PolyJet 6,259,962 (1999) Hanan Gothait (Objet) Liquid plastic

Additive manufacturing equipment producers listed in bold.

a
Trademark of Helisys.
b
Trademark of Stratasys; competitors use the term Fused Filament Fabrication (FFF).
c
Trademark sought by MIT, later abandoned.

Table 3
Key 3D printing companies.

Founded Company Spinoff Parent HQ Printing Process First System Initial Target Exit

1985 Helisys (née Hydronetics) Los Angeles LOM ca. 1990 Industrial 1996: IPO 2000: out of business
1986 3D Systems UVP Los Angeles SLA 1987 Industrial 1987: IPO (Vancouver)
1986 Cubital, Ltd. Scitex Israel SGC 1991 Industrial 2000: out of business
1987 DTM Corporation UT Austin Austin SLS 1992 Industrial 2001: Acquired by 3D Systems
1989 EOS GmbH Germany SLA, SLS 1994 Industrial n/a
1989 Stratasys Minneapolis FDM 1992 Industrial 1994: IPO (NASDAQ)
1990 Materialise NV KU Leuven Belgium Service Bureau – Industrial 2014: IPO (NASDAQ)
1993 Solidscape (née Sanders Prototype) New Hampshire IJP 1994 Commercial 2011: Acquired by Stratasys
1994 Z Corp Boston 3DP 1997 Industrial 2012: Acquired by 3D Systems
1998 Objet Israel PolyJet 2001 Industrial 2012: Merged with Stratasys
1996b ExOne Extrude Hone Pittsburgh 3DP 1999 Industrial 2013: IPO (NASDAQ)
1997 Arcam Chalmers University Sweden EBMa 2002 Industrial 2000: IPO (Nordic Growth Market)
2007c Shapeways Phillips Netherlands Service Bureau – Commercial n/a
2009 Afinia Minneapolis FDM 2012 Consumer n/a
2009 MakerBot New York City FDM 2009 Consumer 2013: Acquired by Stratasys
2011 RepRap Professional RepRap Project UK FDM 2011 Consumer n/a
2011 Ultimaker Netherlands FDM 2011 Consumer n/a
2011 Formlabs Boston SLA 2012 Consumer n/a

Processes invented by a company founder or employee are listed in bold.

a
Exclusive patent license from inventor.
b
Parent company began 3D printing in 1996, spun off in 2005.
c
Spun off as independent company in 2010.

experiments and to fabricate parts, and used his prior 3DP experience to RepRap.org website. The community built printers using a variety of
suggest a variety of nozzles for an FDM-based system that would be open source software and hardware tools.
considerably less expensive (although slower and lower resolution) Over its first five years, the RepRap Project focused on reducing cost
than existing industrial 3DP systems. and improving performance of the open hardware design. Consistent
As the first Stratasys patent was approaching its 2006 expiration date, with other user communities (cf. Lakhani and von Hippel, 2003), it
Bowyer created IP rules for RepRap consistent with open source software also allowed users to share knowledge about 3DP usage, beyond design-
and open design communities (cf. West and O'Mahony, 2008; Raasch ing improvements to the RepRap open hardware. In particular, commu-
et al., 2009). In his initial announcement on the RepRap.org weblog nity members identified possible sources of open source and other free8
(Hodgson, 2012: 1–5), Bowyer said a major goal was to put “ideas into 3D design software that would replace the $1000+ CAD packages used
the public domain as soon as possible, to ensure that they are unpatent- by industrial designers and engineers (cf. Hodgson, 2012).
able”; he sought contributions that were free both in price “as well as in As with open source software (cf. Rossi, 2006), the open source
freedom”, a nod to the free software ideology of Stallman (2002). hardware community of user-innovators focused on “scratching an
To enable hobbyist participation, a major milestone was using a itch”, i.e. eliminating barriers to their own use of a 3DP system. As
RepRap printer to make another printer; the first such copy was made with OSS, they spent less time addressing problems that impacted ordi-
in 2008. RepRap thus appealed to do-it-yourself hobbyists, variously re- nary consumers: in particular, hobbyists had a higher tolerance for ease
ferred to as “DIY”, “makers” or hardware “hackers”.7 Bowyer formed an
online community to develop and refine the hardware designs using the
8
Community members emphasized software that used both open source licensing and
7
The maker movement was organized through Make, a magazine first published in community production processes (cf. West and O'Mahony, 2008). However, 3DP hobby-
2005 by O'Reilly Media, a San Francisco Bay Area company that published numerous books ists also used limited-functionality versions of proprietary software provided free by com-
on open source software. mercial software companies such as Autodesk and Google (Make, 2013).

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
6 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx

of use problems (of the hardware, software and entire system) than Table 4
would the average consumer. Potential barriers to consumer adoption of 3D printing.

Also in parallel to open source software (cf. Dahlander, 2007), the Category Attributes
availability of open hardware enabled entry by firms. Most utilized the Printer performance Speed, quality, output size, output durability
same FDM technology as RepRap in their products, competing with Cost Initial cost, cost of consumables
each other and the RepRap open hardware. Some sold RepRap- Computer performance CPU cycles, RAM, hard disk
derived designs: Bowyer himself founded one such company, RepRap Application software Computer Aided Design (CAD) applications
Ease of use Graphical interfaces, hardware operation, integration
Professional, which launched its first product in late 2011. Other firms
Content (digital designs) Self-designed, standardized components, community
such as Deezmaker and Ultimaker used what they learned from the donated content
RepRap project to create their own designs.9 Many of these companies
had their initial production runs crowdfunded by fundraising cam-
paigns on Kickstarter (e.g., Formlabs) or Indiegogo (RepRap Pro). In
2012, the largest industrial 3D printing company (3D Systems) Only two months after they launched their company, the founders
launched a consumer-oriented Cube printer line. used the annual South by Southwest (SXSW) conference in Austin to
The global market share leader among consumer 3DP companies unveil their first product: Cupcake CNC, a $750 kit to assemble a 3D
was MakerBot Industries, which offered its first printer in 2009 printer. The design was based on the RepRap open hardware design,
(Wohlers Associates, 2013) and controlled the Thingiverse content but moved the base rather than the printing head to reduce complexity
community. When combined with its own custom software, the compa- and weight. MakerBot shipped its first batch of 20 orders in April 2009,
ny was the earliest to offer an end-to-end system of hardware, software and by the end of the year, increased its monthly production batch to
and content for hobbyists and other consumer users. By the end of 2013, 150 units per month (Stern, 2010).
it had sold nearly 44,000 printers and had grown to 450 employees MakerBot built its hardware based on Arduino, an open source elec-
worldwide (Biggs and Kumparak, 2014). tronics controller technology. It made the design of its first printer pub-
licly available using the same free software license (the GNU General
Public License) as used by RepRap. The GPL meant that all hacker chang-
5. MakerBot es had to be shared back with MakerBot. The same license was also used
with its second product, the Thing-O-Matic, which was released the fol-
In January 2009, MakerBot Industries, LLC was founded in Brooklyn lowing year in both kit and assembled form. In an interview, Pettis cited
by Adam Mayer, Bre Pettis, and Zachary “Hoeken” Smith. The three the benefits it gained from open hardware designs:
were active members of NYC Resistor — a “hackerspace” for the New
York City DIY community; Smith was also active in the RepRap online Because we're open source and the community is so smart, we've
community. Pettis became the CEO, Smith focused on hardware and seen a lot of participation in the research and development sector.
Mayer on software. … Because we're open source, our users know that the code and de-
MakerBot developed a series of printers based on both the RepRap signs are theirs to hack on. They also know that if they improve their
printer design and RepRap-compatible object designs distributed via machine, they can share their improvement and everyone in the
Thingiverse. From its open design roots, the company adopted an in- community benefits (Peels, 2010).
creasingly proprietary strategy to differentiate itself from the RepRap
open hardware and other consumer-oriented 3D printer companies. Endorsing this open approach, Pettis criticized an entrepreneurial
This included both proprietary hardware components to improve the hardware designer who was selling a derivative of MakerBot's RepRap
performance of its printers, and new proprietary software applications board without sharing her changes:
to improve the overall ease of use of its system.
To fund their first product, the founders raised $50 k from a local en- Open Source Hardware is hardware that has an open license. You can
trepreneur and $25 k from Adrian Bowyer, founder of RepRap (Pettis, copy it, develop it, and even sell it yourself. You must provide attribu-
2011). After an undisclosed angel round, in August 2011 the company tion to the designer and you must also release the derivative source
became the first consumer 3D printing company to receive venture cap- files under the same license.…Sometimes an individual or a company
ital with a $10 million in Series A venture funding10; soon after that, makes a derivative of an open source project, goes to market with it
both Mayer and Smith left the company. In August 2013, MakerBot and then doesn't share their derivative designs with their changes.
was acquired for more than $450 million by Stratasys, the second largest This is not only against the license, but it's also not ethical. It is a dead
maker of industrial 3D printers.11 The respective company CEOs vowed end for the innovation and development which is the heart of the
that MakerBot would continue to be run as a separate division. open source hardware community. …At MakerBot, we take open
source seriously. It's a way of life for us. We share our design files
when we release a project because we know that it's important for
5.1. Hardware: From open to closed hardware our users to know that a MakerBot is not a black box. …When people
take designs that are open and they close them, they are creating a
During its 56 months of independent existence, MakerBot developed dead end where people will not be able to understand their machines
five consumer-oriented printers (Table 5). All were PC-controlled and they will not be able to develop on them (Pettis, 2010).
printers derived from the RepRap designs, and (like RepRap) used the
FDM approach of melting a plastic filament from a spool. In January 2012, MakerBot made several major changes in its prod-
uct strategy with the first of its Replicator products, named for the ubiq-
9
Bock et al. (2014) identify more than 120 startup companies in consumer 3D printing uitous fabrication device of the Star Trek television show. While the
from 2006-13.
10 product design was publicly available, the design was provided under
Contrasting MakerBot to one of its rivals, in Fall 2012 Formlabs received $2.9 million
from 2068 backers in one of the most successful crowdfunding campaigns in Kickstarter a Creative Commons license, a less restrictive license. With the
history. After MakerBot was acquired, in 2013 Formlabs received $19 million in Series A Replicator, MakerBot stopped selling printer kits, and only offered as-
funding. sembled products, emphasizing ease of use and reliability. The
11
The 4.7 million Stratasys shares offered for MakerBot were valued at $403 million Replicator models also offered enclosed cases that reduced the risk of
when the deal was announced on June 19, but were worth $455 million when the acqui-
sition completed on August 15. The purchase also included an additional 50% stock divi-
having the printer catch something while printing. These attributes sup-
dend incentive based on performance through the end of 2014 (Stratasys, 2013). By ported the company's stated “mission to bring MakerBots to the desk-
comparison, the 1994 Stratasys IPO raised less than $6 million. tops of everyone” (Bilton, 2012).

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx 7

Table 5
MakerBot product history, 2009–2013.
Source: MakerBot website, news reports.

Date Product Pricea Design License Build Volume Min. Layer Resolution Printing Material

March 2009 Cupcake CNC $750b GNU GPL 100 × 100 × 150 mm n.s. PLA (Polylactic Acid)
Sept. 2010 Thing-O-Matic $1300b GNU GPL 110 × 110 × 120 mm 0.25 mm PLA
$2500
Jan. 2012 Replicator $1750 Creative Commons 225 × 145 × 150 mm 0.20 mm PLA or Acrylonitrile Butadiene Styrene (ABS)
Sept. 2012 Replicator 2 $2200 trade secret 285 × 153 × 155 mm 0.10 mm PLA
Jan. 2013 Replicator 2X $2800 trade secret 246 × 163 × 155 mm 0.10 mm PLA or ABS
Jan. 2014 Replicator Minic $1380 trade secret 100 × 100 × 125 mm 0.20 mm PLA
Jan. 2014 Replicator (5th generation)c $2900 trade secret 252 × 199 × 150 mm 0.10 mm PLA
Jan. 2014 Replicator Z18c $6500 trade secret 305 × 305 × 457 mm 0.10 mm PLA
a
Base model at time of introduction.
b
Kit form.
c
Introduced after being acquired by Stratasys.

However, in September 2012, MakerBot reversed its IP strategy with Industries, it makes me ashamed to have my name associated with it
the Replicator 2. Unlike the previous products, the design of the (Smith, 2012).
Replicator 2 (and all subsequent printers) remained a trade secret,
just as the design of earlier industrial 3D printers had been a trade se- In a subsequent blog posting, Pettis questioned the viability of the
cret. Pettis was quoted as saying open hardware model for larger companies:
I wish there were more examples of large, successful open hardware
For the Replicator 2, we will not share the way the physical machine
companies. … There are no models or companies that I know of that
is designed or our GUI because we don't think carbon-copy cloning is
have more than 150 employees that are more open. … We are
acceptable and carbon-copy clones undermine our ability to pay
experimenting so that we can be as open as possible and still have
people to do development (Brown, 2012).
a business at the end of the day. … I don't plan on letting the vulner-
abilities of being open hardware destroy what we've created. …This
The “cloning” was a reference to TangiBot, a (legal) direct copy of the
isn't the first change we've made to become more of a professional
Replicator announced several months earlier to be sold for $1200 (one
business, and it won't be our last (Pettis, 2012).
third less than the Replicator) through manufacturing in China. The de-
signer of the TangiBot sought $500,000 in crowdfunding to launch his
Meanwhile, after its acquisition by Stratasys in 2013, MakerBot also
company, but the effort failed due to resistance from members of the
benefitted from its parent company's extensive patent portfolio. After it
3DP and open hardware communities.
announced three new printers in early 2014, MakerBot published on its
Despite this provocation, MakerBot's decision to switch from being a
website a list of five patents that it asserted covered its five most recent
pioneer of open source hardware to a closed source design raised con-
printers: four Stratasys utility patents and one MakerBot design patent
siderable controversy. This included a blistering attack from co-
(Table 6). One Stratasys patent (6,004,124) was among four listed in a
founder Smith, who had been forced out earlier in the year:
patent infringement lawsuit filed in November 2013 by Stratasys
For me, personally, I look at a move to closed source as the ultimate against Afinia, the importer of a low-cost Chinese FDM printer that com-
betrayal. … Moving from an open model to a closed model is con- peted with MakerBot's products (cf. Weinberg, 2013) (See Table 5).
trary to everything that I stand for, and as a co-founder of MakerBot
5.2. Software: From open source to proprietary

Over its four-year independent existence, MakerBot evolved its soft-

Table 6
ware strategy — as with its hardware design, from a mostly open to a
Self-declared patents applicable to MakerBot products. mostly closed strategy (Table 7).
Source: www.MakerBot.com/patents; also US PTO. Its first printer, the Cupcake CNC, used both existing and new open
Patent Priority Issue Assignee Applicable products
source software for modeling and design. This included Sanguino, an
date date open hardware fork of the Arduino project by Smith originally devel-
oped for the RepRap hardware, that used Arduino's open source soft-
6,004,124 1/26/98 12/21/99 Stratasys Replicator 2,2X,Mini, Z18, 5th gen
6,722,872 6/23/99 4/20/04 Stratasys Replicator Z18 ware libraries for communicating between the computer and the
6,749,414 4/30/01 6/15/04 Stratasys Replicator 2X Sanguino microcontroller board. Smith and Mayer incorporated the
7,384,255 7/1/05 6/10/08 Stratasys Replicator 2,2X,Mini, Z18, 5th gen Arduino software in ReplicatorG, a new open source software project
D677,723 9/18/12 3/12/13 MakerBot Replicator 2,2X
that was the driver for CupCake (and later Thing-O-Matic) printers.
The company also used skeinforge, an open source program popular

Table 7
Commercial and open source software for consumer 3D printing.
Source: project websites, Make (2013).

Category Input Output Open source examples Commercial examples

Modeler Drawing file (.dwg, .dxf), Stereo lithography file (.STL) Blender, OpenSCAD, Art Google Sketchupa, Autodesk
wavefront file (.obj) of Illusion 123D Designa
Slicer STL file Layer definition (G-Code) file skeinforge, Slic3r MakerWare, KISSlicera
Printer Driver G-Code file Printer commands ReplicatorG MakerWare
Controller Driver Library Programmer APIs Low-level serial output Arduino, Sanguino
a
Entry level version available free.

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
8 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx

with RepRap users that converted 3D designs into layers that could be free, but the printers were not — following a common slogan among
printed. open source hardware businesses: bits are free, atoms cost money.
With the 2012 release of its first Replicator printer, the company Because existing 3D content sites were selling designs, the
stopped supporting these two OSS projects, supplanting them with its Thingiverse rules were intentionally designed to encourage sharing
new MakerWare software. The proprietary MakerWare software ad- and thus increase the value of using the open RepRap hardware
dressed two issues. First, it allowed MakerBot to correct some of the (Smith, 2011). This included using the Creative Commons and open
problems the earlier open source software had rendering designs, thus source software licenses utilized by RepRap and other open source
improving output quality. Second, the new software simplified the pro- hardware and software projects, to encourage free revealing and reduce
cess of producing output, improving ease of use. the friction of transaction costs (cf. Harhoff et al., 2003; Balka et al.,
While MakerBot continued to recommend open source modeling 2009).
applications, the company made ongoing improvements in its The MakerBot founders used an unusually open community strategy
MakerWare software. Unlike earlier open source software, these im- in not selling or directly monetizing any of the designs.12 In comparison,
provements were not available to open source hobbyists (e.g. members both the Shapeways and Materialise service bureaus created online
of the RepRap Project) or to proprietary rivals. At an open hardware communities to sell user-generated content that is printed on demand —
conference in 2012, Pettis justified its newly proprietary software strat- as did 3D Systems which acquired an industrial service bureau and used
egy to maintain a competitive advantage in ease of use: it to launch its consumer-facing Cubify service bureau (Table 8). Only
Ultimaker (a direct competitor) emulated MakerBot in creating a com-
We're also not sharing the GUI of MakerBot MakerWare, which is the
munity to enable the free use and sharing of designs.
software that runs it. That's just because we want to have a chance to
control the look and the feel and the experience of the user. And
these things are really valuable to people who wanna clone us and
6.2. Structuring community participation
just make, like, carbon-copy clones, and we're not…and we're not
into that. But, it's still hackable, still modifiable (Ragan, 2012).
In its first five years, the community attracted more than 100,000 3D
designs donated by community members. Reasons for this success in-
clude pent-up interest in 3DP, global online availability, and its first
6. Creating and leveraging Thingiverse mover status as an open repository. However, a key aspect of the suc-
cess of Thingiverse was due to deliberate choices to promote sharing, in-
The Thingiverse community had a consistent policy of openness cluding reducing participation barriers, motivating contributions and
with free downloadable content. The shift of MakerBot to closed hard- enabling cumulative innovation (cf. Murray and O'Mahony, 2007;
ware eliminated the value of Thingiverse to hackers sharing modified West and O'Mahony, 2008).
printer designs — as well as rival printer makers — but value of other While rights to donated designs rested with the designer,
open content to ordinary consumers remained unchanged. Thingiverse encouraged contributors to utilize one of the well-known
Creative Commons licenses. Most of these licenses allowed other com-
munity members to create derivative works. One popular license (Cre-
6.1. Launching the community ative Commons–Attribution–Share Alike) required (as with the open
source GPL license) that required those making modifications must
Before MakerBot, Smith and Pettis started Thingiverse in October share back their changes with other community members. This ap-
2008 to encourage open sharing of digital designs for physical products, proach facilitated learning and modification, and served to bind the
complementary goods that would make 3D printers more valuable. As community of users further through nurturing reciprocity among de-
Smith explained in his announcement on the official RepRap blog: signers. Developing such norms of reciprocity proved an open factor
in motivating contributions to free software communities (cf.
One of the most frequent questions I get after people understand
O'Mahony, 2003).
what a RepRap does, is a variant of either ‘Why do you need a ma-
Another key institutional choice by the community's founders was
chine like this?’ or ‘What do you make once you have one?’. Well,
to emphasize attribution of design works, providing recognition for de-
Thingiverse.com is an answer to that. This is no ordinary object shar-
signers. Even after MakerBot shifted from open to closed hardware —
ing website. Thingiverse.com is a home for all your digital designs. If
and moved to assert tighter control over the operation of the
you can represent a physical object digitally, then we want it on
Thingiverse website — Pettis emphasized the company's commitment
Thingiverse. (Smith, 2008).
to designer attribution: “The legal terms of use are there to keep
Thingiverse legit and protected, not to take away attribution. We
As Pettis explained in a 2010 interview
know attribution is critical in a community of sharing” (Pettis, 2012).
We built Thingiverse because we needed a place to share our designs Attribution facilitates the growth of the community. Thingiverse has
so we wouldn't lose them and so our friends could make what we grown and evolved into much more than an Open Source and Innova-
had made and then modify those designs and make them better. tion Sharing platform, but also into the biggest 3D learning community
The community is amazing and supportive, and it's also a lot of in the digital world (Baichtal, 2008). With more than 100,000 designs
fun. There is no other place that you can share a design for a physical and more than 21 million downloads by June 2013, Thingiverse experi-
thing and people around the world will make their own copies with- enced an “explosion of uploaded and published 3D designs” (Howard,
in minutes … It's that kind of sharing magic that makes Thingiverse 2013), fulfilling its founders' vision of creating a universe of things. At
the closest thing to teleportation that we've got in this solar system. the end of 2013, the most downloaded objects included decorative ob-
(Peels, 2010). jects, toys, small useful objects (such as iPhone cases), and component
designs (such as circuit boards or robot arms) intended for other
While initial uploads were primarily vector drawings for 2D laser makers.
cutters, the user-generated content rapidly expanded to include 3D
models, electronics, and designs, in a range of 2D and 3D object file for-
mats. Content created by an individual “object designer” was shared 12
Over time, MakerBot made its brand more prominent, changing the official name to
freely for other community members to download. This community cre- “MakerBot Thingiverse” and requiring a free MakerBot account to log onto the Thingiverse
ated the first open repository for digital 3D designs. Its design files were site.

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx 9

Table 8
Online content communities sponsored by 3D printing companies.

Community Sponsor Related sponsor revenues Launch date Free designs Designs for sale Printed versions for sale

Thingiverse MakerBot Consumer 3D printers 2008 X

Shapeways Shapeways Consumer service bureau 2009 X X
i.materialise Materialise Industrial service bureau 2010 X X
Cubifya 3D Systems Consumer service bureau 2012 X X
YouMagine Ultimaker Consumer 3D printers 2013 X
a
Launched based on its 2011 acquisition of the Quickparts industrial service bureau.

6.3. Benefits to MakerBot employees of MakerBot that get value from the company. I'll be the
first to say that hardware, open source, and investment are a messy
MakerBot focused on two distinct user audiences. One were the bunch of ingredients to stir up, but we're going to do our best to
makers who — embracing the open hardware ethos — assembled (and make it work (Pettis, 2012).
modified) MakerBot kits, using them to print physical objects. The
others were fabbers (fabricators) who use code (rather than manual While the closed approach deterred contributions to the software
drawing) to describe designs. Most of the fabbers have their attention and firmware that MakerBot used for its printers, it appears to have
to the procedural designs, and release their design files in the form of had limited impact on Thingiverse. Most of the design activities and
openSCAD files, which are intended to facilitate learning and feedback. the sharing of design files remain unaffected, as design files in the
In an interview, hobbyist design engineer Syvvich described the func- Thingiverse are primarily code base subject to the protection of copy-
tion of openSCAD as follows: right. The attribution system works with fabbers as in the way meritoc-
racy works in other open source software communities.
Where you don't design the parts, but you write software and it
MakerBot also increased its branding influence over Thingiverse. In
draws them, which is very cool because you can give various charac-
2011, it successfully registered a US service mark for Thiniverse, and
teristics and every time you need a different gear, you just change
in 2014 was granted a US trademark for a Thingiverse mobile app. In
the software for your parameters, so the 3D printed clock is one
2012, the Thingiverse home page was modified to say a “a MakerBot In-
big computer program-a great way to work.
dustries website”. In 2013, the website was rebranded “MakerBot
Thingiverse,” and users were allowed to log on using either a
Syvvich is a typical hobbyist whose interest is primarily about using
Thingiverse or MakerBot account, but after January 2014 the site re-
code to express design, but seldom 3D prints his designs. The differ-
quired new users to create a MakerBot account.
ences between fabbers and makers was starkly illustrated by the differ-
Thingiverse still remains as the largest repository for 3D designs,
ences in their reaction to MakerBot's decision with the Replicator 2 to
with copyright and attribution used to bind user commitment to ongo-
shift to a closed (i.e. proprietary) hardware design. Steeped in the
ing sharing and modification of each others' designs. Thingiverse forms
ethos of open source software, many of the makers criticized the shift
an important bridge for the transition between the digital to the physi-
away from open IP policies. Makers who are also fabbers are a minority,
cal worlds, and with freely available designs induces not only an in-
although some pulled their designs from Thingiverse to protest
crease in sales of 3D printers, but also the growth of materials. It also
MakerBot's new IP strategy.
remains an online platform for open design, despite the seeming diver-
In contrast to the makers, fabbers are more sympathetic and wel-
gence with MakerBot's IP strategy with its shift from open to closed
come the trade-off between having a better set of tools for design and
design.
the tools becoming closed source, as this Thingiverse community mem-
ber wrote:
As a professional graphic artist and computer animator, if I was re- 7. Discussion
quired to use only open-source software and tools to keep a “Fully-
Open title” I would go insane. I would be anchored to the sophistica- This case study of MakerBot is one of the first studies of firm strate-
tion of the tools and I can tell you, while they are excellent tools and gies for profiting from open hardware designs. Such studies are numer-
are getting better all the time there is a lot of proprietary technology ous for open source software, but previous studies of open hardware
that is nearly essential to functioning in this industry.… Most people have emphasized community rather than firm success.
who use 3d printers professionally couldn't care less how it works As a startup firm, MakerBot quickly led the consumer 3D printing
but THAT it works and right now, the challenges of staying “Open” market in market share and market value. MakerBot delivered a sys-
and being competitive in the marketplace are unimaginably difficult temic innovation by utilizing a three-way value network that linked
for this [open source hardware] business model. 13 existing (externally supplied) open source content generation tools,
the Thingiverse open content community that it sponsored, and a com-
The segmentation of the installed base of Thingiverse users under- bined open source hardware/software printing system that it sold as the
lines the challenges of mixing open and closed strategies for hardware, initial MakerBot 3D printer models. To create barriers to potential rivals,
software and content, dividing the opinions of contributors, users and it migrated the latter from an open to proprietary design, while keeping
investors. In response to the maker community claiming he had violated open the remainder of its value proposition.
the GPL by closed sourcing the hardware, Pettis wrote Here we summarize the complementarity of MakerBot's selective
openness strategies, and offer more general theoretical predictions for
It's important to me that we are not violating the licenses of the soft- other firms utilizing selective openness to gain competitive advantage.
ware. In regards to VC funds, MakerBot does have a duty to do its We conclude with a discussion of limitations and future research.
best to create value for its shareholders, that's part of startup life.
But it's not just the VCs, angel, and seed investors, it's also the
7.1. MakerBot's complementarity of openness

This study points to what we term the complementarity of openness —

13
Comment appended to Pettis (2012) by user Erik J. Durwood II, September 23, 2012. how openness supports a firm's proprietary strategy and vice versa. We

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
10 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx

believe this applies both to firms making tangible and intangible hardware to deter imitative entry, but largely retained the openness of
products. their complementary content (cf. van Burg et al., 2014).
The most oft-studied use of the open source model has been for soft- The founders of Thingiverse used an open strategy to attract contribu-
ware, which has been both used by independent communities and by tors. As with other content communities (and two-sided markets more
entrepreneurial firms that leverage these communities and sponsor generally), the large supply of content on the site provide network effects
their own communities (Dahlander and Magnusson, 2005; Gruber and that fuel a positive-feedback loop of new contributors and new content
Henkel, 2006; West and O'Mahony, 2008). However, Raasch (2011; (cf. Asvanund et al., 2004). The Thingiverse open content was originally
Raasch et al., 2009; Balka et al., 2010) noted the paucity of examples intended to support open source RepRap printers, but soon added value
for this approach being viable for supporting hardware, in part because to MakerBot's printer designs; the open content was complementary
the costs required for designing and producing tangible goods have to (and thus added value) to both open (RepRap, early MakerBot) and pro-
be borne by a firm. Their studies show how open hardware communi- prietary (later MakerBot) hardware. While the open content thus created
ties rely on a symbiotic collaboration with one (and probably only value for both open and proprietary strategies, only the latter provided for
one) firm, suggesting that communities depend on the ability of the enough value capture to fuel the firm's growth ambitions.
sponsoring firm to develop a viable long-term business model. MakerBot used the open components to complete the value proposi-
The MakerBot case allows us to extend knowledge of firm's strategic tion for customers, while gaining advantage by the rapid evolution of its
use of openness in general, and the complementarities of openness in proprietary product that was loosely coupled (through modular inter-
particular. MakerBot's founders used a range of evolving open commu- faces) to the open components. Such loose coupling allows more
nity strategies to create value while increasing its ability to capture rapid, decentralized innovation — as was earlier demonstrated for
value. Much like open source software companies, MakerBot launched open source software (MacCormack, Rusnak & Baldwin, 2006). At the
its first products by building upon the freely-available open source same time, MakerBot's three-part value creation was enabled by
hardware designs, and also openly shared the designs for its initial aligning the IP modularity to the technical modularity — so that interac-
products. However, unlike OSS companies compelled to share designs tions between the different IP strategies were isolated to the well-
by free software licenses, MakerBot stopped sharing its hardware defined technical interfaces (as later advocated by Henkel et al., 2013).
designs to build barriers to imitation and win venture investment.14
The company's CEO asserted that open hardware companies would be 7.2. Selective openness as an entry strategy
unable to grow if they kept an open IP strategy, raising questions as to
whether firms can provide long-term sponsorship of open hardware More generally, MakerBot demonstrates how openness both enables
communities the way they have done for OSS communities. and limits the entrepreneurial strategies of new firms. The availability of
In its software strategy, MakerBot more closely resembled a market- open source technology reduces entry barriers, enabling the formation
leading proprietary software company (cf. West, 2003). It leveraged and market entry of new firms (Gruber and Henkel, 2006); such efforts
open source software when convenient, and even launched an OSS pro- correspond to the bricolage used by new firms to gain competencies de-
ject of its own, but abandoned both when it conflicted with its goals of spite limited resources (Baker et al., 2003). As with many open source
using proprietary advantage in software ease of use and performance software companies (and open hardware companies studied by
as a barrier to imitation. While once-proprietary firms embrace hybrid Raasch et al., 2009; Bock et al., 2014 and others), MakerBot used the
strategies of selective openness by “opening parts” (West, 2003), availability of open source technology to launch its products — in this
MakerBot took the opposite route — in effect, “closing parts”. case leveraging both open hardware and open software. Two of the
However, even after its 2013 acquisition the company continued to three founders launched an open content community even before
sponsor Thingiverse as an open community for user generated content. they had a firm, let alone angel or venture funding. While an open com-
The firm allowed (and encouraged) MakerBot customers and other munity makes it easier for a firm to create value, it makes it harder to
users to freely share their digital designs with little or no restrictions, capture value (Simcoe, 2006).
for use with MakerBot and other 3D printers, as these designs created Such an external innovation community is an example of a comple-
complementary value for the firm's (now proprietary) products. mentary asset as defined by Teece (1986). Assets that are specialized to
This pattern thus suggests a complementarity of openness, between a particular firm's innovation will create value for that innovation, while
the open and proprietary elements of a firm's strategy. It's difficult to those that are generic (and thus more open) will benefit other innova-
support a firm's operations with a fully open offering; this is particularly tions including (in many cases) the firm's competitors. While firms
true for goods with non-zero replication costs (cf. Raasch et al., 2009). might prefer to create or negotiate an external supply of more special-
Conversely, openness increases the value of the non-open part of a ized (proprietary) assets, their ability to do so (particularly for a new
firm's value proposition (West, 2003). With information goods, firms firm) will be limited by their financial resources (Teece, 1986; Gans
have profited by creating complex offerings (such as software systems and Stern, 2003). New firms both lack the resources to make the invest-
or service-based business models) that are a mixture of tightly linked ment themselves, and (often) to incentivize third parties in their eco-
open and proprietary parts (West and Gallagher, 2006). system to make such investments (West, 2014).
In the MakerBot case, the complementary offerings are separate and Given these financial constraints, we would expect that new firms
less tightly coupled: the MakerBot printers and Thingiverse content (particularly resource-limited ones) will be more open than established
have independent (but complementary) value, because the printer firms with greater resources:
can be used to print its owner's unique designs and the content can be
printed on other printers. With both offerings, the three founders Proposition 1. For firms that require new complementary assets to com-
started with free revealing of their fully open offerings, in large part mercialize their innovation, new firms will be more willing than established
due to an ideology of openness that was derived from that of open firms to source these assets as generic assets.
source software (cf. Dedrick and West, 2008; Henkel, 2009). Under
Proposition 2. Among new firms that require new complementary assets
threat of imitation, they largely eliminated the openness of their
to commercialize their innovation, their willingness to source these assets
as generic assets decreases as the availability of financial capital increases.

14
While resource availability makes it possible to be more proprietary,
MakerBot's failure to share its designs — despite the use of the GPL “copyleft” license by
the RepRap community — prompted questions about the efficacy of such licenses for hard-
a certain form of resources makes it desirable to be more proprietary.
ware, and efforts to create new licenses that would be enforceable for hardware (Levine, Specifically, professional investors (whether angels, independent ven-
2013). ture capitalists or corporate venture capitalists) prefer to invest in

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx 11

firms with high barriers to rivals (e.g., Berkery, 2008). In this case, research site was an extreme exemplar, it had a set of unique character-
MakerBot shifted from open to proprietary hardware 13 months after istics, including the importance of open source hardware and software
receiving $10 million in series A funding. Considering both the utilitari- in the formative stage of MarketBot, and the contributions and inputs
an and normative explanations, we predict from various open source user groups and communities to the making
and the success of MarketBot at a later stage. Nevertheless, the case pro-
Proposition 3. After firms receive venture or other professional invest- vides a rich set of insights that can be generalizable to different settings.
ment, their openness will decrease more rapidly than those that do not re- First, it shows how openness in content can afford a switch from an
ceive such investment. open to a proprietary strategy in hardware. This contributes to our un-
derstanding of how an open source company can grow its open content
to support a proprietary model to compete with rival firms.
7.3. Closing parts
Next, our study identified only a few consumer 3DP manufacturers
that operated their own content communities as a source of comple-
While earlier firms have also pursued a partly open strategy (e.g.
mentary value. Yet, we were unable to identify a competing 3DP manu-
Henkel, 2006; Shah, 2006), MakerBot was unusual both in the degree
facturer that went from open to closed: instead, the general pattern
of its shift from an open to proprietary product strategy, and how rapid-
seems to be that open firms stay open and (as with the industrial
ly it executed that shift. However, any firm that chooses to open parts of
makers) closed firms stay closed. Thus, further research is needed to un-
a complex system (West, 2003) must choose what parts to make (or
derstand more generally how hardware producers choose an open or
source) as open technology and which part will be proprietary technol-
closed strategy, the antecedents and consequences of such choices
ogies. In MakerBot's case, even after one round of venture investment, it
and — in particular — the relationship of those decisions to their com-
couldn't replace all of its open technology with purely proprietary tech-
plementary asset portfolio. More generally, we endorse the call of
nologies. This highlights the general question: for a firm “closing parts”,
Raasch et al. (2009) for more research on how openness strategies de-
which parts should be made proprietary?
veloped from open source software do (or do not) apply to companies
Differing parts of a systemic innovation will be more important for
producing tangible goods.
competitive advantage than others. In a complex system, some compo-
Third, the conditions prompting the trajectory from open to closed
nents are more customer-visible or customer-valued than others (West
can be materially different from the opposite trajectory from closed to
and Gallagher, 2006). Some parts are not feasible to modify because
open. Future research can compare and contrast these diametrically op-
standardized components create strong scale economies or switching
posite trajectories, and notably their impacts on user communities and
costs (Langlois and Robertson, 1992). The overall system performance
the growth and success of enterprises and businesses.
may be hampered by components that lag the rest of the system, tech-
The nature of competition and competitive advantage will change as
nologies that Hughes (1987) labels a “reverse salient”; firms choose to
the industry grows, matures and consolidates. Therefore, many open-
improve these components, eliminating the reverse salients to most
ness strategies that worked for firm entry or early in the industry life
rapidly improve system performance and thus product sales (Davies,
cycle will become obsolete if consumers demand low cost (driving com-
1996; Teece, 2006).
moditization), differentiation (increasing the value of R&D) or interop-
Therefore, we would expect
erability (strengthening network effects and thus the value of market
Proposition 4. When a firm switches from an open to partly closed share or interfirm product standardization).
strategy, ceteris paribus, it will make proprietary those parts most impor- Finally, with 3D printers still relatively rare, it is too early to study
tant to its value proposition. the full impact of the digital representations upon tangible consumer
goods. We expect that such representations will have a different impact
Finally, when firms work with an external community, for a variety of
on consumer use value (and industrial organization) than would infor-
reasons they may find certain types of contributions of value easier to
mation goods (such as entertainment or software) that can be con-
source from communities than others. At one extreme, some contribu-
sumed digitally. We suggest that the adoption (and impact) of copy
tions are so easy to source from external communities — in cases such
machines and (2D) color printers on printed pages might suggest ana-
as peer-to-peer tech support — that they may not even require an organi-
logs for studying how 3D printing will change the value created (and
zation to receive or mediate the provision of the value (Lakhani and von
captured) by manufacturing 3-D objects. Researchers could also use
Hippel, 2003). At the other extreme, technical complexity may make it
the 2D analogs to help identify niches both where 3D printed digital
difficult for external contributors to understand how to contribute or be
content will potentially find the earliest adoption, and also those niches
willing to make the effort to contribute (West and O'Mahony, 2008). Sys-
that are most resistant to disruption and displacement by this new
tems with low technical modularity will be harder for outsiders to under-
technology.
stand (and contribute to) than those with high modularity (MacCormack
et al., 2006). Finally, some forms of knowledge or other external innova-
Acknowledgments
tions are more difficult for a firm to acquire and utilize — such as highly
tacit or context-specific knowledge (Von Hippel, 1994) — and thus we
This work was partly supported by the Centre for Creativity, Regula-
would expect such innovations would be less likely to be externally
tion, Enterprise & Technology [grant number AH/K000179/1]. Earlier
sourced. At the same time, firms are more likely to gain competitive ad-
versions of this paper were presented at RWTH Aachen, Academy of
vantage by maintaining trade secrets around tacit or other complex
Management 2014 and Open and User Innovation 2014. We thank the
parts that are difficult for rivals to imitation (Cohen et al., 2000).
conference participants, the special issue editors and particularly the re-
Overall, we predict
viewers for their helpful feedback; all usual disclaimers apply.
Proposition 5. Firms are most likely to make proprietary those compo-
nents that are most difficult to obtain from external contributors, or most dif- References
ficult to assimilate when provided by external sources.
Alexy, Oliver, George, Gerard, Salter, Ammon J., 2013. Cui bono? The selective revealing of
knowledge and its implications for innovative activity. Acad. Manag. Rev. 38 (2),
270–291.
7.4. Limitations and future research Anderson, Chris, 2012. Makers: The New Industrial Revolution. Crown Business, New
York.
Asvanund, Atip, Clay, Karen, Krishnan, Ramayya, Smith, Michael D., 2004. An empirical
Our research was limited to a single case study within an emerging analysis of network externalities in peer-to-peer music-sharing networks. Inf. Syst.
sector within the 3D printing industry. Although our choice of the Res. 15 (2), 155–174.

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
12 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx

Baichtal, John, 2008. Thingiverse.com Launches A Library of Printable Objects. (Wired. Jarvenpaa, Sirkka L., Lang, Karl R., 2011. Boundary management in online communities:
com, Nov. 20, URL: http://www.wired.com/geekdad/2008/11/thingiversecom/). case studies of the nine inch nails and ccMixter music remix sites. Long Range Plan.
Baker, Ted, Miner, Anne S., Eesley, Dale T., 2003. Improvising firms: bricolage, account giv- 44 (5), 440–457.
ing and improvisational competencies in the founding process. Res. Policy 32 (2), Jeppesen, Lars Bo, Frederiksen, Lars, 2006. Why do users contribute to firm-hosted user
255–276. communities? The case of computer-controlled music instruments. Organ. Sci. 17
Balka, Kerstin, Raasch, Christina, Herstatt, Cornelius, 2009. Open source enters the world (1), 45–63.
of atoms: a statistical analysis of open design. First Monday 14 (11). Jeppesen, Lars Bo, Molin, Måns J., 2003. Consumers as co-developers: learning and inno-
Balka, Kerstin, Raasch, Christina, Herstatt, Cornelius, 2010. How open is open source? — vation outside the firm. Tech. Anal. Strat. Manag. 15 (3), 363–383.
Software and beyond. Creat. Innov. Manag. 19 (3), 248–256. Kogut, Bruce, Metiu, Anca, 2001. Open‐source software development and distributed in-
Berkery, Dermot, 2008. Raising Venture Capital for the Serious Entrepreneur. McGraw- novation. Oxf. Rev. Econ. Policy 17 (2), 248–264.
Hill, New York. Kruth, J.P., Leu, M.C., Nakagawa, T., 1998. Progress in additive manufacturing and rapid
Biggs, John, Kumparak, Greg, 2014. MakerBot Unveils The Replicator Mini, Z18, And A prototyping. CIRP Ann. Manuf. Technol. 47 (2), 525–540.
New Prosumer Replicator. TechCrunch (January 6, URL: http://techcrunch.com/ Lakhani, Karim R., von Hippel, Eric, 2003. How open source software works: “free” user-
2014/01/06/makerbot-unveils-the-replicator-mini-and-the-new-replicator-3d- to-user assistance. Res. Policy 32 (6), 923–943.
printing-platform/). Langlois, Richard N., Robertson, Paul L., 1992. Networks and innovation in a modular sys-
Bilton, Ricardo. 2012. “MakerBot CEO Bre Pettis: ‘Before we make 3D printing cheap, we tem: Lessons from the microcomputer and stereo component industries. Res. Policy
have to make it excellent’”, VentureBeat, Sept. 19, URL: http://venturebeat.com/2012/ 21 (4), 297–313.
09/19/makerbot-ceo-bre-pettis-3d-printing-future/. Levine, Gabriella, 2013. Open Hardware Legal Meetup, NYU, Nov. 11. Open Source
Bock, Simon, West, Joel, Greul, Anne, 2014. “Selective Openness Among User Entrepre- Hardware Association (Dec. 5, URL: http://www.oshwa.org/2013/12/).
neurs,” paper presented at World Open Innovation Conference 2014, Napa, California, MacCormack, Alan, Rusnak, John, Baldwin, Carliss Y., 2006. Exploring the structure of
December 5. complex software designs: an empirical study of open source and proprietary code.
Boudreau, Kevin, 2010. Open platform strategies and innovation: Granting access vs. de- Manag. Sci. 52 (7), 1015–1030.
volving control. Manag. Sci. 56 (10), 1849–1872. MacMillan, Douglas, Burrows, Peter, Ante, Spencer E., 2009. Inside the app economy. Bus.
Bresnahan, Timothy F., Greenstein, Shane, 1999. Technological competition and the struc- Week 347–357 (October 22).
ture of the computer industry. J. Ind. Econ. 47 (1), 1–40 (March). Make, 2013. Ultimate Guide to 3D Printing. MakerMedia (November).
Brooks, Nancy Rivera, 1997. “Models of Efficiency; The rapid-prototyping industry is Maula, Markku V.J., Keil, Thomas, Salmenkaita, Jukka-Pekka, 2006. Open innovation in
changing the way manufacturers create products,” Los Angeles Times, July 13, p. D1. systemic innovation contexts. In: Chesbrough, Henry, Vanhaverbeke, Wim, West,
Brown, Rich. 2012. “Pulling back from open source hardware, MakerBot angers some Joel (Eds.), Open Innovation: Researching a New Paradigm. Oxford University Press,
adherents”, CNET News, Sept 27 URL: http://news.cnet.com/8301-32973_3-57520633- Oxford, pp. 241–257.
296/pulling-back-from-open-source-hardware-makerbot-angers-some-adherents. Morgan, Lorraine, Finnegan, Patrick, 2014. Beyond free software: an exploration of the
Cohen, Wesley M., Nelson, Richard R., Walsh, John P., 2000. Protecting their intellectual business value of strategic open source. J. Strateg. Inf. Syst. 23 (3), 226–238.
assets: appropriability conditions and why US manufacturing firms patent (or not). Murray, Fiona, O'Mahony, Siobhán, 2007. Exploring the foundations of cumulative
Working Paper W7552. National Bureau of Economic Research. innovation: implications for organization science. Organ. Sci. 18 (6), 1006–1021.
Cusumano, Michael A., 2004. The Business of Software. Free Press, New York. Nordhaus, William D., 1969. Invention, Growth, and Welfare: A Theoretical Treatment of
Dahlander, Linus, 2007. Penguin in a new suit: a tale of how de novo entrants emerged to Technological Change. MIT Press, Cambridge, MA.
harness free and open source software communities. Ind. Corp. Chang. 16 (5), Nov, Oded, 2007. What motivates wikipedians? Commun. ACM 50 (11), 60–64.
913–943. O'Mahony, Siobhán, 2003. Guarding the commons: how community managed software
Dahlander, Linus, Magnusson, Mats G., 2005. Relationships between open source software projects protect their work. Res. Policy 32 (7), 1179–1198.
companies and communities: observations from Nordic firms. Res. Policy 34 (4), Peels, Joris, 2010. Shapeways interviews Bre Pettis of MakerBot Industries. (Shapeways.
481–493. com, January 27, URL: http://www.shapeways.com/blog/archives/364-shapeways-
Dahlander, Linus, Magnusson, Mats, 2008. How do firms make use of open source com- interviews-bre-pettis-of-makerbot-industries.html).
munities? Long Range Plan. 41 (6), 629–649. Pendleton, Jennifer. 1994. “Firm Makes Impression in Rapid Prototyping” Los Angeles
Davies, Andrew, 1996. Innovation in large technical systems: the case of telecommunica- Times, Dec 27.
tions. Ind. Corp. Chang. 5 (4), 1143–1180. Perens, Bruce, 1999. The open source definition. In: DiBona, Chris, Ockman, Sam, Stone,
Dedrick, Jason, West, Joel, 2008. Movement ideology vs. user pragmatism in the organiza- Mark (Eds.), Open Sources: Voices from the Open Source Revolution. O'Reilly, Sebas-
tional adoption of open source software. In: Kraemer, Kenneth L., Elliott, Margaret topol, Calif., pp. 171–185.
(Eds.), Computerization Movements and Technology Diffusion: From Mainframes Pettis, Bre, 2010. Open Source Ethics and Dead End Derivatives. (MakerBot.com, March
to Ubiquitous Computing. Information Today, Medford, NJ, pp. 427–452. 25, URL: http://www.makerbot.com/blog/2010/03/25/open-source-ethics-and-
Eisenhardt, Kathleen M., Graebner, Melissa E., 2007. Theory building from cases: opportu- dead-end-derivatives/).
nities and challenges. Acad. Manag. J. 50 (1), 25–32. Pettis, Bre, 2011. All-Star Lineup Invests in MakerBot. (MakerBot.com, August 23, URL:
Eisenmann, Thomas R., Parker, Geoffrey, Van Alstyne, Marshall W., 2011. Platform http://www.makerbot.com/blog/2011/08/23/all-star-lineup-invests-in-makerbot/).
envelopment. Strateg. Manag. J. 32 (12), 1270–1285. Pettis, Bre, 2012. Fixing Misinformation with Information. (MakerBot.com, Sept. 20, URL:
Gans, Joshua S., Stern, Scott, 2003. The product market and the market for “ideas”: com- http://www.makerbot.com/blog/2012/09/20/fixing-misinformation-with-
mercialization strategies for technology entrepreneurs. Res. Policy 32 (2), 333–350. information/).
Gawer, Annabelle, 2009. Platform dynamics and strategies: from products to services. In: Piva, Evila, Rentocchini, Francesco, Rossi‐Lamastra, Cristina, 2012. Is open source software
Gawer, Annabelle (Ed.), Platforms, Markets and Innovation. Elgar, Cheltenham, UK, about innovation? Collaborations with the open source community and innovation per-
pp. 45–76. formance of software entrepreneurial ventures. J. Small Bus. Manag. 50 (2), 340–364.
Gibson, Ian, Rosen, David W., Stucker, Brent, 2010. Additive Manufacturing Technologies: Raasch, Christina, 2011. The sticks and carrots of integrating users into product develop-
Rapid Prototyping to Direct Digital Manufacturing. Springer. ment. Int. J. Technol. Manag. 56 (1), 21–39.
Gruber, Marc, Henkel, Joachim, 2006. New ventures based on open innovation — an empir- Raasch, Christina, Herstatt, Cornelius, Balka, Kerstin, 2009. On the open design of tangible
ical analysis of start-up firms in embedded Linux. Int. J. Technol. Manag. 33 (4), 356–372. goods. R&D Manag. 39 (4), 382–393.
Harhoff, Dietmar, Henkel, Joachim, von Hippel, Eric, 2003. Profiting from voluntary infor- Ragan, Sean Michael, 2012. Bre Pettis @OHS: ‘Challenges of Open Source Consumer
mation spillovers: how users benefit by freely revealing their innovations. Res. Policy Products’. (Makezine.com, Sept. 28, URL: http://makezine.com/2012/09/28/bre-
32 (10), 1753–1769. pettis-ohs-challenges-of-open-source-consumer-products/).
Henkel, Joachim, 2006. Selective revealing in open innovation processes: the case of em- Rogers, Everett M., 1995. Diffusion of Innovations. 4th ed. Free Press, New York.
bedded Linux. Res. Policy 35 (7), 953–969. Rossi, Maria Alessandra, 2006. Decoding the open source puzzle: a survey of theoret-
Henkel, Joachim, 2009. Champions of revealing — the role of open source developers in ical and empirical contributions. In: Bitzer, Jürgen, Schröder, Philip J.H. (Eds.),
commercial firms. Ind. Corp. Chang. 18 (3), 435–471. The Economics of Open Source Software Development. Emerald, Bingley, UK,
Henkel, Joachim, Baldwin, Carliss, Shih, Willy C., 2013. IP modularity: profiting from inno- pp. 15–55.
vation by aligning product architecture with intellectual property. Calif. Manag. Rev. Schau, Hope Jensen, Muniz Jr., Albert M., Arnould, Eric J., 2009. How brand community
55 (4), 65–82. practices create value. J. Mark. 73 (5), 30–51.
Henkel, Joachim, Schöberl, Simone, Alexy, Oliver, 2014. The emergence of openness: how Shah, Sonali K., 2006. Motivation, governance, and the viability of hybrid forms in open
and why firms adopt selective revealing in open innovation. Res. Policy 43 (5), source software development. Manag. Sci. 52 (7), 1000–1014.
879–890. Shane, Scott, 2000. Prior knowledge and the discovery of entrepreneurial opportunities.
Hodgson, Gary, 2012. A History of RepRap Development: Posts from the RepRap Organ. Sci. 11 (4), 448–469.
Development Blog URL: http://reprap.org/wiki/File:A_History_of_RepRap_ Simcoe, Tim, 2006. Open standards and intellectual property rights. In: Chesbrough,
Development.pdf. Henry, Vanhaverbeke, Wim, West, Joel (Eds.), Open Innovation: Researching a New
Howard, Jenifer, 2013. The 100,000th Thing on Thingiverse! (MakerBot.com, June 8, URL: Paradigm.Oxford University Press, Oxford, pp. 161–183.
http://www.makerbot.com/blog/2013/06/08/100000th-thing-on-thingiverse/). Smith, Zach, 2008. Announcing Thingiverse.com, or, ‘Why you need a RepRap machine’.
Hughes, Thomas P., 1987. The evolution of large technological systems. In: Bijker, Wiebe (RepRap.org, Nov. 17, URL: http://blog.reprap.org/2008/11/announcing-thing
E., Hughes, Thomas P., Pinch, Trevor (Eds.), The Social Construction of Technological iversecom.html).
Systems. Cambridge, Mass., MIT Press. Smith, Zach, 2011. Thingiverse. In: van Abel, Bas, Evers, Lucas, Klaassen, Roel, Troxler,
Jacobs, Paul F., 1992. Rapid Prototyping and Manufacturing: Fundamentals of Peter (Eds.), Open Design Now. BIS, Amsterdam, p. 235.
Stereolithography. ASME Press, New York. Smith, Zach, 2012. MakerBot vs. Open Source — A Founder Perspective weblog, Sept. 21,
Jacobs, Paul F., 1996. Stereolithography and Other RP&M Technologies: From Rapid URL: http://www.hoektronics.com/2012/09/21/makerbot-and-open-source-a-
Prototyping to Rapid Tooling. ASME Press, New York. founder-perspective/.

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
 J. West, G. Kuk / Technological Forecasting & Social Change xxx (2015) xxx–xxx 13

Spaeth, Sebastian, Stuermer, Matthias, Von Krogh, Georg, 2010. Enabling Knowledge West, Joel, 2003. How open is open enough? Melding proprietary and open source
Creation through Outsiders: Towards a Push Model of Open Innovation. Int. platform strategies. Res. Policy 32 (7), 1259–1285.
J. Technol. Manag. 52 (3-4), 411–431. West, Joel, 2006. Does appropriability enable or retard open innovation? In: Chesbrough,
Stallman, Richard M., 2002. Free Software, Free Society: Selected Essays of Richard M. Henry, Vanhaverbeke, Wim, West, Joel (Eds.), Open Innovation: Researching a New
Stallman. GNU Press, Boston. Paradigm. Oxford University Press, Oxford, pp. 109–133
Stern, Becky, 2010. “The Revolution Will Be Squirted”, Make 21. pp. 46–51 (URL: http:// West, Joel, 2014. Challenges of funding open innovation platforms: lessons from Symbian
makezine.com/magazine/make-21/the-revolution-will-be-squirted/). Ltd. In: Chesbrough, Henry, Vanhaverbeke, Wim, West, Joel (Eds.), New Frontiers in
Stratasys, 2013. Stratasys and MakerBot Complete Merger press release, August 15, URL: Open Innovation. Oxford University Press, Oxford, pp. 71–93.
http://investors.stratasys.com/releasedetail.cfm?ReleaseID=785515. West, Joel, Gallagher, Scott, 2006. Challenges of open innovation: the paradox of firm in-
Teece, David J., 1986. Profiting from technological innovation: implications for integra- vestment in open‐source software. R&D Manag. 36 (3), 319–331.
tion, collaboration, licensing and public policy. Res. Policy 15 (6), 285–305. West, Joel, O'Mahony, Siobhán, 2008. The role of participation architecture in growing
Teece, David J., 2006. Reflections on ‘profiting from innovation’. Res. Policy 35 (8), sponsored open source communities. Ind. Innov. 15 (2), 145–168.
1131–1146. Wohlers, Terry, 1992. The world of rapid prototyping. Proceedings of the Fourth
Teigland, Robin, Gangi, Di, Paul, M., Flåten, Björn-Tore, Giovacchini, Elia, Pastorino, International Conference On Desktop Manufacturing, Sept. 24–25, San Jose, California
Nicholas, 2014. Balancing on a tightrope: managing the boundaries of a firm- (URL: http://wohlersassociates.com/mr.html).
sponsored OSS community and its impact on innovation and absorptive capacity. Wohlers Associates, 2013). Wohlers Report. Ft. Collins, Colo.: Wohlers Associates.
Inf. Organ. 24 (1), 25–47. Wohlers Associates, 2014). Wohlers Report. Ft. Collins, Colo.: Wohlers Associates.
van Burg, Elco, Berends, Hans, van Raaij, Erik M., 2014. Framing and interorganizational Yin, Robert K., 2003. Case Study Research: Design and Methods. 3rd ed. Sage.
knowledge transfer: a process study of collaborative innovation in the aircraft indus-
try. J. Manag. Stud. 51 (3), 349–378.
Joel West is a professor of innovation and entrepreneurship at the Keck Graduate Institute
Varian, Hal R., 1998. “Markets for Information Goods,” Working Paper. University of
of Applied Life Sciences, one of the seven Claremont Colleges. His research examines the
California, Berkeley (October 16).
strategic use of openness in areas such as open innovation, open source software, and
Von Hippel, Eric, 1994. ‘Sticky information’ and the locus of problem solving: implications
open standards. He is the author of the , OpenInnovation.net weblog and co-editor of
for innovation. Manag. Sci. 40 (4), 429–439.
two books on open innovation.
Von Hippel, Eric, von Krogh, Georg, 2003. Open source software and the ‘private-collec-
tive’ innovation model: issues for organization science. Organ. Sci. 14 (2), 209–223.
George Kuk is professor of innovation and entrepreneurship at the Nottingham Business
Weber, Christopher L., Peña, Vanessa, Micali, Maxwell K., Yglesias, Elmer, Rood, Sally A.,
School. His research focuses on open innovation and strategy in software, data, design and
Scott, Justin A., Lal, Bhavya, 2013. The Role of the National Science Foundation in
platform within the creative industries. He examines how companies can attract creative
the Origin and Evolution of Additive Manufacturing in the United States, IDA Science
resources for digital and service innovation.
& Technology Policy Institute. Institute for Defense Analysis, Washington, DC.
Weinberg, Michael, 2013. Stratasys Sues Afinia: Ramifications for the Desktop 3D Printing
Industry. (Makezine.com, Nov. 27, URL: http://makezine.com/2013/11/27/stratasys-
sues-afinia-ramifications-for-the-desktop-3d-printing-industry/).

Please cite this article as: West, J., Kuk, G., The complementarity of openness: How MakerBot leveraged Thingiverse in 3D printing, Technol.
Forecast. Soc. Change (2015), http://dx.doi.org/10.1016/j.techfore.2015.07.025
 Staying at the Technology Forefront by Spotting the Reverse Salient: The Case
of Digital Video Broadcasting

O. Dedehayir1, A. Hornsby2
1
Department of Industrial Management, Tampere University of Technology, Tampere, Finland
2
Department of Signal Processing, Tampere University of Technology, Tampere, Finland
([email protected])

Abstract – This paper traces the historical evolution of reasons. Firstly, the DVB system is undergoing rapid
DVB (digital video broadcasting) terrestrial (DVB-T) and technological evolution (particularly in response to
handheld (DVB-H) technologies subsequent to the resolution alternative technologies, such as broadband, which
of reverse salients – components of technological systems compete in the provision of digital video to end-users) and
which trail in technological performance and hinder
therefore offers plentiful and recent developmental
potential development of the system at large. Early
identification and resolution of reverse salients is an examples for analysis. Secondly, technological
important consideration for organizations, as providing development within the DVB system is governed to a
technological solutions with respect to reverse salients often large extent by the overseeing “DVB” group (a
derives competitive advantage through the establishment of consortium of approximately 300 industry players), which
standards and patents. We argue that organizations exclusive is an influential social component of the system both
of the DVB group (industry-led consortium overseeing the motivating and guiding technological development.
evolution of digital television in Europe) can stay at the Technological evolution as a consequence of the role of a
forefront of technological development by identifying current central governing body is not only illuminating but
as well as forecasting future reverse salients from available
remains relatively under-studied, particularly in the
information such as standards and patent registries.
context of emerging technologies.
Keywords – Digital video broadcasting, handheld, reverse The purpose of this paper is to study the evolution of
salient, technological system, terrestrial DVB technology from a systemic perspective and employ
the reverse salient concept in elaborating the underlying
dynamics of this evolution. We begin our paper by
I. INTRODUCTION providing theoretical background pertaining to
technological systems and the reverse salient concept. In
Technological systems are hierarchically nested, the following section we describe our proposed
comprising both technical components such as physical methodology for studying the DVB technological system
artifacts and software, as well as social components such evolution through an event analysis. Next, we analyze the
as the firms which develop the aforementioned elements temporal evolution of DVB technology, focusing on the
[1-3]. The development of the technological system at evolution from satellite (DVB-S) to terrestrial (DVB-T),
large takes place as a consequence of enhancing the and from terrestrial to handheld (DVB-H) technology,
performance of nested systemic elements, which is the making reference to notable reverse salients along this
end subsequent to the strategic and competitive means timeline. We subsequently discuss the findings of our
deployed by actor firms inside the system. A key driver of analysis and propose strategy formulation considerations
development within the technological system is the particularly for firms external to the DVB group.
identification and correction of a systemic component
which is deficient in technological performance, known as
the reverse salient. The reverse salient hinders the II. THE REVERSE SALIENT
technological performance delivery of the overall system
to end-users and therefore becomes a focal point of We employ a systemic view of technology in
system actors, such as firms providing technological exploring the evolution of digital video broadcasting as
innovations, to resolve the difficulties and ensure desired such view offers the required means to examine, in
development. particular, the dynamics (i.e. the causes and effects) of
In this paper we explore the digital video technological evolution. For our purposes, we describe
broadcasting (DVB) technological system and its technological systems as messy and complex, comprising
temporal development through the resolution of reverse both technical as well as social components [1, 2]. Two
salients. DVB is a leading technological standard, traits of technological systems are of particular interest in
amongst several standards in the world, which provides our analysis. The first is that technological systems are
digital video broadcasting to stationary, nomadic, as well hierarchically structured [4], comprising system, sub-
as mobile devices. The DVB technological system offers system, and component level technologies [3]. For
a highly interesting case to explore at least for two instance, taking a systemic view of DVB technology, we
can identify the digital video broadcasting network at the

978-1-4244-2330-9/08/$25.00
Authorized licensed use limited ©2008
to: HongIEEE 189 on August 20,2023 at 08:56:58 UTC from IEEE Xplore.
Kong Metropolitan University. Downloaded Restrictions apply.
 Proceedings of the 2008 IEEE ICMIT

system level, containing “social” first-order sub-systems the potential is fulfilled by the trajectory reaching its phase
such as broadcast service providers, innovating firms, of maturity or by the “technological paradigm” [8] being
research laboratories (e.g. universities), and standards superseded.
bodies, as well as “technical” first-order sub-systems such Components of the system which evolve at the
as the broadcast transmission infrastructure, and mobile necessary pace and maintain evenness contribute
handheld devices (e.g. mobile telephones). Continuing positively to the collective progress of that system.
down the hierarchy, we can focus on the mobile telephone Conversely, a technological element at any level of the
technology comprising second-order sub-systems such as system which does not develop sufficiently prevents the
the CPU (central processing unit), receiver, battery, technology system achieving desired development: these
operating system, and a display screen. These problematic members are termed reverse salients [1, 4].
technological sub-systems can in turn be subdivided into Literally, a reverse salient is the inverse of the forward
smaller constituents, where sequential subdivision will protrusion along an object’s profile or “a line of battle”
eventually lead to the component technology level of the [4]. With respect to technological system development,
overall system (e.g. transistors or even silicon granules). reverse salients refer to the elements of that system which
The second important characteristic of technological have strayed behind the advancing performance frontier of
systems is their tendency to evolve subsequent to the system [1]. They are therefore the underperforming
fulfillment of systemic objectives and resolution of components which hamper the progress or which prevent
problems [2, 4]. This trait is in fact essentially the fulfillment of potential development of the collective
technological where technology is defined as a means to system. The reverse salient needs to be resolved for
an end, whereby ends are accomplished through the continuation of technological system development, and
resolution of problems and reordering of the material therefore provides an incentive for system stakeholders
world [5]. Hence, technological systems are inherently such as innovating firms, university laboratories, and
compelled to evolve as a result of their own reason for standards bodies, to congregate around the retardant and
existence; to provide ends to paradigmatic problems. strive to remove it through innovations.
Augmenting the two idiosyncrasies of technological Historical developments in a variety of technological
systems considered above, we posit that the evolution of systems have been used in literature to illustrate reverse
technology at any level of the system is brought about by salience. Hughes (1983) has most famously given account
the inherent need of problem resolution in line with system of Thomas Edison’s direct-current electric system and its
objectives. In this light, the problem in question may be development towards the objective of supplying electricity
resolved through technological advances in different sub- within a defined region of distribution. Perhaps the most
systems or components in addition to where the problem is notable limitation (i.e. reverse salient) of this system’s
initially brought to attention. For instance, the provision of development was its low voltage transmission distance,
digital video broadcasting to handheld devices – a problem dictated by the cost of distributing electricity beyond a
which is at the system level of the DVB technological certain range. To reduce costs, Edison introduced a three-
system hierarchy – was enabled, for the most part, through wire system to replace the previously installed two-wire
technical developments at lower-order sub-systems. The alternative, and trialed different configuration of
consequent provision of digital video broadcasting to generators as well as the usage of storage batteries. While
handheld devices was the outcome of certain dynamics of these had a positive impact they did not remove the
system evolution; firstly the identification of limiting reverse salient completely. Inevitably, the satisfactory
technical or social system components which prohibit resolution of the problem of costly transmission and
progress towards an objective, followed by the distribution of low voltage electricity was provided by the
engagement of system actors to remove the limiting radical invention of the alternating-current system [1].
problem through social or technical solutions. Other authors have also offered their accounts based
This dynamics is closely resembled in Dahmén’s on the analysis of different technological systems.
notion of “development blocks” [6], that are a “sequence MacKenzie [9] has discovered the gyroscope sub-system
of complementarities” or a cluster of network elements, as a reverse salient in the ballistic missile technological
such as firms or technologies, which undergo development development, where the systemic objective has been to
as a result of a series of structural tensions [6]. These increase missile accuracy. With the objective of
tensions are outcomes of disequilibria between proliferating mobile music throughout the end-user
interdependent elements, such that the technological market, Takeishi and Lee [10] have argued that music
performance of some system elements are lower than the copyright managing institutions have been the reverse
performance of others, reminiscent of Rosenberg’s salient in the evolution of the mobile music technology
“technological disequilibrium” notion [7]. The system in Japan and Korea. According to Mulder and Knot
technological performance differential between system [11], the development of the PVC (polyvinyl chloride)
components generates a need for technological plastic technology system has been sequentially hampered
development, leading to evolutionary momentum within by several reverse salients, including: difficulty in
the development block. The subsequent technological processing PVC material, poor quality of manufactured
progress follows a particular line or “technological products, health concerns for individuals exposed to
trajectory”, to use Dosi’s [8] terminology, and ends when

190 on August 20,2023 at 08:56:58 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Hong Kong Metropolitan University. Downloaded
 Proceedings of the 2008 IEEE ICMIT

effluent from PVC manufacturing facilities, and finally the development of digital television in Europe. The DVB
carcinogenetic nature of vinyl chloride. group published its first DVB standard in December 1993,
named DVB-S (EN 300 421) for Digital Video
Broadcasting over Satellite. DVB-S is the original Digital
III. METHODOLOGY Video Broadcasting forward error coding and modulation
standard for satellite television and is nowadays serving
We accessed the DVB organization’s website every continent of the world.
(www.dvb.org) to gather information concerning the Following the success of DVB-S, in March 1994, the
evolution of DVB technologies and the significant events DVB group published a new standard, DVB-C (EN 300
connected with this evolution. In particular we read 429), to address the cable distribution systems. Satellite
through the publications available on this website, and Cable distribution systems being addressed, the DVB
although we did not limit ourselves to these publications. group worked on the more traditional terrestrial
We selected the DVB organization’s website due to the distribution system and published the DVB-T standards in
availability of all recent and related literature concerning December 1995 (EN 300 744) which is now used in 66
our research. Additionally, we accessed the DVB-H countries worldwide. Furthermore, in December 2000,
organization’s website (www.dvb-h.org) providing DVB began to discuss work on mobile devices,
literature more specific to handheld DVB technology, as particularly handheld mobile devices, planting the seed for
well as publications referring to DVB technology in what would become, in November 2004, the DVB-H
general from the IEEE Xplore database. We analyzed the standard.
DVB system evolution through event analysis based on The DVB-T standard is now over a decade old and has
the information obtained from these aforementioned set the basis for most of the other digital terrestrial
literature sources. television implementations around the world. The success
of digital television, and in particular DVB-T, has led to
the full transition from Analog to Digital, called the
IV. THE EVOLUTION OF DIGITAL VIDEO Analogue Switch Off (ASO), enabling the new digital
BROADCASTING AND REVERSE SALIENTS distribution to be a reality. However, to fully exploit the
commercial opportunities that may follow ASO, the DVB
The 1980s set the beginning of the digital media group started discussions on the evolution of DVB-T to
revolution. Until then, media relied primarily on analog provide advanced technology and techniques that would
technology which refers to the process of transforming a allow new successful services. In April 2007, the DVB
signal (mainly audio or video) into electronic pulses. group published the DVB-T2 call for technology,
Digital technology on the other hand, translates those same requesting new submissions for technologies which could
signals into binary format represented by a series of “0” meet the requirements for a second generation terrestrial
and “1”s. The advance in digital signal processing and in transmission system
electronic integrated circuits made the conversion from Whether through a call for technology or otherwise,
analog signals into digital signals possible, using analog- the development of digital video broadcasting standards
to-digital converters (ADC), and vice versa, digital-to- have been brought about through the identification and in
analog converters (DAC). Consequently, digital turn the resolution of system inhibiting reverse salients. In
technology became more popular and cheaper, setting the the following paragraphs we illustrate the evolution of
beginning of a transformation that would change the old these standards by considering some of the reverse salients
media into revolutionary new digital media. which were obstacles in the development of the DVB-T
The new digital era allowed the emergence of new and DVB-H systems, respectively.
digital media devices such as computers, televisions, video
cameras, mobile phones, and modems, along with new
media distribution systems such as satellites, cables, CDs, A. The Development of DVB-T from DVB-S Origins
hard disc drives, and later on, DVDs. These new digital
systems altered the meaning of geographic distances, In the DVB-S system the video, audio, and data are
permitting a significant increase of media content and inserted into a fixed length MPEG-2 Transport Stream
permitting faster communications, and subsequently forming the actual payload of the TV system. This stream,
providing a completely new user experience. or TV payload, is then modulated using a single QPSK
In 1991, to answer the new consumer demand for (Quadrature Phase Shift Keying) carrier modulation
digital media, the European Launching Group (ELG), a scheme with additional tools for channel coding and error
group of European broadcasters and equipment protection. QPSK is a phase modulation algorithm, where
manufacturers, came together with the mission of the phase of the carrier wave is modulated to encode bits
overseeing the evolution of digital television, and in of digital information in each phase change. DVB-S was
particular, its distribution in Europe. Two years after, in primarily designed for the distribution of SDTV (Standard
September 1993, the expanded ELG group renamed itself Definition Television) services.
as the Digital Video Broadcasting (DVB) group, As opposed to DVB-S, the digital terrestrial television
concurrently becoming the major driving force in the system, DVB-T, is more complex as it was designed with

191 on August 20,2023 at 08:56:58 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Hong Kong Metropolitan University. Downloaded
 Proceedings of the 2008 IEEE ICMIT

the intention of coping with a different propagation In our illustration we have focused on one such
environment mostly due to the effect of multipaths. technical reverse salient; the receiver’s inability to adapt
Multipath is a propagation phenomenon resulting from its decoding to cater for different signaling methods. Once
radio signals reaching the receiving antenna using two or this reverse salient was resolved through the development
more paths. It is mainly caused by the reflection and of the OFDM method, the technological performance of
refraction from the atmosphere or by objects which lie on the digital video broadcasting system in its entirety was
the signal paths, such as buildings, trees or cars. To cope enhanced to deliver end-users the potential terrestrial
with this relatively more hostile environment, some transmission.
technical changes were required in the development of the
DVB-T system. B. The Development of DVB-H from DVB-T Origins
The DVB-T system was developed to allow the
receiver to adapt its decoding according to a signaling DVB-H is a standard for mobile television based on
method. In fact, the key element of DVB-T transmission is the existing DVB-T terrestrial digital broadcasting
the use of Orthogonal Frequency-Division Multiplexing standard. It enhances the more traditional digital
(OFDM) which uses two modes, namely 2K and 8K television broadcast standard with features required to
modes. The 8K mode is more robust and allows more satisfy requirements specific to handheld devices;
multipath protection, but the 2K mode offers more mobility, smaller screens and antennas, indoor coverage,
Doppler advantages where the receiver is in motion. and reliance on battery power.
OFDM uses a large number of closely-spaced orthogonal The enhancements provided by DVB-H cover a wide
sub-carriers to carry digital data. Sub-carriers overlap in range of technical and service issues. In the DVB
frequency with one another, but do not interfere with each transmission scheme shown in Fig. 1 the DVB-H system
other. Each sub-carrier is modulated with a conventional enhancements are marked in blue. Firstly, on the signal
modulation scheme such as quadrature amplitude transmission level there is an additional COFDM (coded
modulation (QAM) at a low symbol rate, maintaining data OFDM) modulation mode with four thousand carriers
rates similar to conventional single-carrier modulation called 4K and an extension to the signaling elements.
schemes in the same bandwidth. The main advantage of These signaling elements are so-called TPS (Transmission
OFDM over single-carrier schemes is its ability to cope Parameters Signaling) bits indicating DVB-H basic
with severe channel conditions, such as that of the transmission characteristics. The carrier modulation
terrestrial case. options in the DVB-T and DVB-H are the same (QPSK,
From our systemic perspective, we consider the 16QAM, 64QAM). On the transport stream level the
development of the DVB-T technology to be a significant, DVB-H system uses the MPEG-2 packet structure
in other words “radical” change [12]. The emergence of identical to DVB-T. Prior to modulation the transport
the DVB-T technology occurs at the highest level of the stream is coded for error-correction, where the DVB-H
system hierarchy, even though the vital technical standard provides for a new code rate (9/10) which is of
developments necessary for this change – such as the use rather minor importance. Thus, on the transmission level,
of OFDM – took place at lower sub-systems of the digital DVB-H is a superset of DVB-T but practically both
video broadcasting technological system. The dynamics of systems are largely compatible since the DVB-H may
technological evolution in this context commenced with practically be realized with DVB-T parameters.
the identification of a reverse salient at the high echelons The most essential new features of DVB-H are in the
of the system hierarchy – namely the lack of robustness of organization and sending of data. The first of these is
DVB-S technology to cope with terrestrial environments. “time slicing” which allows data to be sent in bursts.
The transmission system being inadequate, the full scale Transmitting in burst significantly contributes to the
broadcasting of digital video to end-users through a reduction of battery power consumption on the terminal
terrestrial network could therefore not reach its potential, side, allowing the terminal receiver to be switched off after
hindered by technical reverse salients at lower sub-system loading its input buffer with a burst of data.
levels.

Fig. 1. The DVB-H Transmission System.

192 on August 20,2023 at 08:56:58 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Hong Kong Metropolitan University. Downloaded
 Proceedings of the 2008 IEEE ICMIT

Indeed, service multiplexing in DVB-H is made in a The organizations (here we refer to organizations
time-division multiplex. Transmission of data for a collectively as firms, research laboratories, universities
particular service is not done in a continuous manner but and so on) belonging to the Digital Video Broadcasting
rather organized in periodical bursts with an off-time (DVB) group are in an advantageous position in
period in between. The receiver, synchronized to the burst, establishing technological standards and trajectories of
can in turn power off during the off-time period and development. As members of the DVB group, these
therefore save power. organizations have immediate access to information
We see the development of the DVB-H system as an concerning not only current developments but also
extension of the already existing DVB-T network, and potential, future developments within the DVB
therefore classify it as an “incremental” change [12] at the technological system. The benefit of early access to
system level rather than a radical one. However, we have information means that member organizations can react
observed some “architectural” [12] and radical changes at quicker to existing technological trends and engage in
lower levels of the system hierarchy which have brought appropriate technological strategies. By spotting reverse
about the system level development. These sub-system salients early, they can deploy appropriate means to
technical changes have been necessitated to overcome resolve the problems and promote their own standards
certain reverse salients. For example, the time slicing ahead of other organizations. Conversely, those
development is a radical technical response to the power organizations which are external to the DVB group may
consumption reverse salient in handheld devices. become aware of reverse salients with some delay and
therefore be placed in a relatively disadvantageous
position concerning the direction of technological
V. DISCUSSION development. Moreover, these organizations may be
compelled to wait for the publication of a “call for
The identification of a reverse salient in a developing technology” (such as the call for technology published in
technological system draws innovating firms to a focal April 2007 for the DVB-T2 system), summoning firms
point as they endeavor to alleviate structural tensions [6]. both internal and external of the DVB group to provide
However, the resolution of persisting prohibitive problems solutions to technological problems – in other words a
are manifested in circumstances often marked by rivalry public call for the resolution of reverse salients – before
between alternative solutions with respect to reverse engaging in the development of technological solutions.
salients - a battle of solutions therefore ensues (e.g. the so- To become more competitive organizations external to
called battle of the systems which took place between the the DVB group must actively seek information concerning
direct current and alternating current electric systems [1]) - trends in DVB technology. While calls for technology
leading eventually to the selection of one alternative provide valuable sources of information, additional
solution over others. At closer inspection we encounter the information related to future technological developments
incentive of different actors to establish their solution as a can also be obtained from other sources. We propose four
winning standard irrespective of the relative superiority or generic sources of information related to technological
inferiority of the offered solution in question (relative changes, although not all are accessible to organizations
superiority or inferiority is typically revealed in external to the DVB group. Fig. 2 illustrates our proposed
retrospect). Examples in history abound – often referred to sources of information as well as their sequential
as “technological lock-in” [13, 14] – for example the connections.
electric refrigerator’s victory over the gas refrigerator
despite comparable inferiorities [15].
Once a technological dominant design [16, 17] is
established as a consequence of reverse salient
identification and resolution, subsequent technological
developments follow specific trajectories within the
bounds of the technological paradigm [8]. Research and
development activities as well as setting of technical
standards support technological progress along these
trajectories. Considering the case of the DVB
technological system, innovating firms have historically
strived to push their own standards and establish winning
solutions. Here, the incentive for pushing a technological
solution is that victory in the standardization battle lends
the victorious firm room to direct future technological
development through the setting of standards and filing of Fig. 2. The sources of information connected with technological
patents, which translates into business success. The early development.
identification of a reverse salient therefore emerges as a
significant factor in the strategic consideration of While the call for technology provides accessible
organizations. information to organizations outside of the DVB group,

193 on August 20,2023 at 08:56:58 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Hong Kong Metropolitan University. Downloaded
 Proceedings of the 2008 IEEE ICMIT

the reverse salients upon which the call for technology is similar strategic considerations are valid also in different
developed may not be directly accessible. It is likely that technological domains.
organizations inside the group identify the reverse salient
and are therefore aware of it in advance of the call for
technology. The call for technology encourages REFERENCES
organizations to engage in activities such as research and
development, patent filing, and proposition of standards. [1] T. P. Hughes, Networks of Power: Electrification in Western
Information regarding these actions however may not Society, 1880-1930. USA: The John Hopkins University
Press, 1983.
become accessible until their official registration as
[2] T. P. Hughes, "Technological Momentum," in Does
standards (with institutions such as ETSI – European Technology Drive History?, M. R. Smith and L. Marx, Eds.
Telecommunications Standards Institute) and as patents. USA: The MIT Press, 1994, pp. 101-113.
Registered standards and patents provide valuable [3] J. P. Murmann and K. Frenken, "Toward a systematic
information concerning existing technological trajectories framework for research on dominant designs, technological
in digital video broadcasting, illuminating existing innovations, and industrial change. ," Research Policy, vol.
technological gaps (i.e. reverse salients) which present 35, pp. 925-952, 2006.
focal areas of future innovations. Assessment of end-user [4] T. P. Hughes, "The Evolution of Large Technological
needs can additionally enhance such forecasting. In Systems," in The Social Construction of Technological
Systems, W. E. Bijker, T. P. Hughes, and T. P. Pinch, Eds.
summary, the early identification of possible reverse USA: The MIT Press, 1987, pp. 51-82.
salients through active search is a necessary [5] M. Heidegger, The Question Concerning Technology and
entrepreneurial act of organizations exclusive of the DVB Other Essays. New York: Harper & Row, Publishers, Inc.,
group, which can bolster their competitiveness in 1977.
providing technological solutions. [6] E. Dahmén, "Developmen blocks in industrial economics,"
in Industrial dynamics, B. Carlsson, Ed. Boston: Kluwer
Academic Publishers, 1989, pp. 332.
VI. CONCLUSIONS [7] N. Rosenberg, Perspectives on Technology. Cambridge:
Cambridge University Press, 1976.
In this paper we have analyzed the technological [8] G. Dosi, "Technological paradigms and technological
evolution of the DVB-T (terrestrial) and DVB-H trajectories," Research Policy, vol. 11, pp. 147-162, 1982.
[9] D. MacKenzie, "Missile accuracy: a case study in the social
(handheld) digital video broadcasting standards through a
processes of technological change.," in The Social
systemic perspective. In our analysis we have illustrated Construction of Technological Systems., W. E. Bijker, T. P.
the dynamics of such technological evolution through the Hughes, and T. J. Pinch, Eds. USA: The MIT Press, 1987,
sequence of identification and resolution of system pp. 195-222.
performance hindering components referred to as reverse [10] A. Takeishi and K.-J. Lee, "Mobile music business in Japan
salients. Taking into account the incentives of and Korea: copyright management institutions as a reverse
organizations which provide technological solutions and salient.," Journal of Strategic Information Systems, vol. 14,
sponsor system development, we asserted the early pp. 291-306, 2005.
identification of reverse salients as an important strategic [11] K. Mulder and M. Knot, "PVC plastic: a history of systems
development and entrenchment.," Technology in Society,
consideration. We believe that organizations stand to
vol. 23, pp. 265-286, 2001.
benefit from establishing winning solutions to [12] R. M. Henderson and K. B. Clark, "Architectural
technological problems as these typically translate into Innovation: The Reconfiguration of Existing Product
registered standards and patents as well as subsequent Technologies and the Failure of Established Firms,"
financial gains. The competitive nature of emerging Administrative Science Quarterly, vol. 35, pp. 9-30, 1990.
technologies, such as digital video broadcasting, renders [13] W. B. Arthur, "Competing Technologies, Increasing
the early recognition of reverse salients and the focus of Returns, and Lock-in by Historical Events," The Economic
innovation activity as highly significant factors. In this Journal, vol. 99, pp. 116-131, 1989.
light, organizations which are members of the “DVB” [14] P. A. David, "Clio and the Economics of QWERTY," The
American Economic Review, vol. 75, pp. 332-337, 1985.
group (the industry-led consortium overseeing the
[15] C. Bissell and S. Bennett, "The Role of the History of
evolution of digital television in Europe) have the Technology in the Engineering Curriculum.," European
advantage of accessing information concerning Journal of Engineering Education, vol. 22, pp. 267-275,
technological developments ahead of organizations outside 1997.
of this group. Notwithstanding, organizations exclusive of [16] P. Anderson and M. L. Tushman, "Technological
the DVB group can stay at the forefront of technological discontinuities and dominan designs: A cyclical model of
development by employing an active rather than a passive technological change," Administrative Science Quarterly,
technology strategy. In addition to the calls for vol. 35, pp. 604-633, 1990.
technology, these organizations can identify current as [17] M. L. Tushman and P. Anderson, "Technological
discontinuities and organizational environments,"
well as forecast future reverse salients from available
Administrative Science Quarterly, vol. 31, pp. 439-465,
information such as standard and patent registries. Finally, 1986.
while we have explored the evolution of the DVB
technology standards in this article, we propose that

194 on August 20,2023 at 08:56:58 UTC from IEEE Xplore. Restrictions apply.
Authorized licensed use limited to: Hong Kong Metropolitan University. Downloaded
 INTERNATIONAL CONFERENCE ON ENGINEERING DESIGN, ICED19
5-8 AUGUST 2019, DELFT, THE NETHERLANDS

A COMPARISON OF CONTEMPORARY PROTOTYPING

METHODS

Coutts, Euan Ross (1,2); Wodehouse, Andrew (2); Robertson, Jason (2)

1: University of Canterbury; 2: University of Strathclyde

ABSTRACT
Prototypes are a common feature of many product design and development endeavours. An ever
widening range of prototyping options are available to designers and engineers. May particular options
be superior to others, or more appropriate for particular endeavours? This paper reviews current
literature on the nature of what constitutes a prototype and the benefits they offer to the discipline. They
principally facilitate communication, aid learning, help gain and provide feedback, inform decision
making and generally provide superior design outcomes. In order to determine if any particular manner
of prototype is preferable for achieving these benefits a comparative study of some of the contemporary
prototyping methods is subsequently conducted: A 3D printed prototype (physical prototype), a CAD
prototype (represented using a computer monitor), an augmented reality prototype (represented using a
tablet device) and a virtual reality prototype (represented using a stereo projector and polarised glasses).
The results indicate that while all provide benefits, overall the physical prototype performs best and the
augmented reality prototype performs most poorly.

Keywords: Virtual reality, 3D printing, Product modelling / models, Prototyping

Contact:
Coutts, Euan Ross
University of Canterbury, New Zealand
School of Product Design
New Zealand
[email protected]

Cite this article: Coutts, E.R., Wodehouse, A., Robertson, J. (2019) ‘A Comparison of Contemporary Prototyping
Methods’, in Proceedings of the 22nd International Conference on Engineering Design (ICED19), Delft, The
Netherlands, 5-8 August 2019. DOI:10.1017/dsi.2019.137

ICED19 1313
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
 1 INTRODUCTION
Prototypes are a common feature of many product design and development endeavours. As technology
continues to evolve, the formats and types of prototype available to designers has widened considerably.
In recent years, the emergence of digital technologies has presented options in additive manufacture,
augmented reality and virtual reality that can accelerate and enhance design decision making. However,
the nature and benefits of these are not yet well understood - how should these differing formats of
prototypes be deployed, presented and utilised? This paper advances understanding of what constitutes a
prototype and progresses to compare several prototyping options through a practical study. This paper
advances understanding of the benefits of different prototyping formats, and evaluates communication
performance and designer preferences through a comparative study.
Terminology in this domain - prototype, model, mock-up etc. - are often used interchangeably. The
Academic Press Dictionary of Science and Technology gives the definition of prototype as “an early or
original form… (in engineering) a full-scale model of a structure or piece of equipment used in
evaluating form, design, fit, and performance” (Morris, 2018). Similarly, the definition given for a mock-
up is given as “a scale model, often full size, of a structure, apparatus or vehicle; used for study, training,
or testing and to determine if the apparatus can be manufactured easily and economically.” Both
definitions describe a scale model that is used for analysis and testing, this leads to the conclusion that
both terms can be used to describe the same thing. This also falls into line with both literature (Wang,
2002) and real world practises that suggests they are commonly used as interchangeable terms.
Lauff, Kotys-Schwartz and Rentschler (2018) ask the question “what is a prototype?” within the context
of the design process, stating “A prototype is a physical or digital embodiment of critical elements of the
intended design, and an iterative tool to enhance communication, enable learning, and inform decision-
making at any point in the design process”. This definition was derived through ethnographic
observations and inductive analysis of three design companies and states three key aspects:
That a prototype is an embodiment of a design.
It is a tool to improve communication, learning and decision making.
It can be used at any stage in the design process.
However, Houde and Hill (1997) raise the observation that the term “prototype” is ambiguous, meaning
different things to different people, for example a designer may call a moulded foam object a prototype
and a programmer would call a test program a prototype. Additionally, an example asking “is a brick a
prototype?” is given, the answer to which “depends on what it is used for, if it is used to represent the
weight and scale of a future artefact then it must be a prototype”. This illustrates that a prototype does
not rely on how it is manifest, what defines a prototype is that it is used to explore or demonstrate some
aspect of a future artefact, by such logic any representation of a design may arguably be considered a
prototype. This illustrates that the fundamental function of a prototype is to communicate design intent,
and to this extent any suitable representation of a design may arguably considered a prototype. Indeed,
other research has included sketches and illustrations within the working definition. This research,
however, focuses on 3D representations that provide some form of interaction in the demonstration and
communication of a design concept.
In this vein, a prototype is often considered as a means to answer a question, such as proving an
application of a technology, or clearly communicating some key functionality of a design, frequently
referred to as “proof of concept” (Ullman, 2010). This work’s definition of a prototype is aligned with
Lauff et al. (2018), Lidwell et al. (2003) and galvanised by Houde and Hill’s earlier definition. It is
asserted that a prototype is an external artefact which embodies certain, or all, features of an intended
design. Such artefacts are external to the designer (does not exist only in designers mind) and may take
many forms and be created by a variety of prototyping methods.

2 PROTOTYPES AND PROTOTYPING METHODS

Prototypes can be enormously useful within the design process, having many purposes and uses.
Multiple papers and books propose different forms of prototyping each with their own benefits and
methods. Ulrich and Eppinger, (2012) describe the purpose of a prototype as being able to aid learning,
communication, integration and milestones. This categorisation of prototypes allows designers to
establish, from the beginning of a prototyping exercise, what the prototype they are about to create is for.
This has the benefit of allowing the prototype to be created with a fixed purpose in mind, helping to

1314 ICED19
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
 ensure that it meets that purpose, and may additionally reduce the possibility of hardware swamping
(Clausing, 1998). Ullman, (2010) takes a different point of view, by describing different types of
prototypes in terms of their purpose and at what phase in a product development they should be used,
describing proof of concept, proof of product, proof of process and proof of production. While Ullman
does provide more specifics and guidelines on what each prototype is for and how it should be manifest,
his list is somewhat constrained in terms of scope, there is no mention of a prototype used to
communicate. An additional observation would be that each type is mainly focused on validation and
verification.
Prototyping not only refers to the creation of prototypes, it also implies that prototypes are created and
used for a specific purpose (Pidaparti, 2018; Ulrich and Eppinger, 2012). It can be viewed as a step, task
or process within the design process, where the output is ultimately evaluated, with this evaluation
feeding back into the design process, seeding the development or validation of further concepts. The way
in which someone conducts a prototyping process will have some effect on the outputs of the process,
not just in what prototypes are created but what is gained from the activity. Structures for prototyping
processes have been proposed in some papers and even tested for validity (Christie, et al., 2012;
Camburn, et al., 2013a). While purposes and intent can vary across design contexts, for any model that is
shared across the design team and with key stakeholders, the accuracy and resolution of representation is
critical to its effectiveness. In this regard, the research here assumes that the purpose of the prototype is
to communicate design ideas from designer to designer, or from designer to client, in order to support
discussions and decision making as part of the design process. Prototypes constructed for the purposes of
individual exploration, or ‘internal conversation’, as discussed by Goldschmidt (2006) are not considered
at this time.
Prototypes can come in a near infinite number of shapes and sizes, however they can be generally be
placed on a spectrum of high-fidelity (hi-fi) to low-fidelity (lo-fi), the fidelity of a prototype refers to
how closely it resembles or how similar it is to the desired product (Yang, 2005; Houde, S. and Hill,
C., 1997). Lo-fi prototyping it characterised by quick and easy creation of an artefact. Lo-fi is also
known as low-tech as the means to create them consists mostly of time, with some cardboard, sticky
tape, post-it notesor similar. This type of prototyping has the benefit of being low cost and that it
allows non- professionals to actively participate in the design process (Egger, 2000).
At the opposite end of the spectrum there is hi-fi prototyping, in which a prototype is often created using
the same materials and production methods as the final product, intended to have the same functionality
and appearance as a final product (Walker, Takayama and Landay, 2002; Yang, 2005). Hi-fi prototypes
take more time, money and resources to make that lo-fi prototypes (Egger, 2000). The question of which
is best used depends highly on what they are being used for and is more than the simple question of how
much time and money is available. Several studies have investigated this question, Walker, Takayama
and Landay (2002), found few differences between the two when testing web prototypes, as the feedback
gained from both was equally as good. This is supported by the findings of Sauer, Seibel and Rüttinger
(2010), who also found that users are capable of extrapolating from lo-fi prototypes, meaning they could
infer functionality and problems that were not there. Both types of prototypes have their place within the
design process, with each being useful for different things (Rudd, Stern and Isensee, 1996).

3 EXPERIMENTAL DESIGN
An experiment was designed to further investigate how different prototyping methods may affect
communication of a design’s properties. Four prototyping methods were used to create proof-of-concept
prototypes (Ullman, 2010), and more specifically learning prototypes (Ulrich and Eppinger, 2012). They
were evaluated on how well they act as integration prototypes within Houde and Hill’s (1997) model.
This evaluation therefore has three categories: “role”, “look and feel”, and “implementation”. While
Houde and Hill’s work may initially appear a little dated, the principles for determining the efficacy of a
prototype are robust. However, the means of prototyping have moved on considerably since the 90’s;
with the advent and increasing affordability of virtual and augmented reality platforms for prototyping
how do these contemporary methods of prototyping compare in achieving the purpose of a prototype?
Additive manufacturing (3D printing), augmented reality, virtual reality and Computer Aided Design
(CAD) represent the primary emerging digital prototyping methods available to designers. All of
these, at their core, require some form of digital model that they represent. Using an identical model
for each method, allows for a comparison of the prototyping methods; the same artefact (be it digital

ICED19 1315
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
 or physical) is being displayed in each method, meaning that any differences observed are a result of
only the prototyping methods used. To prevent bias due to variation of the artefact, a physical
prototyping method that can utilise the same digital modal used by the virtual methods is key. For this
purpose, two possible methods were identified, these being CNC machining and 3D printing. While
both were considered the difference would only be materials used , rather than differences in the
methods themselves. As a result, 3D printing was included as the sole physical prototyping method.
The artefact chosen to be prototyped in the study would require some form of mechanism and yet
would need to be sufficiently familiar, or ubiquitous, to the participants. To fulfil these requirements a
CAD assembly of a cork screw was selected, as it is relatively simple while still having a mechanism
that required a degree of interrogation in order to review configuration and functionality, while at the
same time being a familiar object. The part file was obtained from Grab CAD (Crasto, 2018).
Figure 1 shows the different incarnations of the corkscrew product used in the study:
(a) Virtual reality; displayed using a stereo projector and polarised glasses.
(b) Augmented reality displayed on an iPad that was running the e drawings app.
(c) CAD displayed on a laptop.
(d) 3D printed prototype created in blue ABD plastic at a 0.2 mm layer height.

Figure 1. Different prototyping methods used in the study

3.1 Evaluation
A survey was constructed consisting of 3 sections, the first section contained two screening questions,
used to ensure participants were designers familiar with prototypes, preventing those who do not
belong to the research population being included (Brace, 2004). A total of 30 trainee design engineers,
with a minimum of three years training, participated in the study. The main body of the questionnaire
consisted of three questions that were repeated for each of the four prototypes. These questions asked:
How well could you determine the final product’s functionality, what it’s function or job is?
How well could you determine the look, feel and sound of the final product?
How difficult was it to understand how it operates?
These questions were answered on a semantic differential scale, with the dimensions being composed
of 6 points, that were rated from “with great difficulty” to “very easily” for the first two questions and
“very difficult” to “very easy” for the third. The participants were additionally prompted to answer an
open question for any additional information.
The last part of the survey was more general, it asked:
Which of the prototyping methods did you like the most?
Which of the prototyping methods did you find the most useful?
These questions were also followed up by an additional probing question, asking “could you please
explain your answer”. This last section was included as it is more open, and opinion based, than the
first two. This provided participants a chance to answer the questions however they liked, this allowed
for their opinions on what makes a “good” prototyping method.

4 RESULTS
The results below show the total mean scores for each prototyping method. These were generated by
taking the mean of each aspect that was rated (roll, look & feel and implementation) and then taking
the mean of those 3, the result is the average score the participants gave each prototyping method.

1316 ICED19
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
 4.1 Rating
Figure 2 shows that the physical prototype, on average, was the best of the four at communicating all
aspects, while the AR prototype was rated the poorest, on average. Figure 3 compares the percentage
of the total ratings that were positive and negative for each prototyping method. This perspective on
the results shows that overall there were more positive ratings for the virtual reality prototype than the
CAD prototype, helping to reinforce the mean score findings. Overall the AR prototype is the only one
that has more negative votes than positive.

Figure 2. Mean scores for each prototyping method.

Overall positive and negative ratings for each

method

100%
80%
60%
40%
20%
0%
phisical CAD augmented reality virtual reality

Number of overall positive votes Number of overall negative votes

Figure 3. Overall positive and negative ratings for each prototyping method.

4.2 Preference
Figure 4 details what method the participants identified as their preferred method, they did not need to
provide a reason, although some did, and this question was based entirely on their own opinion. More
than 50% of the participants in the investigation voted the AR prototype as their favourite, which was
twice that of second place. Most of the reasons stated for AR prototype preference were that it was
“cool”, “new”, “interesting” and “techy”. While the reasons given for selecting the physical prototype
were entirely based around it being functional and the participant could hold it in their hands.
Participants were then asked what method they found to be the most useful, as shown in Figure 5.

ICED19 1317
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
 Most participants thought that the physical prototype was the most useful, with most reasons given
mentioning that it is the only one that allows the prototype to be held in the hand and felt.

Score for preferred method

CAD prototype,
7% Physical
prototype, 27%

Augmented virtual reality

reality prototype, prototype, 13%
53%

Figure 4. Score for preferred method.

Score for method perceieved to be most useful

CAD prototype,
14%

Augmented
reality prototype,
11%

Physical
prototype, 75%

Figure 5. Score for method perceived to be most useful.

5 DISCUSSION & CONCLUSION

Prototypes and prototyping in general have existed for many years and presumably will do for years to
come, for this to be the case there are benefits to these artefacts and processes. While individual
prototypes can differ vastly from each other, prototyping when considered as a process or technique is
far more clearer in terms of what it can offer individuals and teams. Many authors have asked the
question “What do you gain from prototyping?” often only looking at specific areas, although some
have tried to cover the entire subject. There is some consensus between authors on a few of the general
answers, although many approaches each differently.
Prototyping helps communicate ideas (Buchenau and Suri, 2000). This is achieved by synchronising the
mental model individuals have of a concept, ether passively acting as a representation or actively by
providing a mechanism for discussion, these faculties in turn allow the concept to be explained, allow
others to provide focused feedback on the specifics of a design and these two combined allow for
meaningful negotiations to be conducted about the prototype, product or idea, (Lauff et al., 2018). This
communication between team members does not only apply to designers it may apply to managers,
clients and other stakeholders (Gerber and Carroll, 2012). Part of this learning process is feedback,
prototypes provide feedback in many ways, such as showing how components interact and fit together.
Within design and the design of a new product feedback is essential, and arguably the most important
feedback is gained from the intended end user. Prototypes can act as excellent tools to gain user

1318 ICED19
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
 feedback, (Steven Dow et al., 2009; Lui and Khooshabeh, 2003), as they allow the user to share the
identical mental image of a product and provide something that they can test and interact with.
Such feedback can represent a paradox; it is both essential and innately useless. Regardless of how much
feedback is gained if it is not acted upon to make decisions, it is only data. Decision making is
fundamentally important to the design process, not only for improvements to be implemented but to
move between key stages in the design process and ensuring the deadlines are met, (Yilmaz et al., 2013).
Prototypes have been shown to be useful for decision making where there are various competing
recruitments involved (Doke and Swanson, 1995; Smith and Eppinger, 1997). While feedback it still the
heart of good decision making prototyping and prototypes can provide this feedback, (Krishnan and
Ulrich, 2001). Prototypes inform decision making, as they provide a combined ability to aid
communication (including gaining feedback) and facilitate learning (Lauff et al., 2018).
Informed decision making can be the making or breaking of a project, this also applies to the decision
of when and how prototyping could be used within a project. Implementing it early has been shown to
increase the quality of a final product (Yang, 2005). This increase in quality through prototyping is not
limited to only the earlier stages however, an increase in technical quality can be seen though the
implementation of prototyping within a project (Steven Dow et al., 2009). One of the faculties of
prototyping that can provide this improvement of results is its ability to allow individuals and teams to
explore parallel design concepts (Steven Dow et al., 2011; Walker, Takayama and Landay, 2002).
From literature there appears to be a lot to be gained from prototyping, and by extension prototypes.
Generalising and combining the finding, prototypes and prototyping provide five benefits, which are
that they:
facilitate communication
aid learning
help gain and provide feedback
inform decision making
provide superior design outcomes
The results of the investigation indicate that physical prototyping should be employed if the purpose of
the prototype is to communicate the role and implementation of a product. However, if the scale and
cost of a physical prototype become prohibitive, CAD prototypes should be considered as this method
is regarded as similar to that of the physical but with the added benefit of not affected as much by cost
and size constraints.
The apparent favouring of physical prototyping methods over virtual based prototyping methods
appears to be a result of participants being able to hold and feel the prototype. This tactile feedback is
lacking in other methods, although the question of scale arises again, as this may not be the case for
large products that do not fit into the hand. Further study could investigate how the results of an
investigation like this one change when multiple products of various sizes are considered.
Virtual reality only excelled in one area, describing the look & feel of a product, this is assumed to be
a result of it being 3D, which may have allowed it to describe the form and shape of the corkscrew
well. On the other hand, its performance was comparable to that of the physical prototype. This
ultimately leads back to the questions of scale and cost however, if it is more economical and practical
to use a virtual reality prototype then it should be considered, but only if it is for communicating look
and feel, if there are other considerations than physical may be worth the investment.
The AR prototype performed the worst in all the aspects considered but it was rated by a large
majority to be there favourite method. The reasons given all related to the method being “cool” or
“new”, which strongly suggest this is a result of “the wonder of the new”. However, this is speculative
at this stage although the results suggest this it could be a potentially fruitful area of investigation. The
results would appear to suggest that physical prototyping performed best overall. However, there are
caveats which prevent it from being recommended in every purpose.

6 CONCLUSION
It is concluded from review of literature that for prototyping to be successful there must be purpose
behind each prototype and the process must have an end goal in mind. The results of the experimental
study would appear to indicate that physical prototyping should be employed if the purpose of the
prototype is to communicate the role and implementation of a product. Which would appear to support
widely held assumptions that the construction of physical models is important and effective in

ICED19 1319
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
 communicating (and hence exploring) the functional and mechanical aspects of a design concept, the
prototype as a stimulus for discussion, as it were. However, results would also appear to indicate that
augmented reality models brought the engagement, but did not necessarily deliver on the information,
given subjective participants responses in explaining this preference there may be a “wonder of the
new” involved in the use of these methods, although this would require further investigation. In
practice, and in certain settings, this may be enough in itself and there will likely be future
opportunities to make more sophisticated AR models with aspects of physical feedback. Areas for
future study would be to ascertain prototypes created with different purposes and mind and how
different prototyping methods may facilitate these purposes.

REFERENCES
Brace, I. (2004), Questionnaire Design : How to Plan, Structure and Write Survey Material for Effective Market
Research. 2nd ed. Kogan Page, Limited, London, pp. 63–126.
Dunlap, B.U., Hamon, C.L., Camburn, B.A., Crawford, R.H., Jensen, D.D., Green, M.G., Otto, K. and Wood,
K.L. (2014), “Heuristics-based prototyping strategy formation: Development and testing of a new
prototyping planning tool”, (2014), In: ASME 2014 International Mechanical Engineering Congress and
Exposition. ASME 2014, p. 9.
Buchenau, M. and Suri, J.F. (2000), Experience Prototyping Proceedings of the Conference on Designing
Interactive Systems Processes Practices Methods and Techniques, pp. 424–433.
Camburn, Dunlap, B.U., Kuhr, R., Viswanathan, V.K., Linsey, J.S., Jensen, D.D., B., Dunlap, B., Kuhr, R.,
Viswanathan, V., Linsey, J. and Jensen, D. et al. (2013a), “Methods for prototyping strategies in
conceptual phases of Design: Framework and experimental assessment,” In: ASME 2013 Power
Transmission and Gearing Conference. Portland: Volume 5: 25th International Conference on Design
Theory and Methodology, p. 10.
Camburn, B.A., Dunlap, B.U., Linsey, J., Viswanathan, V., Jensen, D., Crawford, R., Otto, K. and Wood, K.L.
(2013b), “Using design problem characteristics to build a prototyping strategy,” ASEE Annual Conference,
Atlanta, GA.
Crasto, M.C. (2018), Grab CAD. [Online]. [3 December 2018]. Available from:
https://grabcad.com/library/cork-screw-3
Christie, E.J., Jensen, D.D., Buckley, R.T., Menefee, D.A., Ziegler, K.K. and Wood, P.K.L. et al. (2012),
“Prototyping strategies: Literature review and identification of critical variables”, Proceedings of the 2012
American Society for Engineering Education Conference
Clausing, D. (1998), Total Quality Development, 2nd ed, ASME Press, New York.
Doke, E. and Swanson, N. (1995), “Decision variables for selecting prototyping in information systems
development: A Delphi study of MIS managers”, Information & Management, Vol. 29 No. 4, pp. 173–182.
Dow, S.P., Glassco, A., Kass, J., Schwarz, M. and Klemmer, S.R. (2009), “The effect of parallel prototyping on
design performance, learning, and self-efficacy”, Stanford Tech Report. [online] Stanford: Stanford
University HCI Group. Available at:
https://hci.stanford.edu/publications/2009/EffectOfParallelPrototyping.pdf [Accessed 1 Dec. 2018].
Dow, S., Fortuna, J., Schwartz, D., Altringer, B., Schwartz, D. and Klemmer, S. (2011), “Prototyping dynamics:
sharing multiple designs improves exploration, group rapport, and results”, In: Conference on Human
Factors in Computing Systems, AMC, New York, NY, USA, pp. 2807–2816.
Egger, F.N. (2000), “Lo-Fi vs. Hi-Fi Prototyping: how real does the real thing have to be?”, Teaching HCI
Workshop, OZCHI2000, Sydney (Australia), December 14–19, 2000
Gerber, E. and Carroll, M. (2012), “The psychological experience of prototyping,” Design Studies, Vol. 33 No.
1, pp. 64–84.
Goldschmidt, G. and Smolkov, M. (2006), “Variances in the impact of visual stimuli on design problem solving
performance,” Des. Stud., Vol. 27 No. 5, pp. 549–569.
Houde, S. and Hill, C. (1997), “What Do Prototypes Prototype”, Handbook of Human-Computer Interaction,
2nd ed, Elsevier Science, Amsterdam, pp. 367–381.
Krishnan, V. and Ulrich, K. (2001), “Product development decisions: A review of the literature”, Management
Science, Vol. 47 No. 1, pp. 1–21.
Lauff, C., Kotys-Schwartz, D. and Rentschler, M. (2018), “What is a prototype? What are the roles of prototypes
in companies?”, Journal of Mechanical Design, Vol. 140 No. 6, p. 061102.
Lidwell, W., Holden, K. and Butler, J. (2003), Universal Principles of Design, Rockport Publishers, Gloucester,
MA.
Lui, L. and Khooshabeh, P. (2003), “Paper or interactive?: a study of prototyping techniques for ubiquitous
computing environments”, In: Extended Abstracts on Human Factors in Computing Systems, AMC, New
York, NY, USA, pp. 1030–1031.

1320 ICED19
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
 Morris, C. (2018), “Prototype, engineering”, In: Academic Press Dictionary of Science and Technology, 1st ed.
[online] Academic Press, San Diego, p. 1744.
Rudd, J., Stern, K. and Isensee, S. (1996), “Low vs. high-fidelity prototyping debate”, Interactions, Vol. 3 No. 1,
pp. 76–85.
Sauer, J., Seibel, K. and Rüttinger, B. (2010), “The influence of user expertise and prototype fidelity in usability
tests”, Applied Ergonomics, Vol. 41 No. 1, pp. 130–140.
Smith, R. and Eppinger, S. (1997), “A Predictive Model of Sequential Iteration in Engineering Design”,
Management Science, Vol. 43 No. 8, pp. 1104–1120
Ullman, D. (2010), The Mechanical Design Process, 4th ed, McGraw-Hill, Boston, pp. 117–131.
Ulrich, K. and Eppinger, S. (2012), Product Design and Development, 5th ed, Irwin/McGraw-Hill, Boston,
Mass., pp. 289–309.
Wall, M., Ulrich, K. and Flowers, W. (1992), “Evaluating prototyping technologies for product design”,
Research in Engineering Design, Vol. 3 No. 3, pp. 163–177.
Wang, G. (2002), “Definition and review of virtual prototyping”, Journal of Computing and Information Science
in Engineering, Vol. 2 No. 3, pp. 232–236.
Walker, M., Takayama, L. and Landay, J. (2002), “High-fidelity or low-fidelity, paper or computer? Choosing
attributes when testing web prototypes”, Proceedings of the Human Factors and Ergonomics Society
Annual Meeting, Vol. 46 No. 5, pp. 661–665.
Yang, M. (2005), “A study of prototypes, design activity, and design outcome”, Design Studies, Vol. 26 No. 6,
pp. 649–669.
Yilmaz, S., Daly, S., Christian, J., Seifert, C. and Gonzalez, R. (2013), “Can experienced designers learn from
new tools? A case study of idea generation in a professional engineering team”, International Journal of
Design Creativity and Innovation, Vol. 2 No. 2, pp. 82–96.

ICED19 1321
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
https://doi.org/10.1017/dsi.2019.137 Published online by Cambridge University Press
J PROD INNOV MANAG 2015;••(••):••–••
2016;33(3):320–341
©
C 2015 Product Development & Management Association
V Association
DOI: 10.1111/jpim.12276
10.1111/jpim.12276

Perspective: A Review of Marketing Research on Product

Design with Directions for Future Research*
Michael G. Luchs, K. Scott Swan, and Mariëlle E. H. Creusen

The authors provide synthesized summaries of research on product design conducted over the 20-year period from 1995
to 2014, as well as suggestions for future research. Building on the conceptual model of product design proposed by
Luchs and Swan, the current project describes research findings based on a review of 252 articles drawn from eight of
the academic journals most influential to marketing thought, and identified by their inclusion of the terms “product
design” or “industrial design” within their abstracts, subject terms, and/or author supplied keywords. Specifically, the
authors provide integrated summaries of 25 product design subtopics organized within Luchs and Swan’s original 11
product design research topic categories, which, in turn, address the following three general product design research
categories: context and strategy, product design process, and product design consequences. These summaries are
followed by suggested future research opportunities to address gaps in the literature. In addition to seeking inspiration
for future research based on a review of extant research, the authors illustrate an approach for exploring research
opportunities based on current and emerging industry trends, such as sustainability, the sharing economy, and the
emergence of consumer-oriented health and performance management products. For each identified industry trend, the
authors provide illustrative design implications with consequent illustrative research opportunities. This balanced
approach to identifying near-term research opportunities based on extant research and based on industry trends, i.e.,
looking forward and externally, may in turn improve the potential impact of future research on both knowledge
development and on industry practice.

Practitioner Points strategy and market success,” Luchs and Swan (2011,
p. 327) provided an analysis of research on product
• Offers managers a comprehensive, structured view of design, addressing both interdisciplinary journals, such
design and an ability to better plan their use of design. as the Journal of Product Innovation Management, and
• Identifies and develops a broad array of strategic ele- marketing oriented journals, such as the Journal of Mar-
ments of design to balance a traditional focus on its keting Research. Their analysis describes the growing
aesthetic role. body of academic research and responds to prior calls for
• Highlights tactical suggestions to improve efficiency, the development of supporting elements of product
optimization, and performance. design research, such as conceptual frameworks and defi-
• Identifies seven design trends which have implications nitions of key terms. The current research is intended to
for managerial challenges and opportunities. complement their analysis by providing a detailed and
current content review along with insight and guidance
Introduction for future research.
Luchs and Swan’s (2011) review of product design

P
roduct design is an important subject—both as a research covered a 14-year period from 1995 to 2008.
process and outcome—for academics and practi- Their focus was on developing a conceptual framework
tioners alike. Given the emergence of product for product design, describing research on product design
design as “an integrated practice fundamental to firm relative to other related marketing topics, and describing
publication trends. In addition to covering a longer and
more recent time period, specifically the 20-year period
Address correspondence to: Michael G. Luchs, The College of William from 1995 to 2014, the current project describes in
& Mary, P.O. Box 8795, Williamsburg, Virginia 23187-8795. E-mail:
[email protected]. Tel: +757-221-2906. greater detail what has been collectively learned within
* Thanks to Abbie Griffin for her guidance through the development of each of the product design topic categories, illustrated in
paper; thanks also to Gloria Barczak and other participants’ feedback at the
International Product Development Management Conference session Figure 1. A synthesis of these findings and suggestions
where some of these ideas were presented in Limerick, Ireland (2014). for future research are provided. While Luchs and Swan
2PRODUCT DESIGN
J PRODRESEARCH
INNOV MANAG J PROD INNOV MANAG 321
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

Figure 1. Conceptual Model of Product Design Research (Luchs and Swan, 2011)

(2011, p. 341) provided some guidance for future

BIOGRAPHICAL SKETCHES research, they acknowledge that “comprehensive and
Dr. Michael G. Luchs is an associate professor at the College of William specific recommendations are beyond the scope” of their
& Mary’s Raymond A. Mason School of Business, and Founding Direc-
tor of the Jim and Bobbie Ukrop Innovation & Design Studio
analysis and suggest that “. . . a more detailed analysis of
(www.wmidstudio.com). His research interests include sustainable con- the research within each topic may aid in the identifica-
sumption and product design. Dr. Luchs’s research has been published tion of apparent gaps in our knowledge.” In addition to
in the Journal of Marketing, Journal of Product Innovation Manage- identifying new research opportunities inspired by gaps
ment, Journal of Public Policy & Marketing, Journal of Business Ethics,
Journal of Consumer Policy, Journal of Marketing Management,
in existing research, an approach that is inspired by trends
Journal of Research for Consumers, and Journal of Business Research. in industry is provided, i.e., looking forward and exter-
He is also coeditor of Design & Design Thinking: Essentials in the nally. This balanced approach to identifying near-term
PDMA’s New Product Development Series to be published by John
research opportunities may, in turn, improve the potential
Wiley & Sons, Fall 2015.
impact of future research on both knowledge develop-
Dr. K. Scott Swan is a professor of international business, design, and ment and on industry practice.
marketing at The College of William & Mary. Prof. Swan was awarded
a Fulbright and named the 2015–2016 Hall Chair for Entrepreneurship First, the methodology employed for identifying and
in Central Europe at WU (Vienna, Austria) and The University of Bra- reviewing research on product design published over the
tislava (Slovakia). His latest publications are Design & Design Think- last 20 years is described. This is followed by a synthe-
ing: Essentials in the PDMA’s New Product Development Series (John
sized summary of the research and a discussion of
Wiley & Sons Publishers, Fall 2015, coeditor) and Innovation and
Product Management: A Holistic and Practical Approach to Uncer- research opportunities within each of the product design
tainty Reduction (Springer Science & Business Media, 2014). His topic areas. Finally, the authors’ approach for identifying
research has appeared in journals such as Strategic Management product design relevant trends in industry, research
Journal, Journal of International Management, Journal of International
Business Studies, Management International Review, Journal of Busi-
opportunities based on these trends, as well as a specific
ness Research, and Journal of Product Innovation Management. illustrative list of near-term, externally inspired research
Dr. Mariëlle E.H. Creusen is an assistant professor of marketing and
opportunities are described.
consumer research in the Faculty of Industrial Design Engineering at
Delft University of Technology in the Netherlands. Her research inter-
ests include consumer research methods in new product development Methodology
and consumer response to product design. She has published in journals
such as International Journal of Research in Marketing, Journal of
Product Innovation Management, European Journal of Marketing, and For consistency, our approach for selecting articles for
International Journal of Design. this review parallels the approach used by Luchs and
Swan (2011), albeit for a longer time period. Specifically,
322
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.3
2016;33(3):320–341 2015;••(••):••–••

Table 1. Subtopics Addressed within Each Category of the Product Design Conceptual Model (Adapted from Luchs
and Swan, 2011)
General Category Topic Category Subtopics

Context and strategy External context Integrating customer information

Trends in the environment
Competitors
Firm strategy, objectives, and capabilities Dynamic capabilities
Organization of the design function
Interfirm engagement Codevelopment
Product complexity and the supply chain
Product design process Idea generation and screening Customer needs assessment
Creativity and ideation
Concept development and evaluation Attribute-based models of consumer preference
Integrating user needs and firm capabilities
Beyond discrete attributes
From consumer research to mass customization
Customer involvement
Technical implementation Platforms and modularity
Efficiency and optimality
Commercialization Product design and commercialization
Package design
Product design consequences Consumer evaluation and choice Product form
Product function
Choosing within product lines
Post-consumer choice Product use
Product disposal and sustainability
Product success Product success
Firm performance Firm performance

eight of the journals1 most influential to marketing education were excluded), yielding a focal set of 252
thought identified by Baumgartner and Pieters (2000), articles.
who incorporate both the journal’s level of influence and Next, an integrated summary of research on product
span of influence in their rankings, are analyzed. This design conducted during this time period is provided,
latter measure is especially important in the current organized within the 11 topic categories identified by
context given the interdisciplinary nature of product Luchs and Swan (2011). These 11 topic categories are
design. Within these journals, articles for our analysis organized within the general categories of “context and
were identified by searching within the EBSCOhost data- strategy,” “product design process,” and “product design
base “Business Source Complete” for articles that contain consequences.” Within each topic category, this summary
the term “product design” or “industrial design” within is further organized by the subtopics identified in Table 1.
their abstracts, subject terms, and/or author supplied key-
words. The time period begins in January 1995, coincid-
ing with Bloch’s (1995) seminal article on product
Product Design Context and Strategy
design, and continues through December 2014. The Bloch’s (1995) influential model of the design process
search yielded 337 articles, although 85 did not suffi- begins with the identification of design objectives and
ciently address product design or were otherwise incon- constraints: performance, ergonomic, production/cost,
sistent with our focus (e.g., articles focused on design regulatory/legal, marketing program, and designer
related. Subsequent research has expanded the definition
of product design to address the strategic role of design
1
In alphabetical order: Journal of the Academy of Marketing Science, (see Luchs and Swan, 2011; Ulrich, 2011). Broadly, strat-
Journal of Business Research, Journal of Consumer Research, Journal of egy topics are classified into three areas: (1) the external
Marketing, Journal of Marketing Research, Journal of Product Innovation
Management, Management Science, and Marketing Science. See Luchs and context; (2) firm strategy, objectives, and capabilities; and
Swan (2011) for details about journal selection. (3) interfirm engagement.
4PRODUCT DESIGN
J PRODRESEARCH
INNOV MANAG J PROD INNOV MANAG 323
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

External Context approach to new product development (Raja, Bourne,

Goffin, Çakkol, and Martinez, 2013).
Chernov (2012) identifies three broad aspects of environ-
mental information gathering for strategic analysis, Trends in the environment. While a variety of macro
which are used to organize research within this topic: environmental issues are important, the dominant focus
• Identifying and selecting target customers along with in this area has been on ecological and social issues—i.e.,
performing value analysis; sustainability attributes of products (Eppinger, 2011;
• Understanding broad environmental variables; Esslinger, 2011). Sustainability generally refers to a goal
• Analyzing competitor information. of “meeting the needs of the present without compromis-
ing the ability of future generations to meet their own
Integrating customer information. One of the funda- needs” (World Commission on Environment and Devel-
mental roles of marketing is to provide information about opment, 1987). While Chen (2001) suggests that “green
customers, competitors, and the external context as input product development” presents a challenge given the
for the design process (Bailetti and Litva, 1995; potential trade-off between traditional product attributes
Michalek, Feinberg, and Papalambros, 2005). On one and environmental constraints, Dangelico, Pontrandolfo,
hand, too much customer input can increase time to and Pujari (2013) find that the acquisition of technical
market (Datar, Jordan, Kekre, Rajiv, and Srinivasan, know-how and the creation of collaborative networks are
1996), introduce biases (Antioco, Moenaert, and important for integrating environmental issues into
Lindgreen, 2008), conflict with the channel structure product design. Further, green product development leads
(Williams, Kanna, and Azarm, 2011), and produce dispa- to increased opportunities though not necessarily better
rate perceptions across functional areas (Haggblom, financial performance of new product development
Calantone, and Di Benedetto, 1995). On the other hand, (NPD) programs. Furthermore, Luchs, Brower, and
design-driven innovation makes it possible for firms to Chitturi (2012) show how aesthetic design can help to
innovate based on new product meanings that can ulti- compensate for consumer perceived trade-offs between
mately diffuse through society (Verganti, 2008). This a product’s functional performance and level of
approach addresses future as well as current customers’ sustainability. Fuller and Ottman (2004) suggest that
needs (Souder and Song, 1997; Veryzer and de Mozota, while sustainable product design can counter ecosystems
2005). degradation, current processes generally ignore this
While some designers endeavor to shape consumers’ increasingly important desire.
preferences through design, firms achieve greater success
in the marketplace by addressing customers’ needs, Competitors. As technology-driven markets mature,
across the entire product-market life cycle, and with a customers may reach “satiation” with the performance of
diversity of options (Kahn, 1998). This ability is culti- existing products (Adner and Levinthal, 2001). Thus,
vated through an understanding of the relationships increasing performance and maintaining price in a com-
among innovation, product design, and cross-cultural dif- petitive market are often less productive than addressing
ferences (Moon, Miller, and Kim, 2013). However, the unsatisfied customer needs. This observation has impor-
sociocultural context and other environmental factors tant implications as firms attempt to balance user needs
complicate the design process and create uncertainty with corporate objectives. Specifically, design can be a
(Giloni, Seshadri, and Tucci, 2008). For instance, driver of innovation addressing both customer needs and
Amaldoss and Jain (2008) note that social forces, such as corporate objectives (Brown and Katz, 2011; Verganti,
reference group effects, can lead customers to identify 2008).
costly features that offer little true value-in-use. Other research shows that lower external information
The ability to address customer needs is also affected search costs for buyers may lead to a greater focus by
by the increased technical complexity of many products firms on product design and, as a result, higher product
(Goffin, 1998; Nambisan, 2002) and complicated by the differentiation that decreases price competition (Kuksov,
wide variety of products available from different compa- 2004). Alternatively, competition and changes in market
nies. One suggested solution is to accomplish cross- conditions influence the stage-to-stage information
product integration through “interfirm modularity,” dependencies within the firm. For example, Jespersen
whereby components of different firms work together (2012) demonstrates that “information dependencies
(Staudenmayer, Tripsas, and Tucci, 2005). This integra- increase priority given to financial decision criteria at
tion of products and services requires a different gates and lower priority given to customer and market
324
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.5
2016;33(3):320–341 2015;••(••):••–••

decision criteria.” Thus, the ease of price comparisons renewal, growth, and adaptation as the environment
offers an opportunity to firms that invest in long-term changes (Marsh and Stock, 2006). For instance, Cachon
product design differentiation over a more short-term and Swinney (2011) identify quick response production
product cost focus. capabilities with enhanced product design capabilities
that can offer significant value. Much of this research has
Research opportunities. The overall opportunity is to focused on efficiency, often associated with modeling
address how firms can best manage the uncertainty approaches (Smith and Eppinger, 1997). Efficiency goals
created by evolving technology, globalization, and social are connected with strategic planning, functionally inte-
change. Design provides the opportunity to address con- grated decision-making, and core capabilities (Michalek
sumers’ emerging needs and to respond to dynamic tech- et al., 2005; Srinivasan, Lovejoy, and Beach, 1997; Ulrich
nological change as well as social shifts that require more and Ellison, 1999; Ulrich and Pearson, 1998). Finally,
sustainable solutions. The nature of the design process is visual brand recognition is emerging as a design topic
to embrace early failure and uncertainty so as to continu- (Karjalainen and Snelders, 2010). Design capabilities in
ously iterate toward better solutions. While many practi- this context lead to competitive advantage by serving as
tioners understand this and other core “design thinking” an expression of brand values.
principles, researchers have not sufficiently incorporated
this mindset into our theories. Specific research opportu- Organization of the design function. Another signifi-
nities include: cant stream of research addresses team capabilities and
their influence on performance (see Dahl, 2011). Organi-
• Assessing customer needs requires understanding zational and technical practices, such as cross-functional
information. What are the design tools best used for teams and colocation, are effective for increasing new
comparison, for experimentation, and for learning? design integration with other areas (Datar et al., 1996;
• Examining how design can create opportunities in Liker, Collins, and Hull, 2003; Veryzer and de Mozota,
an environment of uncertainty and disruption. For 2005) and subsequent commercial success (Ettlie, 1995),
example, can design offer more incremental com- even across cultures (Souder and Jenssen, 1999). Adding
petitive advantages through quick iteration? How new team members increases product enhancements at a
can it be used toward developing discontinuous faster pace than more intense use of process technology,
advantages? but increases the need for repairs at almost the same rate
• Gaining a better understanding of how design can be as enhancements (Barry, Kemerer, and Slaughter, 2006).
used to: Specifically, the role of design in the NPD process and
○ Find differentiating factors that compensate for
combining design knowledge with prior knowledge (mar-
ignoring customers’ less important concerns keting or technological) related to NPD performance are
○ Improve the profitability of lagging segments
critical to our understanding of the role of design in
○ Move unprofitable customers toward the relation to other functional areas (Abecassis-Moedas and
competition. Mahmoud-Jouini, 2008).
Research has begun to tackle how industrial designers
Firm Strategy, Objectives, and Capabilities work within the firm. Micheli, Jaina, Goffin, Lemke, and
Verganti (2012) investigate how managers and designers
The fundamental question in strategy is how do firms can interact more productively. Interpersonal communi-
achieve and sustain a competitive advantage? Generally, cation between designers and others in the fabrication
resources, including dynamic capabilities, are used to stages is positively related to the efficiency of a vertically
create positional advantages that are important to the integrated organizational structure (Monteverde, 1995).
customer with the goal of conversion into positive per- The development of discontinuous new products is dif-
formance outcomes (Day and Wensley, 1988). ferent from more conventional, incremental product
development and requires unique, contextual roles along
Dynamic capabilities. Design is increasingly seen as with complex integration for marketing and industrial
a strategic tool to develop dominant brands with lasting design (Veryzer, 2005; Zhang, Hu, and Kotabe, 2011).
advantages (Bloch, 1995; Noble and Kumar, 2010; Implementing design thinking offers end-user profiles
Srinivasan, Lilien, Rangaswamy, Pingitore, and Seldin, across the organization, helps cultivate an organic and
2012). As such, dynamic capabilities related to design collaborative organizational structure, establishes a
become a powerful source of competitive advantage, design language, and stimulates further design thinking.
6PRODUCT DESIGN
J PRODRESEARCH
INNOV MANAG J PROD INNOV MANAG 325
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

Research opportunities. Much of knowledge develop- tion can also be a problem (Lau, Yam, and Tang, 2011).
ment on product design has been accomplished by prac- Since there is no cooperative approach that is universally
titioners, not researchers. However, the integration of applicable, managers may need to engage in a portfolio of
practitioner knowledge within the theoretical structure of relationship approaches with buyers during NPD
dynamic capabilities would be a very influential contri- (Athaide and Klink, 2009). The difficulties of relation-
bution. How does design augment a theory of the firm? ship management, both within and outside the boundaries
Strategic design research opportunities include: of the firm, require creative and sometimes counter-
intuitive wisdom (Staudenmayer et al., 2005) but can lead
• Explaining how design capabilities build and sustain
to further success even in such things as green product
competitive advantage. For instance, market orienta-
development (Dangelico et al., 2013).
tion and design orientation are both concerned with
consumers but emphasize different resources, capabili-
Product complexity and the supply chain. Product
ties, approaches, tools, etc.
complexity rises with the number of product components,
• Determining whether and how designers can serve as
the extent of interactions among components, and the
strategists. While there are examples of designers as
degree of product novelty (Tidd, 1995). Using a property
chief executive officers, what training; personality
rights approach in the auto industry, Novak and Eppinger
traits; and ability to engender a design orientation,
(2001) argue that in-house production is more attractive
understand customers, or deliver creative break-
when product complexity is high so firms can capture
throughs allow designers to contribute at the strategic
benefits of their investment. Further, Baiman, Fischer,
level? Similarly, how might the adoption of “design
and Rajan (2001) highlight the importance of the rela-
thinking” by nondesigners within the firm change the
tionships among performance measures agreed upon by
role of those with formal design roles?
cooperating members of the supply chain, the architec-
• Understanding the effect of more people acting in a
ture of the product produced by the supply chain, and the
designer role with the popularity of design thinking
efficiency of the supply chain. Also, manufacturers can
training. How does design thinking training benefit the
use product line design to induce their intermediaries
firm and what are its limitations?
to have a consumer targeting strategy consistent with
their own, for example, carry their full product line
Interfirm Engagement (Villas-Boas, 1998), manage the trade-off between profits
from the sale of the product and after-sales service
Firms adopt cooperative strategies when they can achieve (Cohen and Whang, 1997), manage “days of supply”
their goals better together than apart. Despite the obvious (Cachon and Olivares, 2010), and the complexity quality
benefits of interfirm engagement, the downsides are trade-off (Gokpinar, Hopp, and Iravani, 2010).
myriad. New product development complexity often Positive design outcomes from complex products and
leads to partnerships (Muffatto and Panizzolo, 1996), but complex supply chains require difficult trade-offs among
many successful high-growth firms are having problems a host of idiosyncratic characteristics and variables.
moving to concurrent design as well as involving suppli- Black-box products (i.e., hidden internal parts and opera-
ers, dealers, and customers in the design process tions) require a highly interactive design process among
(Dickson, Schneier, Lawrence, and Hytry, 1995) (see also suppliers and original equipment manufacturers
the Customer involvement section). The trend has been to (OEMs)—firms cannot avoid changes in specifications;
outsource more of the parts and components, along with rather, this interactive approach should be a learning
complementary products and services, while cooperating opportunity (Karlsson, Nellore, and Soderquist, 1998).
more closely with fewer partners to address increased The distinctive composition of components determines
product complexity concerns. the performance and cost benefits of product architec-
tures (Desai, Kekre, Radhakrishnan, and Srinivasan,
Codevelopment. Cooperative research concerns 2001; Mikkola, 2006). In one example, an Italian motor-
include supply chain management, contract manufactur- cycle industry supplier-provided component design is
ing, and design outsourcing (e.g., Carson, 2007; Mikkola, used in only a few firms and only by specialist and niche
2006; Wasti and Liker, 1997). Ineffective relationship producers, never by volume producers who usually prefer
management with potential buyers during NPD can be a codesign relationship (Muffatto and Panizzolo, 1996).
an important contributor to new product failure in Uncertainty, stage in the life cycle, and position in the
technology-based industrial markets. Poor communica- supply chain influence design decisions and the focus of
326
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.7
2016;33(3):320–341 2015;••(••):••–••

innovation. Firms facing uncertainty adjust by increasing Product Design Process

the knowledge overlap with their collaborators. For
example, suppliers expand their architectural knowledge Next, topics within each of the four design phases are
relatively more during times of uncertainty, while assem- discussed: idea generation and screening, concept devel-
blers put greater emphasis on component innovation (Lee opment and evaluation, technical implementation, and
and Veloso, 2008). In summary, product complexity and commercialization. These activities focus on enabling the
cooperative strategies can offer opportunities but require rapid and effective creation, prototyping, and testing of
sophisticated management and creative design. product concepts.

Research opportunities. The big opportunity in inter- Idea Generation and Screening
firm engagement is to elucidate the importance of the
supply chain to competitive advantage within the context Idea generation requires a high level of integration
of globalization. The complexity of the product, logistics, between external factors, such as user needs, and internal
and value analysis over time require comfort with uncer- factors, such as technical capabilities (Veryzer and de
tainty and tools that can cope with these challenges. Spe- Mozota, 2005). Our review suggests that idea or concept
cific research opportunities include: generation often is treated either as a creative process
based on user requirements, or as an engineering design
• Finding approaches that allow firms to make decisions process that seeks to identify the “ideal concept” among a
on relative complexity versus value for current and large set of possible feature combinations (see the
future customers. Attribute-based models of consumer preference section).
• Understanding the relative contributions of firm attri- The former is used more for newer products, while the
butes in a design context that contribute to innovation latter is used more for incremental product improvements.
and product success.
• Examining the in-house/outsourcing decision in the Customer need assessment. Understanding custom-
context of the design function. Under what conditions ers’ needs is a robust, emergent theme, and promotes a
does it make sense to have the capabilities to design user-centered perspective early in the product develop-
wholly within the firm versus using outside ment process (cf. Rosenthal and Capper, 2006) while
consultancies (e.g., IDEO, frog), and when might these aiding the generation of ideas that appeal to customers.
two approaches complement each other? Finally, how Veryzer and de Mozota (2005) discuss the concept of
can transitions among design sourcing strategies best user-oriented design (UOD), also known as user-centered
be managed? design. UOD focuses on explicit and deep consideration
of customer needs, emphasizing user experience and con-
Summary: Product Design Context and Strategy sidering both future as well as current customers. This
approach goes beyond considering the performance
Three topics were examined in this section. The most aspects of a product to include issues of usability,
developed issue from a research perspective is product support, and compatibility with other products.
complexity, and it will likely continue to attract signifi- The pursuit of customer insights, specifically unad-
cant research attention. An integration of assessing (cus- dressed and latent needs, has led to a growing interest in
tomer) needs, dynamic capabilities (company), and research methods such as contextual observation and eth-
codevelopment (collaborators) could also offer profound nography. Rosenthal and Capper (2006) describe the
insights into how to deal with product complexity. process steps and alternative approaches that uncover
While sustainability is an important topic, design insights as well as illustrate the benefits and challenges of
research has left out much of the rest of the macro- ethnographic research in the front end of the product
environmental analysis—likely to its detriment. In this development process.
regard, one of the key research opportunities involves
identifying technologies and knowledge from other Creativity and ideation. Ideation research has led to
areas and industries to be absorbed (i.e., lateral think- many insights. Dahl, Chattopadhyay, and Gorn (1999)
ing). This could help to avoid the traditional fate of focus on designers’ use of visual mental imagery based
many industry leading firms: discontinuous innovations on imagination, which leads to more original designs
brought about by firms from outside the industry, versus memory imagery. Including the customer in
leading to creative destruction. imagination visual imagery also leads to designs that are
8PRODUCT DESIGN
J PRODRESEARCH
INNOV MANAG J PROD INNOV MANAG 327
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

perceived as more useful than those based on memory Concept Development and Evaluation
imagery. Dahl and Moreau (2002) suggest that analogical
thinking—the transfer of knowledge from a related The traditional focus on the use of multi-attribute models
domain, such as power tools, to a new domain, such as of consumer preference has more recently been supple-
powered tooth brushes—increases the originality of ideas mented with methods to incorporate more subjective
which in turn positively influences consumers’ willing- attributes and to extend applications. Attribute-based
ness to pay. Solution novelty increases with a larger con- models, optimization approaches, perceptual methods,
ceptual distance between the analogy and current and applications to mass customization offer abundant
offerings (Kalogerakis, Lüthje, and Herstatt, 2010). insights.
Richness of vision and touch input enhances perceived
functionality and decreases novelty of consumer- Attribute-based models of consumer preference. Sig-
derived product concepts (Rosa, Qualls, and Ruth, 2014). nificant prior research has addressed the difficulties of
When multiple people generate ideas, the frequency of predicting a product’s market acceptance. There has been
similar ideas indicates value so unique ideas are not gen- an emphasis on identifying the optimal product configu-
erally the most valuable ones (Kornish and Ulrich, 2011). ration from a set of possible product attributes and attri-
Another ideation technique is the use of modular bute levels, which is used as input for product design.
architectures for concept prototyping by combining Attribute-based techniques include conjoint analysis,
modules and learning about what needs have been multidimensional scaling, and multi-attribute value
addressed (Sanchez, 1999) versus creating concepts analysis (Shi, Olafsson, and Chen, 2001; Tomkovick and
based on customer needs (see also the Technical imple- Dobie, 1995). Recent research extends beyond the
mentation section). In addition, consumers are recog- optimal initial choice for a single product to address
nized as a direct source of new product ideas, for limitations inherent in attribute-based analytical
example, in “crowd sourcing.” Poetz and Schreier (2012) techniques.
show that, compared to firm professionals, user ideas Discrete choice models, using hypothetical choices
score higher on novelty and customer benefit, though derived through surveys, are widely used to estimate the
lower on feasibility (see also the Customer involvement importance of product attributes as input for product
section). design and marketing mix decisions (Feit, Beltramo, and
Feinberg, 2010). Several researchers have proposed
refinements to attribute-based models of consumer pref-
Research opportunities. A traditional problem is the
erence and share-of-choice problems, specifically for
integration of consumer information into the ideation
large numbers of attributes and attribute levels (e.g.,
phase of NPD. Insights into ways of increasing designer
Balakrishnan and Jacob, 1996; Camm, Cochran, Curry,
empathy with consumers and the effectiveness of these
and Kannan, 2006; Shi et al., 2001). Several researchers
tools in alternative contexts are important. Involvement of
demonstrate the benefits of hierarchical Bayesian
users in the idea generation phase is another way to
methods in recovering individual-level parameters in an
incorporate consumer information (see the Customer
effort to account for respondent heterogeneity (Andrews,
involvement section). Investigating the use of such
Ansari, and Currim, 2002; Sándor and Wedel, 2005).
methods as brainstorming, external idea contests, and the
Bayesian methods are also mixed with choice experi-
use of analogies with the number and quality of ideas is
ments (Arora, Henderson, and Qing, 2011).
an important but under-researched area for marketing
Other researchers emphasize the benefits of address-
(e.g., Dahl and Moreau, 2002). Specific research oppor-
ing respondents with “extreme preferences” (Allenby and
tunities in this area are:
Ginter, 1995). For example, one might focus on satisfying
• Identifying which techniques to use and how to the “lower fringe” of users in an effort to design a product
communicate consumer information to designers in that is robust to variability in consumer preferences. A
order to enhance engagement and empathy with combination of analytical tools can be used to provide
consumers. greater precision to predict market demand (see the
• Assessing the effects of specific creativity techniques Commercialization section). For example, Liechty,
on the originality and usefulness of ideas and on their Ramaswamy, and Cohen (2001) propose the use of con-
overall consumer appeal. sumer “choice menus,” with various feature and price
• Increasing insight into when and how crowd sourcing is level scenarios, in order to learn more about consumer
useful for idea generation purposes. preferences.
328
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.9
2016;33(3):320–341 2015;••(••):••–••

Integrating user needs and firm capabilities. Firms Likewise, actual product usage/trial may not aid concept
are challenged to find the best fit between market accep- selection significantly in situations where the product is
tance and firm factors such as technical capability and already relatively familiar to consumers (Dickinson and
product manufacturability. To address this concern, Wilby, 1997). Thus, the value of representations in
Pullman, Moore, and Wardell (2002) propose the simul- concept evaluation may depend upon the product attri-
taneous use of quality function deployment (QFD) and butes being evaluated (e.g., their novelty and complexity)
conjoint analysis. While QFD enables novel ideas and the as well as consumer characteristics, such as product
identification of engineering trade-offs, conjoint analysis knowledge.
forecasts the market reaction to alternative designs (also
see Katz, 2004). Luo (2011) proposes a product line opti- From consumer research to mass customization. A
mization method that simultaneously considers market- noteworthy trend is research on mass customization.
ing and engineering factors. Michalek et al. (2005) “Mass customization is a strategy that creates value by
combines analytical target cascading with traditional some form of company–customer interaction at the
marketing methods, such as conjoint analysis, for a more fabrication/assembly stage of operations . . . to create
optimal solution. customized products with production cost and monetary
Arora, Allenby, and Ginter (1998) integrate both con- price similar to those of mass-produced products”
sumer choice and quantity decisions. To provide a better (Kaplan and Haenlein, 2006, p. 176). Most notable is the
assessment of market success, their model targets consum- exclusion of services, which they argue are necessarily
ers based on their purchase context. While conjoint analy- customized, and a focus on the operations of the firm
sis is traditionally used to optimize design of a single (e.g., component assembly into a finished product) as
product, it can address the problem of optimal product line opposed to the processes of the firm (e.g., technology
and product platform design (Chen and Hausman, 2000; development). Fueled in part by the interactive capabili-
Moore, Louviere, and Verma, 1999). Other firm contexts ties of the Internet, mass customization often involves
matter, such as government regulations. For example, outsourcing many of the tasks depicted in Figure 1 to
Chen (2001) considers interactions of consumer prefer- consumers, from customer needs identification through
ences and environmental standards to optimize product idea generation and concept selection. A variety of per-
design decisions. Complex factors often require a contin- spectives are apparent including how firms can benefit,
gency approach; firm capabilities serve as flexible strate- when it is most likely to succeed, its limitations, and ways
gic options that firms can deploy in ways consistent with to overcome limitations.
environmental forces (Moorman and Slotegraaf, 1999). Several researchers examine when mass custom-
ization is likely to succeed. According to Simonson
Beyond discrete attributes. Incorporating consumers’ (2005), the advantage of individual marketing-based
perceptions of subjective characteristics, using proto- approaches over segmentation have been greatly exagger-
types, improves product preference predictions, as ated. He argues that firms must carefully consider a
attribute-based predictions do not adequately include variety of consumer individual differences such as the
consumer response to product aesthetics, product usabil- customer’s knowledge and the stability of their prefer-
ity, or integrative effects of design decisions (Luo, ences, as well as the type of purchase and presentation
Kannan, and Ratchford, 2008; Srinivasan et al., 1997). format. Further, Gosh, Dutta, and Stremersch (2006)
Simply put, a product is more than the sum of its parts. contend that firms must consider the limits of what the
Also, prototypes allow the collection of information in a customer should and can customize. Subsequently, firms
variety of usage situations. Subsequently, firms better should take more control of design as technology
understand which designs are robust, producing lower unpredictability increases and as modularity and cus-
variations in product performance and consumer prefer- tomer knowledge decrease. Randall, Terwiesch, and
ences across usage situations (Luo, Kannan, Besharati, Ulrich (2007) seek to overcome the limits of customer
and Azarm, 2005). knowledge and describe the advantages of needs-based
The benefits of developing prototypes are significant, systems over traditional feature-based systems. With this
but there are costs and challenges. While using realistic approach, the user specifies the relative importance of
pictorial representations can help respondents understand their needs—while the firm retains control over how
design attributes, verbal representations may actually these needs are met. Franke, Keinz, and Steger (2009)
facilitate judgment when attributes are well understood show benefits gained from customization are higher when
(Vriens, Loosschilder, Rosbergen, and Wittink, 1998). consumers have better insight into their own preferences,
PRODUCT DESIGN
10 RESEARCH
J PROD INNOV MANAG J PROD INNOV MANAG 329
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

are better able to express their preferences, and have understand user needs in detail, in favor of transferring
greater product involvement. In addition, a customer’s need-related aspects of product and service development
base category consumption frequency and need satisfac- to users” (p. 821). The value of user toolkits is shown in
tion positively influence their intention to adopt a mass- business-to-business contexts, where demand is hetero-
customized product (Kaplan, Schoder, and Haenlein, geneous and users are willing to invest the time and effort
2007). Defining consumer design and the factors that required (von Hippel and Katz, 2002), as well as in
influence value creation are important questions and business-to-consumer markets (Franke and Piller, 2004),
opportunities for research (Moreau, 2011). specifically for mass customization purposes (see also the
Several researchers investigate the influence of self- previous section).
customization procedures and toolkit configuration on Carbonell, Rodríguez-Escudero, and Pujari (2009)
self-design evaluations (e.g., Franke and Schreier, 2010; show that technical novelty and turbulence positively
Franke, Schreier, and Kaiser, 2010; Moreau, Bonney, and influence customer involvement in service development.
Herd, 2011; Valenzuela, Dhar, and Zettelmeyer, 2009). Customer involvement, in turn, is positively associated
For example, Franke and Schreier (2010) show that with technical quality and innovation speed, promoting
process enjoyment and effort influence the perceived competitive superiority and sales performance. In addi-
value of products that are self-customized using a toolkit, tion, they find that the importance of customer involve-
while Franke, Keinz, and Schreier (2008) promotes the ment is independent of the development process stage,
benefits of mass customization user communities, where suggesting that customers should be involved throughout
customers benefit from each other’s preliminary ideas as the entire development process.
well as design solutions. Further, for complex products, Openness increases user innovation community
the use of starting solutions leads to enhanced choice members’ involvement and contribution to the innovation
satisfaction and the purchase of more feature-rich and project (Balka, Raasch, and Herstatt, 2014). A challenge
higher priced products (Hildebrand, Haubl, and concerning virtual user communities is selecting the most
Herrmann, 2014). attractive ideas. Jensen, Hienerth, and Lettl (2014) offer
heuristics that can help firms select promising designs
Customer involvement. More recent approaches from the abundance often generated. In addition, idea
involve consumers in the idea generation and concept quality measures from online consumer panels are better
development phases, and sometimes even in the commer- at determining good ideas over expert evaluations
cialization phase. Methods include the lead–user method (Kornish and Ulrich, 2014).
(e.g., von Hippel, 1998), cocreation, and toolkits for user
innovation and design (e.g., Franke and Piller, 2004; von
Research opportunities. Incorporation of subjective
Hippel and Katz, 2002). These approaches either provide
experiential product attributes in concept optimization
designers with need-related information or provide them
has received more attention in the last decade. Technical
with product concepts ready for production.
developments such as 3D printing, computer simulation,
The lead user method involves understanding innova-
augmented reality, and online mass customization tool-
tive uses of products as an important potential market
kits have increased the possibilities of assessing con-
trend (e.g., von Hippel, 1998). The use of peer commu-
sumer response to such attributes. Taken one step further,
nities offers additional value in lead user research, as
this enables consumers’ influence on these elements
community members provide feedback on ideas, make
directly along with designing and producing their “own”
concrete development contributions, test new product
product. In such cases, the role of the designer may
concepts, and even help to diffuse innovations (Hienerth
change from a designer of a product to a designer of a set
and Lettl, 2011). Hoffman, Kopalle, and Novak (2010)
of possible product options that can be chosen and com-
show consumers high in “emergent nature” have a unique
bined by consumers. The design of the interface of the
capability to imagine product concepts that are more
toolkit is important as it determines the boundaries and
appealing and useful to mainstream consumers than those
possibilities for consumers to configure their product.
developed by typical, lead user, or innovative consumers.
Specific research opportunities include:
Toolkits can allow potential customers to design their
own product from a set of predefined elements, then • Improving insight into how the design and configura-
iterate and even simulate the product’s use in their own tion of a mass customization toolkit influences con-
environment to test performance (von Hippel and Katz, sumer satisfaction and the quality of the final
2002). The goal of a toolkit is to “abandon the attempt to configured product.
330
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.11
2016;33(3):320–341 2015;••(••):••–••

• Increasing understanding of when and how to best Thomson, 2004; Erat and Kavadias, 2008; Loch,
apply rapid prototyping techniques such as 3D printing, Terwiesch, and Thomke, 2001; Michalek et al., 2005;
computer simulation, and augmented/virtual reality Netessine and Taylor, 2007). Thomke (1998) suggests
techniques in concept testing. that economical switching between different experimen-
tation modes in product design (and the related trial and
Technical Implementation error learning), such as computer simulation and rapid
prototyping, reduces total product development cost and
Implementation is a varied topic that includes sharing time. In situations where the use of visual representations
physical attributes or processes across products along or prototypes in consumer tests is justified (see the
with addressing concerns of efficiency and optimization. Beyond discrete attributes section), a related question is
This area holds some of the greatest interest for global how many prototypes to develop and test. Once again,
managers because of their concern with uncertainty and managers must balance the benefits of using prototypes
the consequent opportunities. against their costs. Dahan and Mendelson (2001) demon-
strate that investing in multiple prototypes makes the
Platforms and modularity. Sharing parts, compo- most sense in situations where the profit uncertainty is
nents, platforms, processes, or resources across products high and the cost of testing is low. Collaborative
is of great interest in practice and research (Fisher, prototyping can offer evaluative benefits and may be suc-
Ramdas, and Ulrich, 1999; Sköld and Karlsson, 2013). cessfully undertaken (depending on the design problem
For example, Meyer, Tertzakian, and Utterback (1997) and the market characteristics) at a range of costs
discuss the efficient construction of products on succes- (Terwiesch and Loch, 2004). In addition, collaborative
sive generations of underlying product architectures, prototyping leads to a better balance between functional-
product platform renewals, and derivative product gen- ity and usability (Bogers and Horst, 2014).
erations. The use of modular techniques enables a firm to Smith and Eppinger (1997) offer insight into sequen-
leverage investments (Krishnan and Gupta, 2001), creates tial iteration through optimally ordered design tasks to
performance advantages (Meyer and Dalal, 2002), offers minimize the duration of the design process. Other
greater product variety, and lowers time to market, while approaches highlight leveraging interdependencies
it decreases the cost of developing new product designs between design and manufacturing phases in NPD (Bajaj,
(Sanchez, 1999). Finally, proper feature commonality Kekre, and Srinivasan, 2004) or identifying conditions in
selection across product lines can improve consumer which design activity coupled with strategy minimizes
valuation, capture economies-of-scale in manufacturing, lead time but maximizes other performance measures
and reduce logistical concerns (Kim and Chhajed, 2001). (Joglekar, Yassine, Eppinger, and Whitney, 2001).
For extreme market diversity or high levels of Finally, top management involvement and participation
nonplatform scale economies, platforms are not appropri- can allow processes to successfully account for the level
ate (Krishnan and Gupta, 2001). Excessive levels of inte- of ambiguity alone, with less concern for the degree of
gration and overly refined modularization can be the volatility (Carson, Wu, and Moore, 2012).
effect of modest levels of search and a premature fixation
on inferior designs (Ethiraj and Levinthal, 2004). John,
Research opportunities. Platform sharing, modular-
Weiss, and Dutta (1999) suggest that (1) a high cost of
ity, upgradability, and optimized designs will require
producing the first unit (versus cost of reproduction)
additional research to understand appropriate contexts
favors platform designs over optimized designs, (2) tech-
and interrelationships. Opportunities include:
nology diversity favors modular designs over optimized
designs, and (3) knowledge tacitness favors optimized • Improving measures to help designers understand the
designs over modular designs. Finally, modular context, diminishing returns, and limits of manufactur-
upgradable architecture reduces product obsolescence ing modularity. The three basic modularity drivers (i.e.,
but increases design inconsistency (Krishnan and creation of variety, utilization of similarities, and
Ramachandran, 2011) and speed of improvements (Ülkü, reduction of complexities) could also be incorporated
Dimofte, and Schmidt, 2012). with the modularity challenges for better prescriptive
advice and theoretical understanding.
Efficiency and optimality. Technical engineering • Enhancing the ability of designers to make early deci-
design approaches offer tactical suggestions to improve sions with respect to the type, number, and quality of
efficiency and optimization (e.g., Bhuiyan, Gerwin, and prototypes, simulations, and the use of augmented
PRODUCT DESIGN
12 RESEARCH
J PROD INNOV MANAG J PROD INNOV MANAG 331
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

reality—this has implications across the product devel- package design can enable (Van der Lans, Pieters, and
opment process, functional areas, and a range of Wedel, 2008).
diverse environments. Although few package design-related articles were
identified using the keyword “product design,” there is
Commercialization research in the focal journals under the keyword
“package design” that concentrates on the influence of
Product design has many important implications after package shape on perceived size/volume (e.g., Krider,
products are manufactured. As the product is being devel- Raghubir, and Krishna, 2001) along with which food
oped, there remains a range of additional design deci- package size is most profitable and best for consumer
sions. These include package design, as well as those that welfare (e.g., Jain and Kannan, 2002).
are indirect consequences of earlier decisions, such as
pricing. Research on product design’s role in commer- Research opportunities. Commercialization and
cialization has addressed customers’ interest in mass package design appear to offer significant opportunity for
customization, willingness to pay, level of differentiation, new research. Design decisions are highly influential at
and package design. this stage since it is the last opportunity to position the
product. Visual product and package design may be the
Product design and commercialization. Research on best way to draw attention, gain interest, and engender
commercialization has primarily been focused on pricing desire from customers (see also the Product form
and self-designed products with little research focused on section). Related research opportunities include:
distribution or communication/promotional activities. In
attribute-based models of consumer preference, commer- • Developing ideas related to how consumers encounter
cialization aspects such as price sensitivities can be taken products: in person, virtually, or in print. Overall,
into account during concept optimization. For example, research is needed on the differences in designing prod-
Tomkovick and Dobie (1995) combine hedonic pricing ucts for alternative forums (e.g., on shelf, online) and
models and factorial surveys to gauge price sensitivity their interactions.
and market receptivity to new product designs. In addi- • Analyzing the interaction of design with distribution,
tion, market simulation can include competitive reactions supply chain, and communication/promotional activi-
(Giloni et al., 2008). Competition offers interesting ties. Given the importance of product design decisions
implications for mass customized products. Syam, Ruan, on cost to transport (e.g., ability to be efficiently
and Hess (2005) argue that firms need to carefully con- stacked, containerized, disassembled), and on point of
sider which features to customize with respect to com- purchase decisions including the role of packaging, this
petitors’ customized offerings. area provides significant opportunity for more research.
One advantage of mass customization is that consum-
ers show greater willingness to pay for self-designed Summary: Product Design Process
products (Franke and Piller, 2004). Such products offer
increased functionality, uniqueness, pride of authorship, Besides improving attribute-based models of consumer
and joy of performing a creative act. For fashion products preference, research has focused on the use of iterative
in particular, enhanced product design mitigates strategic prototyping during product design as a way to better
consumer purchasing behavior (e.g., purchase delay until incorporate more subjective and experiential product
clearance sale) (Cachon and Swinney, 2011). attributes. In addition, recent research has addressed mass
customization, the conditions under which it is success-
Package design. Package design incorporates engi- ful, and its limits. Customer involvement in the design
neered functional attributes (e.g., ergonomics, durability, process is another topic that has received increased atten-
recyclability) and a package’s visual attributes (Bloch, tion from researchers, with a focus on user design tool-
1995). Orth and Malkewitz (2008) suggest that package kits. Increasing our understanding of successfully using
designs have five holistic types and can be matched newer technologies—such as 3D prototyping—computer
with brand types. Raghubir and Greenleaf (2006) find simulations, and virtual reality in product design and
that the ratio of the sides of a rectangular product or development remain highly relevant research areas.
package can influence purchase intentions. In addition, Efficiency and optimality has been a significant
the extent to which a brand visually stands out, or brand concern for researchers. However, research on various
salience, is a vital competitive advantage and one that forums in which the product is encountered (e.g., online)
332
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.13
2016;33(3):320–341 2015;••(••):••–••

and packaging have strong managerial interest but have on consumers’ aesthetic responses to products. Raghubir
exhibited weaker levels of research interest. The greatest and Greenleaf (2006) state that the ratio of the length of the
opportunity to move this area forward may be in helping sides of a rectangular product or package can influence
design lessons learned migrate in the digital context. Spe- consumer preference and purchase intention. The sym-
cifically, as retailers and manufacturers embrace digital bolic value of product form has also been a research topic.
opportunities, evidence suggests that some translate well Orth and Malkewitz (2008) offer a taxonomy of design
while other design knowledge must be adapted, and types within product categories (e.g., bottles of wine)
others invented. Additionally, globalization will continue where each design type conveys a specific symbolic
to have a large influence on the need to understand the meaning, such as “natural” (see also the Package design
design implications for modularity, product platforms, section). They suggest that firms should select a design
and supply chain partners, along with a desire to find type based on the intended personality of the brand.
global lead users. Product design can be strategically used to communicate a
brand’s core values (Karjalainen and Snelders, 2010).
Product Design Consequences Prior research highlights the role of product form in
communicating functional attributes, i.e., the capabilities
Consequences, or outcomes, of product design include of a product (Bloch, 1995). Hoegg and Alba (2011) dem-
both consumer responses, such as consumer evaluation, onstrate that product form alters judgment about feature
choice and post-consumer choice, as well as product and function even with conflicting feature information. Luchs
firm performance. Our review addresses these two levels et al. (2012) show that superior aesthetic design increases
separately, i.e., at the level of the consumer and the firm. the confidence consumers feel toward sustainable prod-
Although consumer response and firm performance are ucts, thereby increasing choice likelihood. While func-
clearly interdependent, research has tended to focus on tional attribute information is often conveyed through the
either one or the other. visual elements of a product’s form, more recent research
has addressed other sensory cues. Krishna and Morrin
Consumer Evaluation and Choice (2007) addresses the influence of haptic cues, i.e., touch,
on perceptions of product performance.
Research has focused on consumer responses to product Research on how to design easy-to-use products has
form and product function, and their interrelations. For traditionally been published in design and ergonomic
example, product form influences consumer perception journals, not in the marketing literature. One exception is
of product function (e.g., Hoegg and Alba, 2011), and Oppenheimer (2005), which offers the metaphor of “con-
product function can be accentuated by product form. versation” to describe user–product interaction. Besides
creating ease of operation and ease of use, the design of a
Product form. Product form represents a variety of product communicates impressions that may influence
elements “chosen and blended into a whole . . . to achieve consumer perceptions at the point of purchase (Creusen
a particular sensory effect . . . including shape, scale, and Schoormans, 2005).
tempo, proportion, materials, color, reflectiveness, orna- Finally, product appearance influences the way con-
mentation and texture” (Bloch, 1995 p. 17). Bloch’s sumers categorize a product (Bloch, 1995). A product’s
(1995) seminal article identifies several ways in which visual “goodness-of-fit” with a category, or visual
the form of a product contributes to a product’s success. prototypicality, can enhance evaluation (Veryzer and
First, the form can help a product gain attention. Second, Hutchinson, 1998). Landwehr, McGill, and Hermann
the form can have an inherent aesthetic value that (2011) show prototypicality and complexity of frontal car
affects consumers’ evaluations and is consequential designs significantly improves sales forecasting. Despite
throughout the ownership of the product. Third, the form the benefits of clear category membership achieved
can communicate functional, experiential, and symbolic through visual prototypicality, firms may instead be inter-
information to consumers. Finally, the form enables cat- ested in finding ways to alter the categorization of a
egorization of the product. product and, thereby, strategically extend perceived brand
The relationship of form factors with the aesthetic breadth. Kreuzbauer and Malter (2005) show how
preference of consumers has drawn attention from several designers alter a product’s form to support line extension
researchers. For example, Veryzer and Hutchinson strategies. In some cases, design differentiation is strate-
(1998) demonstrate that both the perceived unity and gically advisable to create a positional advantage or to
prototypicality of a product design have a positive effect communicate new functional attributes. Talke, Salomo,
PRODUCT DESIGN
14 RESEARCH
J PROD INNOV MANAG J PROD INNOV MANAG 333
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

Wieringa, and Lutz (2009) highlight how design newness Choosing within product lines. Kahn (1998) argues
positively impacts car sales, while Radford and Bloch that companies are more effective when offering a high-
(2011) demonstrate that products high in visual newness variety product line, making it more likely that consum-
engender more positive affective consumer reactions. ers will find a desirable option as well as allowing
However, Mugge and Dahl (2013) show that low levels of consumers to enjoy a diversity of options over time.
design newness are better in the case of radical Others have shown that consumers’ evaluations of prod-
innovations. ucts are influenced by the presence of related products
The effect of product form on consumers’ evaluations within the same product line. Kim and Chhajed (2001)
varies, however, based on individual differences. Krishna demonstrate that when products share similar features
and Morrin (2007) demonstrate that the influence of within a product line, the valuation of the low-end
tactile feedback on perceptions of product performance product can increase while the valuation of the high-end
depends upon an individual’s need for touch. Bloch, product can decrease. Thus, designers must consider the
Brunel, and Arnold (2003) offer a scale on consumers’ effect that this context has on consumers’ valuations of
propensity to be influenced by visual product the individual products.
aesthetics—the Centrality of Visual Product Aesthetics.
An innovative product form increases liking for innova- Research opportunities. Illuminating the relationship
tive and high-design acumen consumers (Truong, Klink, between product form characteristics and consumer
Fort-Rioche, and Athaide, 2013). Further, while much of evaluations provides companies with a basis for their
the research on product design and product form focuses decisions about product appearance, rather than
on initial purchase decisions, Cox and Cox (2002) estab- leaving it solely to the intuition of designers. To
lish the value of studying consumers’ product evaluations date, most of the focus has been on the influence of
over time. They show that consumers’ liking of visually product form characteristics on perceived aesthetic
complex forms increases with repeated exposure. value. Research into the interrelationships between
form and function has increased over the last few
Product function. A significant amount of research years, but more research is needed into how product
approaches design as the process of determining which form can influence consumer perceptions of functional-
function/performance attributes to embody in a product ity, ease of use, and quality. Also, more research into
(see the Attribute-based models of consumer preference whether and how perceptions based on product form
section). In addition, research has focused on consumers’ differ across individuals, cultures, contexts, and time
responses to the holistic outcome of these discrete attri- would be valuable. Specific opportunities in this area
bute decisions. Designers may be tempted to include include:
many attributes with the underlying assumption that each
contributes independently to consumers’ evaluation of • Assessing the influence of product form characteristics
the product, that is, “more is better.” Companies may add on the perception of different types of product value
features to a product that have relatively little inherent (e.g., aesthetic value, functional value, perceived ease
functional value to most consumers based upon reference of use, and perceived quality).
group effects, for example, with publicly consumed • Increasing insight into how product form influences
luxury goods (Amaldoss and Jain, 2008). perception of product function and vice versa in differ-
Consumers may choose products with many features ent contexts and across target groups (e.g., different
given their initial focus on perceived product capability degrees of consumer product knowledge and involve-
(Thompson, Hamilton, and Rust, 2005). However, this ment, different types of product categories, different
increased capability may come at the expense of product countries).
usability and thus decrease satisfaction with the product
in use. Thompson et al. (2005) suggest that rather than Post-Consumer Choice
bundle a large set of attributes into a single product,
companies interested in maximizing consumer satisfac- In addition to changes in preference for design and fea-
tion and long-term sales are better off designing a larger tures across time (e.g., Cox and Cox, 2002; Thompson
set of more specialized products. Finally, other contexts, et al., 2005), researchers have addressed product design
specifically bottom of the pyramid countries, are a effects beyond initial product choice. Popular research
growing research concern and source of innovation in this subjects in this area include product use, disposal, and
context (Viswanathan and Sridharan, 2012). sustainability.
334
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.15
2016;33(3):320–341 2015;••(••):••–••

Product use. Chitturi, Raghunathan, and Mahajan • Analyzing how product sustainability can better be
(2008) show that a product’s hedonic and utilitarian ben- communicated to consumers and how product and
efits differentially affect post-consumption feelings. Spe- package appearance influence their perceptions of
cifically, products that exceed customers’ utilitarian product sustainability.
needs enhance satisfaction, whereas products that exceed • Determining how best to incorporate various
hedonic wants enhance customer delight, which improves sustainability-related frameworks, tools, and principles
customer loyalty beyond that derived from satisfaction. within the product design and development process.
Srinivasan et al. (2012) relate design elements of func-
tionality, aesthetics, and meaning to customer satisfaction Product/Firm Performance
and define customer segments that place different impor-
tance on these elements. As mentioned in the Product Product success and firm success are intimately con-
function section, consumers in general tend to focus on nected. Firms attempt to manage short-term individual
product capabilities during purchase, while usability is product success within the context of longer term firm
more important during use. Further, Ziamou, Gould, and level performance in order to afford the reinvestment
Venkatesh (2012) illustrate the importance of manufac- necessary to fund future growth.
turers accounting for consumers’ learning process
through design within the context of technology-based Product success. The strategic significance of domi-
products. nant designs, the balance of risks, investments, and
success rates, all feed into the desire to increase profits,
Product disposal and sustainability. Attention to sales, and develop strong brands (Srinivasan, Lilien, and
sustainability issues has emerged as significant Rangaswamy, 2006). There are a number of firm incen-
(Eppinger, 2011). Beyond considering the effect of tives to offer both customized and standard products in a
design decisions on consumers after initial product competitive environment (Syam and Kumar, 2006). Prod-
choice, research has begun to address other long- ucts customized on the basis of expressed preferences
term effects (see also the Trends in the environment produce significantly higher benefits from customers in
section). “Green product development” is becoming terms of willingness to pay, purchase intention, and atti-
increasingly important to customers (Chen, 2001), tude toward the product than standard products (Franke
although current marketing decision processes have et al., 2009). Finally, industrial design and cost engineer-
historically ignored this growing customer preference ing, implemented together, enhance both the effective-
(Fuller and Ottman, 2004). Thus, in addition to consid- ness and efficiency of NPD in early-stage firms by
ering the effects of design decisions on individuals, reducing development duration and cost for the compa-
there are opportunities to consider the indirect effects on ny’s first commercial product (Marion and Meyer, 2011).
the environment and, hence, on the sustainability. New product development managers must consider
Design frameworks and principles such as “cradle-to- the changing nature of competition and innovative prod-
cradle” responsibility (i.e., production, recycling, and ucts with respect to design and design strategy (Gemser
subsequent production of products from post- and Leenders, 2001). Also, selecting an architecture that
consumption materials) and “product life-cycle analy- maximizes product performance and facilitates develop-
sis” are increasingly being integrated within the ment process flexibility is associated with better-
standard design process. performing projects (MacCormack, Verganti, and Iansiti,
2001). Approaches for firms to adapt product definitions
to the level of uncertainty can be adjusted through “fre-
Research opportunities. Traditionally, the marketing
quent, repeated interactions with customers and using a
literature has focused on consumer product preference at
flexible development process” (Bhattacharya and
the point of purchase. More recently, attention has broad-
Krishnan, 1998) or creating robust products that work
ened to post-purchase satisfaction and accounting for
across multiple contexts (Swan, Kotabe, and Allred,
post-use effects during product design. Therefore,
2005).
research opportunities include:
• Understanding how to better communicate ease of use Firm performance. Product success measures are
in consumer purchase decisions. This is an important often tied together with firm performance variables like
determinant of satisfaction but often not strongly market share, sales, quality, brand development,
weighted when making the purchase decisions. innovativeness, speed to market, and profit (Amaldoss
PRODUCT DESIGN
16 RESEARCH
J PROD INNOV MANAG J PROD INNOV MANAG 335
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

and Jain, 2008; Dell’Era and Verganti, 2007; Moore in use and experiential product aspects. This is an impor-
et al., 1999; Novak and Eppinger, 2001). Gemser and tant research area and efforts to understand ways to
Leenders (2001) explicitly show integrating industrial reflect this in the design and testing of products during
design into product development increases company development would contribute greatly. Also, research on
performance. In addition, Chiva and Alegre (2009) product form and its influence on consumer perceptions
demonstrate that investment in design and design man- of product function have increased in the last decade; this
agement enhances firm performance, while Hertenstein, area still offers significant research opportunities.
Platt, and Veryzer (2005) corroborate by identifying
industrial design’s relationship to increased corporate Market Trends as Inspiration for
financial and stock market performance. The most Future Research
successful firms “achieve a balance between creative
innovation and cost discipline in the NPD process with In addition to identifying new research opportunities
third-party design and manufacturing resources . . . and inspired by gaps in existing research, inspiration can also
add additional creative resources to the core entrepre- be gained by exploring emerging technological and
neurial team, maximizing the ability to catalyze benefi- sociocultural trends in the market. The latter approach is
cial tension between creativity and cost discipline” speculative to some degree given that it depends on
(Marion and Meyer, 2011). assumptions about which trends are most important. It
Customer involvement in NPD, by allowing custom- also depends on extrapolating their relevance to industry
ers to create or select product designs, also has positive and to academic research. Nonetheless, this is an impor-
performance implications for firms through improving tant externally focused and forward-looking complement
corporate attitudes and behavioral intentions (Fuchs and to the approach used thus far. While researchers consider
Schreier, 2011; Schreier, Fuchs, and Dahl, 2012). In addi- many criteria as they begin a new project, thinking
tion, customization increases product differentiation, through issues and opportunities that managers may face
leading to less intense price competition (Loginova, in the near future will improve the practical relevance of
2012). the research.
The authors’ approach began with current and emerg-
Research opportunities. An important challenge with ing technological and sociocultural trends that the popular
respect to long-term product and firm success is how to press has identified as especially relevant to managers
engender “thinking outside the box” in order to bring in involved in product design and development. To do this,
ideas from outside the industry since many innovations, widely read business publications, such as Forbes, Wall
especially discontinuous innovations, come from outside. Street Journal, and Bloomberg/Businessweek as well as
The benefit of hiring a company like IDEO is that they leading publications focused on emerging trends in busi-
have access to (and can) broker innovation and technol- ness, such as Fast Company and Wired were reviewed.
ogy from diverse sectors. Survival is dependent on scan- One of the authors developed a list of current and emerg-
ning, accessing, transforming, and applying innovations ing topics based on patterns of articles published on spe-
from across industries to avoid being blindsided. Specific cific topics, such as “the Internet of Things,” as well as
research opportunities include: based on specific articles that explicitly identified emerg-
ing trends, such as Fast Company’s 2014 article, “The
• Understanding how design attributes from adjacent World-Changing Ideas of 2014.” This initial list was
industries can create value (e.g., improving the revised based on feedback from the other two authors.
experience—emotion, aesthetics, identity, social Next, the list of 10 trends was vetted with other aca-
impact, ergonomics, technology, and quality). demic researchers at the 21st International Product
• Developing the ability to identify and commit to Development Management Conference held in June 2014
courses of action that are likely to lead to discontinuous in Limerick, Ireland. Twenty-nine conference attendees
innovations. This also will require an ability to scan and completed a survey in which, for each trend, they were
incorporate technologies from outside the industry. asked to rate the extent to which they agreed that “. . . this
is an important trend worthy of product/service design
Summary: Product Design Consequences related research” on a scale of 1 (strongly disagree) to 7
(strongly agree). Survey participants were also asked to
Research attention is shifting from consumer product identify any other trends that they believed should be
evaluation in the purchase decision to include satisfaction included, but no additional trends were mentioned by
336
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.17
2016;33(3):320–341 2015;••(••):••–••

Table 2. Illustrative Research Opportunities Derived from Current and Emerging Technological and Sociocultural
Trends
Trends In the Context of Design Illustrative Design Implications Illustrative Research Opportunities

Sustainability A concern for the direct and indirect Extending product life by developing Understanding how to balance the need for
effects of product purchase, usage, products easier to upgrade and repair, proprietary and differentiated products
and disposal with respect to e.g., modular products that make it while adopting standards that enable
environmental and social issues possible to upgrade a product’s modular products (1.2, 1.3, 2.3)
Examples: Nike “Considered” shoes, capabilities (robustness), e.g., cell Determining how to integrate principles of
Toyota Prius phone’s software and chips rather than sustainable design within traditional
purchasing a new product Stage-Gate development processes (2.1,
Responsibly sourcing to protect 2.2, 2.3, 2.4)
communities and eliminate
environmental degradation
Sharing Exemplified by collaborative Development of more durable and reliable Defining how to simulate product usage and
economy consumption, whereby consumers products given heavier usage measure product quality assuming multiple
purchase access to products rather Product design that enables usage of the users of the same product (2.3, 3.4)
than owning them in order to gain same product by different customers Identifying decision rules that optimize design
desired utility of usage without with different needs and preferences trade-offs between product simplicity and
incurring the full costs of ownership ease of use while enabling different users
Examples: Zipcar, Airbnb and usage contexts (2.1, 2.2)
Internet of A network system that enables Development of products that self-monitor Defining design processes and choice models
Things information sharing and interaction and report on the status of the device that consider a significantly expanded set of
between devices and appliances in and/or the user technical possibilities, many of which are
the home, the office, as well as Development of devices that collectively external to and uncontrollable by the firm
mobile devices optimize use of shared resources, such (1.3, 2.2, 2.3)
Examples: Nest thermostat, GPS with as traffic management on highways Understanding how consumers will relate and
live traffic updates respond to products with increasing levels
of built-in intelligence (3.3, 3.4)
3D printing Use of additive manufacturing Development of parts and products that Exploring how design processes could be
technology that creates specifically take advantage of 3D modified to rapidly and cost-effectively
three-dimensional objects in printing technology develop accurate prototypes, e.g., toward
virtually any shape using an Development of product platforms that more iterative development (2.2, 2.3, 2.4)
increasingly diverse set of materials enable “micro-production” of small Understanding how to develop user design
Examples: cell phone case, prosthetic production runs or customized toolkits so that consumers can design and
arm individual products make products (2.4, 3.3)
Ease of quick, quality prototyping and Effects of better, cheaper prototypes and
user feedback subsequent feedback on iteration and
product acceptance.(2.2, 3.4)
Experience Conceived as the next evolution, Definition and development of fully Understanding when and how to use
economy following the “service economy,” in integrated experiences that are delivered qualitative research to gain insight into
which fully integrated experiences over a period of time customer needs—cognitive and
are “staged” using goods and/or Procurement of additional components of affective—with respect to experiences, and
services as enablers of the an integrated experience, beyond the using these insights to define “experience
experience firm’s traditional boundaries specifications” for integrated experience
Examples: Coffee house, Disney development (1.1, 2.1)
Cruise Understanding how to design an experience,
which could be a combination of product
and service (2.1, 3.3)
Health and Use of personal sensors to collect and Development of software to enable data Determining when and how to use open
performance analyze health and performance collection and analysis of data across versus proprietary data standards (1.2, 1.3)
monitoring information. users and platforms Understanding and responding to consumer
Examples: Polar heart monitor, Integration of sensor functionality into concerns about data use and privacy (3.3,
Nike + fuelband existing products, e.g., clothing, contact 3.4)
lenses, bandages
Social media Online software and networks for Using user-provided information, such as Understanding how to collect and synthesize
consumers that enable the creation product reviews, to improve existing online, user-generated information about
and exchange of information products and to inspire new products products to inform new product design
including user-generated content Development of products that employ efforts (1.1, 2.1)
Examples: Instagram, Facebook social media as a product feature, such Exploring the effects of widespread adoption
as cell phones and use of social media on users’
experiences in the home and workplace in
an effort to identify emerging product and
feature needs (2.1, 3.4)
Numbers following research opportunities indicate the relevant subtopic(s) within the conceptual model of product design research (see Figure 1).
PRODUCT DESIGN
18 RESEARCH
J PROD INNOV MANAG J PROD INNOV MANAG 337
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

more than two participants and therefore the analysis Arora, N., G. M. Allenby, and J. L. Ginter. 1998. A hierarchical Bayes
model of primary and secondary demands. Marketing Science 17 (1):
focused on the ratings of the initial 10 trends. 29–44.
Based on the ratings, seven trends were retained, iden- Arora, N., T. Henderson, and L. Qing. 2011. Noncompensatory dyadic
tified in Table 2, all of which had ratings above 5 (“some- choices. Marketing Science 30 (6): 1028–47.
what agree”). Each of these trends is listed and briefly Athaide, G. A., and R. R. Klink. 2009. Managing seller–buyer relationships
during new product development. Journal of Product Innovation Man-
described in Table 2, along with illustrative examples of agement 26 (5): 566–77.
potential implications, followed by related examples of Bailetti, A. J., and P. F. Litva. 1995. Integrating customer requirements into
potential research opportunities to consider. For example, product designs. Journal of Product Innovation Management 12 (1):
3–15.
based on the trend toward a greater concern about
Baiman, S., P. E. Fischer, and M. V. Rajan. 2001. Performance measure-
sustainability, one implication is the need and opportu- ment and design in supply chains. Management Science 47 (1): 173–88.
nity to consider ways to significantly extend product life, Bajaj, A., S. Kekre, and K. Srinivasan. 2004. Managing NPD: Cost and
such as development of easily upgradeable, modular schedule performance in design and manufacturing. Management
Science 50 (40): 527–36.
electronics—which is, indeed, currently being explored
Balakrishnan, P. V., and V. S. Jacob. 1996. Genetic algorithms for product
by Google and a host of other companies. A research design. Management Science 42 (8): 1105–17.
opportunity to consider, based on this trend, is the devel- Balka, K., C. Raasch, and C. Herstatt. 2014. The effect of selective open-
opment of an understanding of how to balance the need ness on value creation in user innovation communities. Journal of
Product Innovation Management 31 (2): 392–407.
for proprietary and differentiated products while adopting Barry, E. J., C. F. Kemerer, and S. A. Slaughter. 2006. Environmental
standards that enable modular products. Table 2 also volatility, development decisions, and software volatility: A longitudi-
notes how each of these research opportunities relates to nal analysis. Management Science 52 (3): 448–64.
the topics within our conceptual model of product design Baumgartner, H., and R. Pieters. 2000. The influence of marketing journals:
A citation analysis of the discipline and its sub-areas. CiteSeer.
(Figure 1). Available at: https://pure.uvt.nl/portal/files/537586/123.pdf.
Overall, Table 2 provides some specific ideas that may Bhattacharya, S., and V. Krishnan. 1998. Managing new product definition
inspire new research projects. There are many other in highly dynamic environments. Management Science 44 (11): S50–
64.
important trends that could be considered, however. Fur- Bhuiyan, N., D. Gerwin, and V. Thomson. 2004. Simulation of the new
thermore, it is not the authors’ intent to advocate research product development process for performance improvement. Manage-
that is exclusively defined by, or bound by, industry ment Science 50 (12): 1690–703.
trends. Instead, our intent is to illustrate a thought process Bloch, P. H. 1995. Seeking the ideal form: Product design and consumer
response. Journal of Marketing 59 (3): 16–29.
whose purpose is to provide a complementary way of Bloch, P. H., F. F. Brunel, and T. J. Arnold. 2003. Individual differences in
identifying new research opportunities; one that is the centrality of visual product aesthetics: Concept and measurement.
inspired both by gaps in research, i.e., looking inward and Journal of Consumer Research 29 (4): 551–65.
across research domains, and by trends in industry, i.e., Bogers, M., and W. Horst. 2014. Collaborative prototyping: Cross-
fertilization of knowledge in prototype-driven problem solving.
looking forward and externally. A balanced approach like Journal of Product Innovation Management 31 (4): 744–64.
this may, in turn, improve the potential impact of future Brown, T., and B. Katz. 2011. Change by design. Journal of Product
research, both on knowledge development and on indus- Innovation Management 28 (3): 381–83.
Cachon, G. P., and M. Olivares. 2010. Drivers of finished-goods inventory
try practice. in the U.S. automobile industry. Management Science 56 (1): 202–16.
Cachon, G. P., and R. Swinney. 2011. The value of fast fashion: Quick
References response, enhanced design, and strategic consumer behavior. Manage-
ment Science 57 (4): 778–95.
Abecassis-Moedas, C., and S. B. Mahmoud-Jouini. 2008. Absorptive
Camm, J. D., J. J. Cochran, D. J. Curry, and S. Kannan. 2006. Conjoint
capacity and source-recipient complementarity in designing new prod-
optimization: An exact branch-and-bound algorithm for the share-of-
ucts: An empirically derived framework. Journal of Product Innovation
choice problem. Management Science 52 (3): 435–47.
Management 25 (5): 473–90.
Adner, R., and D. Levinthal. 2001. Demand heterogeneity and technology Carbonell, P., A. I. Rodríguez-Escudero, and D. Pujari. 2009. Customer
evolution: Implications for product and process innovation. Manage- involvement in new service development: An examination of anteced-
ment Science 47 (5): 611–28. ents and outcomes. Journal of Product Innovation Management 26 (5):
536–50.
Allenby, G. M., and J. L. Ginter. 1995. Using extremes to design products
and segment markets. Journal of Marketing Research 32 (4): 392–403. Carson, S. J. 2007. When to give up control of outsourced new product
development. Journal of Marketing 71 (1): 49–66.
Amaldoss, W., and S. Jain. 2008. Research note-trading up: A strategic
analysis of reference group effects. Marketing Science 27 (5): 932–42. Carson, S. J., T. Wu, and W. L. Moore. 2012. Managing the trade-off
Andrews, R. L., A. Ansari, and I. S. Currim. 2002. Hierarchical Bayes between ambiguity and volatility in new product development. Journal
versus finite mixture conjoint analysis models: A comparison of fit, of Product Innovation Management 29 (6): 1061–81.
prediction, and partworth recovery. Journal of Marketing Research 39 Chen, C. 2001. Design for the environment: A quality-based model for
(1): 87–98. green product development. Management Science 47 (2): 250–63.
Antioco, M., R. K. Moenaert, and A. Lindgreen. 2008. Reducing ongoing Chen, K. D., and W. H. Hausman. 2000. Technical note: Mathematical
product design decision-making bias. Journal of Product Innovation properties of the optimal product line selection problem using choice-
Management 25 (6): 528–45. based conjoint analysis. Management Science 46 (2): 327–32.
338
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.19
2016;33(3):320–341 2015;••(••):••–••

Chernov, Y. 2012. Design of a state-of-the-art wheelchair. Available at: Fisher, M., K. Ramdas, and K. Ulrich. 1999. Component sharing in the
http://upcommons.upc.edu/pfc/bitstream/2099.1/13692/1/ management of product variety: A study of automotive braking
Mem%C3%B2ria.pdf. systems. Management Science 45 (3): 297–315.
Chitturi, R., R. Raghunathan, and V. Mahajan. 2008. Delight by design: The Franke, N., P. Keinz, and M. Schreier. 2008. Complementing mass
role of hedonic versus utilitarian benefits. Journal of Marketing 72 (3): customization toolkits with user communities: How peer input
48–63. improves customer self-design. Journal of Product Innovation Man-
agement 25 (6): 546–59.
Chiva, R., and J. Alegre. 2009. Investment in design and firm performance:
The mediating role of design management. Journal of Product Inno- Franke, N., P. Keinz, and C. J. Steger. 2009. Testing the value of
vation Management 26 (4): 424–40. customization: When do customers really prefer products tailored to
their preferences? Journal of Marketing 73 (5): 103–21.
Cohen, M. A., and S. Whang. 1997. Competing in product and service: A
product life-cycle model. Management Science 43 (4): 535–44. Franke, N., and F. Piller. 2004. Value creation by toolkits for user innova-
tion and design: The case of the watch market. Journal of Product
Cox, D. A., and A. D. Cox. 2002. Beyond first impressions: The effects of Innovation Management 21 (6): 401–15.
repeated exposure on consumer liking of visually complex and simple
product designs. Marketing Science 30 (2): 119–30. Franke, N., and M. Schreier. 2010. Why customers value self-designed
products: The importance of process effort and enjoyment. Journal of
Creusen, M. E. H., and J. P. L. Schoormans. 2005. The different roles of Product Innovation Management 27 (7): 1020–31.
product appearance in consumer choice. Journal of Product Innovation
Franke, N., M. Schreier, and U. Kaiser. 2010. The “I designed it myself”
Management 22 (1): 63–81.
effect in mass customization. Management Science 56 (1): 125–40.
Dahan, E., and H. Mendelson. 2001. An extreme-value model of concept Fuchs, C., and M. Schreier. 2011. Customer empowerment in new product
testing. Management Science 47 (1): 102–16. development. Journal of Product Innovation Management 28 (1):
Dahl, D. W. 2011. Clarity in defining product design: Inspiring research 17–32.
opportunities for the design process. Journal of Product Innovation Fuller, D. A., and J. A. Ottman. 2004. Moderating unintended pollution: the
Management 28 (3): 425–27. role of sustainable product design. Journal of Business Research 57
Dahl, D. W., A. Chattopadhyay, and G. J. Gorn. 1999. The use of visual (11): 1231–38.
mental imagery in new product design. Journal of Marketing Research Gemser, G., and M. A. A. M. Leenders. 2001. How integrating industrial
36 (1): 18–28. design in the product development process impacts on company per-
Dahl, D. W., and P. Moreau. 2002. The influence and value of analogical formance. Journal of Product Innovation Management 18 (1): 28–38.
thinking during new product ideation. Journal of Marketing Research Giloni, A., S. Seshadri, and C. L. Tucci. 2008. Neo-Rawlsian fringes: A new
39 (1): 47–60. approach to market segmentation and new product development.
Dangelico, R. M., P. Pontrandolfo, and D. Pujari. 2013. Developing sus- Journal of Product Innovation Management 25 (5): 491–507.
tainable new products in the textile and upholstered furniture indus- Goffin, K. 1998. Evaluating customer support during new product
tries: Role of external integrative capabilities. Journal of Product development—An exploratory study. Journal of Product Innovation
Innovation Management 30 (4): 642–58. Management 15 (1): 42–56.
Datar, S., C. Jordan, S. Kekre, S. Rajiv, and K. Srinivasan. 1996. New Gokpinar, B., W. J. Hopp, and S. M. Iravani. 2010. The impact of misalign-
product development structures: The effect of customer overload on ment of organizational structure and product architecture on quality in
post-concept time to market. Journal of Product Innovation Manage- complex product development. Management Science 56 (3): 468–84.
ment 13 (4): 325–33. Gosh, M., S. Dutta, and S. Stremersch. 2006. Customizing complex prod-
Day, G. S., and R. Wensley. 1988. Assessing advantage: A framework for ucts: When should the vendor take control? Journal of Marketing
diagnosing competitive superiority. Journal of Marketing 52 (2): 1–20. Research 43 (4): 664–79.
Dell’Era, C., and R. Verganti. 2007. Strategies of innovation and imitation Haggblom, T., R. J. Calantone, and C. A. Di Benedetto. 1995. Do new
of product languages. Journal of Product Innovation Management 24 product development managers in large or high-market-share firms
(6): 580–99. perceive marketing-R&D interface principles differently? Journal of
Product Innovation Management 12 (4): 323–33.
Desai, P., S. Kekre, S. Radhakrishnan, and K. Srinivasan. 2001. Product
Hertenstein, J. H., M. B. Platt, and R. W. Veryzer. 2005. The impact of
differentiation and commonality in design: Balancing revenue and cost
industrial design effectiveness on corporate financial performance.
drivers. Management Science 47 (1): 37–52.
Journal of Product Innovation Management 22 (1): 3–21.
Dickinson, J. R., and C. P. Wilby. 1997. Concept testing with and without Hienerth, C., and C. Lettl. 2011. Exploring how peer communities enable
product trial. Journal of Product Innovation Management 14 (2): 117– lead user innovations to become standard equipment in the industry:
25. Community pull effects. Journal of Product Innovation Management
Dickson, P., W. Schneier, P. Lawrence, and R. Hytry. 1995. Managing 28 (s1): 175–95.
design in small high-growth companies. Journal of Product Innovation Hildebrand, C., G. Haubl, and A. Herrmann. 2014. Product customization
Management 22 (1): 42–62. via starting solutions. Journal of Marketing Research 51 (6): 707–25.
Eppinger, S. 2011. The fundamental challenge of product design. Journal Hoegg, J., and J. W. Alba. 2011. Seeing is believing (too much): Design bias
of Product Innovation Management 28 (3): 399–400. in perceived product functionality. Journal of Product Innovation Man-
Erat, S., and S. Kavadias. 2008. Sequential testing of product designs: agement 28 (3): 346–59.
Implications for learning. Management Science 54 (5): 956–68. Hoffman, D. L., P. K. Kopalle, and T. P. Novak. 2010. The “right” consum-
Esslinger, H. 2011. Sustainable design: Beyond the innovation-driven busi- ers for better concepts: Identifying consumers high in emergent nature
ness model. Journal of Product Innovation Management 28 (3): 401–4. to develop new product concepts. Journal of Marketing Research 47
(5): 854–65.
Ethiraj, S. K., and D. Levinthal. 2004. Modularity and innovation in
Jain, S., and P. Kannan. 2002. Pricing of information products on online
complex systems. Management Science 50 (2): 159–73.
servers: Issues, models, and analysis. Management Science 48 (9):
Ettlie, J. E. 1995. Product-process development integration in manufactur- 1123–42.
ing. Management Science 41 (7): 1224–37. Jensen, M. B., C. Hienerth, and C. Lettl. 2014. Forecasting the commercial
Feit, E. M., M. A. Beltramo, and F. M. Feinberg. 2010. Reality check: attractiveness of user-generated designs using online data: An empirical
Combining choice experiments with market data to estimate the impor- study within the LEGO user community. Journal of Product Innovation
tance of product attributes. Management Science 56 (5): 785–800. Management 31 (S1): 75–93.
PRODUCT DESIGN
20 RESEARCH
J PROD INNOV MANAG J PROD INNOV MANAG 339
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

Jespersen, K. R. 2012. Stage-to-stage information dependency in the NPD Liechty, J., V. Ramaswamy, and S. H. Cohen. 2001. Choice menus for mass
process: Effective learning or a potential entrapment of NPD gates? customization: An experimental approach for analyzing customer
Journal of Product Innovation Management 29 (2): 257–74. demand with an application to a web-based information service.
Joglekar, N. R., A. A. Yassine, S. D. Eppinger, and D. E. Whitney. 2001. Journal of Marketing Research 38 (2): 183–96.
Performance of coupled product development activities with a dead- Liker, J. K., P. D. Collins, and F. M. Hull. 2003. Flexibility and standard-
line. Management Science 47 (12): 1605–20. ization: Test of a contingency model of product design manufacturing
John, G., A. M. Weiss, and S. Dutta. 1999. Marketing in technology- integration. Journal of Product Innovation Management 16 (3): 248–
intensive markets: Toward a conceptual framework. The Journal of 67.
Marketing 63 (1): 78–91. Loch, C. H., C. Terwiesch, and S. Thomke. 2001. Parallel and sequential
Kahn, B. E. 1998. Dynamic relationships with customers: High-variety testing of design alternatives. Management Science 47 (5): 663–78.
strategies. Marketing Science 26 (1): 45–53. Loginova, O. 2012. Competitive effects of mass customization. Review of
Marketing Science 10 (1): 1–32.
Kalogerakis, K., C. Lüthje, and C. Herstatt. 2010. Developing innovations
based on analogies: Experience from design and engineering consul- Luchs, M. G., J. Brower, and R. Chitturi. 2012. Product choice and the
tants. Journal of Product Innovation Management 27 (3): 418–36. importance of aesthetic design given the emotion-laden trade-off
between sustainability and functional performance. Journal of Product
Kaplan, A. M., and M. Haenlein. 2006. Toward a parsimonious definition of
Innovation Management 29 (6): 903–16.
traditional and electronic mass customization. Journal of Product Inno-
vation Management 23 (2): 168–82. Luchs, M. G., and K. S. Swan. 2011. Perspective: The emergence of
product design as a field of marketing inquiry. Journal of Product
Kaplan, A. M., D. Schoder, and M. Haenlein. 2007. Factors influencing the
Innovation Management 28 (3): 327–45.
adoption of mass customization: The impact of base category consump-
tion frequency and need satisfaction. Journal of Product Innovation 24 Luo, L. 2011. Product line design for consumer durables: An integrated
(2): 101–16. marketing and engineering approach. Journal of Marketing Research
48 (1): 128–39.
Karjalainen, T., and D. Snelders. 2010. Designing visual recognition for the
brand. Journal of Product Innovation Management 27 (1): 6–22. Luo, L., P. K. Kannan, B. Besharati, and S. Azarm. 2005. Design of robust
new products under variability: Marketing meets design. Journal of
Karlsson, C., R. Nellore, and K. Soderquist. 1998. Black box engineering: Product Innovation Management 22 (2): 177–92.
Redefining the role of product specifications. Journal of Product Inno-
vation Management 15 (6): 534–49. Luo, L., P. K. Kannan, and B. T. Ratchford. 2008. Incorporating subjective
characteristics in product design and evaluations. Journal of Marketing
Katz, G. M. 2004. Practitioner note: A response to Pullman et al.’s. 2002. Research 45 (2): 182–94.
Comparison of quality function deployment versus conjoint analysis.
Journal of Product Innovation Management 21 (1): 61–63. MacCormack, A., R. Verganti, and M. Iansiti. 2001. Developing products
on “internet time”: The anatomy of a flexible development process.
Kim, K., and D. Chhajed. 2001. An experimental investigation of valuation Management Science 47 (1): 133–50.
change due to commonality in vertical product line extension. Journal
of Product Innovation Management 18 (4): 219–30. Marion, T. J., and M. H. Meyer. 2011. Applying industrial design and cost
engineering to new product development in early-stage firms. Journal
Kornish, L. J., and K. T. Ulrich. 2011. Opportunity spaces in innovation: of Product Innovation Management 28 (5): 773–86.
Empirical analysis of large samples of ideas. Management Science 57
(1): 107–28. Marsh, S. J., and G. N. Stock. 2006. Creating dynamic capability: The role
of intertemporal integration, knowledge retention, and interpretation.
Kornish, L. J., and K. T. Ulrich. 2014. The importance of the raw idea in Journal of Product Innovation Management 23 (5): 422–36.
innovation: Testing the sow’s ear hypothesis. Journal of Marketing
Research 51 (1): 14–26. Meyer, M. H., and D. Dalal. 2002. Managing platform architectures and
manufacturing processes for nonassembled products. Journal of
Kreuzbauer, R., and A. J. Malter. 2005. Embodied cognition and new Product Innovation Management 19 (4): 277–93.
product design: Changing product form to influence brand categoriza-
tion. Journal of Product Innovation Management 22 (2): 165–76. Meyer, M. H., P. Tertzakian, and J. M. Utterback. 1997. Metrics for man-
aging research and development in the context of the product family.
Krider, R. E., P. Raghubir, and A. Krishna. 2001. Pizzas: π or square? Management Science 43 (1): 88–111.
Psychophysical biases in area comparisons. Marketing Science 20 (4):
405–25. Michalek, J. J., F. M. Feinberg, and P. Y. Papalambros. 2005. Linking
marketing and engineering product design decisions via analytical
Krishna, A., and M. Morrin. 2007. Does touch affect taste? The perceptual target cascading. Journal of Product Innovation Management 22 (1):
transfer of product container haptic cues. Journal of Consumer 42–62.
Research 34 (6): 807–18.
Micheli, P., J. Jaina, K. Goffin, F. Lemke, and R. Verganti. 2012. Percep-
Krishnan, V., and S. Gupta. 2001. Appropriateness and impact of platform- tions of industrial design: The “means” and the “ends.” Journal of
based product development. Management Science 47 (1): 52–68. Product Innovation Management 29 (5): 687–704.
Krishnan, V., and K. Ramachandran. 2011. Integrated product architecture Mikkola, J. H. 2006. Capturing the degree of modularity embedded in
and pricing for managing sequential innovation. Management Science product architectures. Journal of Product Innovation Management 23
57 (11): 2040–53. (1): 128–46.
Kuksov, D. 2004. Buyer search costs and endogenous product design. Monteverde, K. 1995. Technical dialog as an incentive for vertical integra-
Marketing Science 23 (4): 490–99. tion in the semiconductor industry. Management Science 41 (10):
Landwehr, J. R., A. McGill, and A. Herrmann. 2011. It’s got the look: The 1624–39.
effect of friendly and aggressive “facial” expressions on product liking Moon, H., D. R. Miller, and S. H. Kim. 2013. Product design innovation
and sales. Journal of Marketing 75 (3): 132–46. and customer value: Cross-cultural research in the United States and
Lau, A. K. W., R. C. M. Yam, and E. Tang. 2011. The impact of product Korea. Journal of Product Innovation Management 30 (1): 31–43.
modularity on new product performance: Mediation by product Moore, W. L., J. J. Louviere, and R. Verma. 1999. Using conjoint analysis
innovativeness. Journal of Product Innovation Management 28 (2): to help design product platforms. Journal of Product Innovation Man-
270–84. agement 16 (1): 27–39.
Lee, J., and F. M. Veloso. 2008. Inter-firm innovation under uncertainty: Moorman, C., and R. J. Slotegraaf. 1999. The contingency value of comple-
Empirical evidence for strategic knowledge-partitioning. Journal of mentary capabilities in product development. Journal of Marketing
Product Innovation Management 25 (5): 413–525. Research 36 (2): 239–57.
340
PRODUCT J PROD
DESIGN INNOV MANAG
RESEARCH J PROD INNOV MANAGM. G. LUCHS ET AL.21
2016;33(3):320–341 2015;••(••):••–••

Moreau, C. P. 2011. Inviting the amateurs into the studio: Understanding Smith, R. P., and S. D. Eppinger. 1997. A predictive model of sequential
how consumer engagement in product design creates value. Journal of iteration in engineering design. Management Science 43 (8): 1104–20.
Product Innovation Management 28 (3): 409–10. Souder, W. E., and S. A. Jenssen. 1999. Management practices influencing
Moreau, C. P., L. Bonney, and K. B. Herd. 2011. It’s the thought (and the new product success and failure in the United States and Scandinavia:
effort) that counts: How customizing for others differs from customiz- A cross-cultural comparative study. Journal of Product Innovation
ing for oneself. Journal of Marketing 75 (5): 120–33. Management 16 (2): 183–203.
Muffatto, M., and R. Panizzolo. 1996. Innovation and product development Souder, W. E., and M. Song. 1997. Contingent product design and market-
strategies in the Italian motorcycle industry. Journal of Product Inno- ing strategies influencing new product success and failure in U.S. and
vation Management 13 (4): 348–61. Japanese electronics firms. Journal of Product Innovation Management
Mugge, R., and D. W. Dahl. 2013. Seeking the ideal level of design 14 (1): 21–34.
newness: Consumer response to radical and incremental product Srinivasan, R., G. L. Lilien, and A. Rangaswamy. 2006. The emergence of
design. Journal of Product Innovation Management 30 (S1): 34–47. dominant designs. Journal of Marketing 70 (2): 1–17.
Nambisan, S. 2002. Complementary product integration by high-
Srinivasan, R., G. L. Lilien, A. Rangaswamy, G. M. Pingitore, and D.
technology new ventures: The role of initial technology strategy. Man-
Seldin. 2012. The total product design concept and an application to the
agement Science 48 (3): 382–98.
auto market. Journal of Product Innovation Management 29 (S1):
Netessine, S., and T. A. Taylor. 2007. Product line design and production 3–20.
technology. Marketing Science 26 (1): 101–17.
Srinivasan, V., W. S. Lovejoy, and D. Beach. 1997. Integrated product
Noble, C. H., and M. Kumar. 2010. Exploring the appeal of product design: design for marketability and manufacturing. Journal of Marketing
A grounded, value-based model of key design elements and relation- Research 34 (1): 154–63.
ships. Journal of Product Innovation Management 27 (5): 640–57.
Staudenmayer, N., M. Tripsas, and C. L. Tucci. 2005. Interfirm modularity
Novak, S., and S. D. Eppinger. 2001. Sourcing by design: Product com- and its implications for product development. Journal of Product Inno-
plexity and the supply chain. Management Science 47 (1): 189–204. vation Management 22 (4): 303–21.
Oppenheimer, A. 2005. From experience: Products talking to people—
Conversation closes the gap between products and consumers. Journal Swan, K. S., M. Kotabe, and B. B. Allred. 2005. Exploring robust design
of Product Innovation Management 22 (1): 82–91. cap abilities, their role in creating global products, and their relation-
ship to firm performance. Journal of Product Innovation Management
Orth, U. R., and K. Malkewitz. 2008. Holistic package design and con- 22 (2): 144–64.
sumer brand impressions. Journal of Marketing 72 (3): 64–81.
Syam, N. B., and N. Kumar. 2006. On customized goods, standard goods,
Poetz, M. K., and M. Schreier. 2012. The value of crowdsourcing: Can and competition. Marketing Science 25 (5): 525–37.
users really compete with professionals in generating new product
ideas? Journal of Product Innovation Management 29 (2): 245–56. Syam, N. B., R. Ruan, and J. D. Hess. 2005. Customized products: A
competitive analysis. Marketing Science 24 (4): 569–84.
Pullman, M. E., W. L. Moore, and D. G. Wardell. 2002. A comparison of
quality function deployment and conjoint analysis in new product Talke, K., S. Salomo, J. E. Wieringa, and A. Lutz. 2009. What about design
design. Journal of Product Innovation Management 19 (5): 354–64. newness? Investigating the relevance of a neglected dimension of
Radford, S. K., and P. H. Bloch. 2011. Linking innovation to design: product innovativeness. Journal of Product Innovation Management 26
Consumer responses to visual product newness. Journal of Product (6): 601–15.
Innovation Management 28 (S1): 208–20. Terwiesch, C., and C. H. Loch. 2004. Collaborative prototyping and the
Raghubir, P., and E. A. Greenleaf. 2006. Ratios in proportion: What should pricing of custom-designed products. Management Science 50 (2):
the shape of the package be? Journal of Marketing 70 (2): 95–107. 145–58.
Raja, J. Z., D. Bourne, K. Goffin, M. Çakkol, and V. Martinez. 2013. Thomke, S. H. 1998. Managing experimentation in the design of new
Achieving customer satisfaction through integrated products and ser- products. Management Science 44 (6): 743–62.
vices: An exploratory study. Journal of Product Innovation Manage- Thompson, D. V., R. W. Hamilton, and R. T. Rust. 2005. Feature fatigue:
ment 30 (6): 1128–44. When product capabilities become too much of a good thing. Journal
Randall, T., C. Terwiesch, and K. T. Ulrich. 2007. User design of custom- of Marketing Research 42 (4): 431–42.
ized products. Marketing Science 26 (2): 268–80. Tidd, J. 1995. Development of novel products through intraorganizational
Rosa, J. A., W. J. Qualls, and J. A. Ruth. 2014. Consumer creativity: Effects and interorganizational networks. Journal of Product Innovation Man-
of gender and variation in the richness of vision and touch inputs. agement 12 (4): 307–22.
Journal of Business Research 67 (3): 386–93.
Tomkovick, C., and K. E. Dobie. 1995. Applying hedonic pricing models
Rosenthal, S. R., and M. Capper. 2006. Ethnographies in the front end: and factorial surveys at Parker Pen to enhance new product success.
Designing for enhanced customer experiences. Journal of Product Journal of Product Innovation Management 12 (4): 334–45.
Innovation Management 23 (3): 215–37.
Truong, Y., R. R. Klink, L. Fort-Rioche, and G. A. Athaide. 2013. Con-
Sándor, Z., and M. Wedel. 2005. Heterogeneous conjoint choice designs. sumer response to product form in technology-based industries.
Journal of Marketing Research 42 (2): 210–18. Journal of Product Innovation Management 31 (4): 867–76.
Sanchez, R. 1999. Modular architectures in the marketing process. Journal
Ulrich, K. T. 2011. Design is everything? Journal of Product Innovation
of Marketing 63 (4): 92–111.
Management 28 (3): 394–98.
Schreier, M., C. Fuchs, and D. W. Dahl. 2012. The innovation effect of user
design: Exploring consumers’ innovation perceptions of firms selling Ulrich, K. T., and D. J. Ellison. 1999. Holistic customer requirements and
products designed by users. Journal of Marketing 76 (5): 18–32. the design-select decision. Management Science 45 (5): 641–58.
Shi, L. Y., S. Olafsson, and Q. Chen. 2001. An optimization framework for Ulrich, K. T., and S. Pearson. 1998. Assessing the importance of design
product design. Management Science 47 (12): 1681–92. through product archaeology. Management Science 44 (3): 352–69.
Simonson, I. 2005. Determinants of customers’ responses to customized Ülkü, S., C. V. Dimofte, and G. M. Schmidt. 2012. Consumer valuation of
offers: Conceptual framework and research propositions. Journal of modularly upgradeable products. Management Science 58 (9): 1761–
Marketing 69 (1): 32–45. 76.
Sköld, M., and C. Karlsson. 2013. Stratifying the development of product von Hippel, E. 1998. Economics of product development by users: The
platforms: Requirements for resources, organization, and management impact of “sticky” local information. Management Science 44 (5):
styles. Journal of Product Innovation Management 30 (S1): 62–76. 629–44.
PRODUCT DESIGN
22 RESEARCH
J PROD INNOV MANAG J PROD INNOV MANAG 341
M. G. LUCHS ET AL.
2015;••(••):••–•• 2016;33(3):320–341

von Hippel, E., and E. Katz. 2002. Shifting innovation to users via toolkits. Viswanathan, M., and S. Sridharan. 2012. Product development for the
Management Science 48 (7): 821–33. BOP: Insights on concept and prototype development from university-
Valenzuela, A., R. Dhar, and F. Zettelmeyer. 2009. Contingent to self- based student projects in India. Journal of Product Innovation Man-
customization procedures: Implications for decision satisfaction and agement 29 (1): 52–69.
choice. Journal of Marketing Research 46 (6): 754–63. Vriens, M., G. H. Loosschilder, E. Rosbergen, and D. R. Wittink. 1998.
Van der Lans, R., R. Pieters, and M. Wedel. 2008. Competitive brand Verbal versus realistic pictorial representation in conjoint analysis with
salience. Marketing Science 27 (5): 922–31. design attributes. Journal of Product Innovation Management 15 (5):
455–67.
Verganti, R. 2008. Design, meanings, and radical innovation: A metamodel
and a research agenda. Journal of Product Innovation Management 25 Wasti, S. N., and J. K. Liker. 1997. Risky business or competitive power?
(5): 436–56. Supplier involvement in Japanese product design. Journal of Product
Innovation Management 14 (5): 337–55.
Veryzer, R. W. 2005. The roles of marketing and industrial design in
discontinuous new product development. Journal of Product Innova- Williams, N., P. K. Kanna, and S. Azarm. 2011. Retail channel structure
tion Management 22 (1): 22–41. impact on strategic engineering product design. Management Science
57 (5): 897–914.
Veryzer, R. W., and B. B. de Mozota. 2005. The impact of user-oriented
design on new product development: An examination of fundamental World Commission on Environment and Development. 1987. Our common
relationships. Journal of Product Innovation Management 22 (2): 128– future. Oxford: Oxford University Press.
43. Zhang, D., P. Hu, and M. Kotabe. 2011. Marketing-industrial design inte-
Veryzer, R. W., and J. W. Hutchinson. 1998. The influence of unity and gration in new product development: The case of China. Journal of
prototypicality on aesthetic responses to new product designs. Journal Product Innovation Management 28 (3): 360–73.
of Consumer Research 24 (4): 374–94. Ziamou, P. L., S. Gould, and A. Venkatesh. 2012. Am I getting it or not?”
Villas-Boas, J. M. 1998. Product line design for a distribution channel. The practices involved in “trying to consume” a new technology.
Marketing Science 17 (2): 156–70. Journal of Product Innovation Management 29 (2): 216–28.
 Chapter 14

The Advanced Turboprop Project:

Radical Innovation
in a Conservative Environment
by Mark D. Bowles and Virginia P. Dawson

In 1987, a Washington Post headline read, "The aircraft engine of the future has
propellers on it."1 To many this statement was something like heralding "the
reincarnation of silent movies."2 Why would an "old technology" ever be chosen over
a modem, new, advanced alternative? How could propeller technology ever supplant
the turbojet revolution? How could the 'Jet set mind-set" of corporate executives, who
demanded the prestige of speed and "image and status with a Jet," ever be satisfied
with a slow, noisy, propeller-driven aircraft? 3 A Washington Times correspondent
predicted that the turbojet would not be the propulsion system of the future. Instead,
the future would witness more propellers than Jets and if "Star Wars hero Luke
Skywalker ever became chairman of a Fortune 500 company, he would replace the
corporate jet with a ... turboprop."4 It appeared that a turboprop revolution was
underway.

NASA Lewis Research Center's Advanced Turboprop Project (1976-1987) was the
source of this optimism. The energy crisis of the early 1970s served as the catalyst for
renewed government interest in aeronautics and NASA launched this ambitious
project to return to fuel saving, propeller-driven aircraft. The Arab oil embargo
brought difficult times to all of America, but the airlines industry, in particular,
suffered and feared for its future in the wake of a steep rise in fuel prices. NASA
responded to these fears by creating a program to improve aircraft fuel efficiency. Of
the six projects NASA funded through this program, the Advanced Turboprop Project
promised the greatest payoffs in terms of fuel savings, but it was also the most
conceptually radical and technically demanding.

The project began in the early 1970s with the collaboration of two engineers, Daniel
Mikkelson from NASA Lewis, and Carl Rohrbach of Hamilton Standard, the nation's
last major propeller manufacturer. Mikkelson, then a young aeronautical research
engineer, went back to the old NACA wind tunnel reports where he found a "glimmer
of hope" that propellers could be redesigned to make propeller-powered aircraft fly
faster and higher than those of the mid to late-1950s.5 Mikkelson and Rohrbach came
up with the concept of sweeping the propeller blades to reduce noise and increase
efficiency and NASA received a joint patent with Hamilton Standard for the
development of this technology. At Lewis, Mikkelson sparked the interest of a small
cadre of engineers and managers. They solved key technical problems essential for the
creation of the turboprop, while at the same time they attracted support for the project.
After a project office was established, they became political advocates, using
technical gains and increasing acceptance to fight for continued funding. This
involved winning government, industry, and public support

1. Martha M. Hamilton, "Firms Give Propellers a New Spin," Washington Post,

February 8, 1987.

2. Robert J. Serling, "Back to the Future with Propfans," USAIR (June 1987).

3. R.S. Stahr, Oral report on the RECAT study contract at NASA, April 22, 1976, Nored
papers, NASA, Lewis Research Center, box 224.

4. Hugh Vickery, "Turboprops are Back!," Washington Times, November 1, 1984, p.

5B.

5. Interview with Daniel Mikkelson, by Virginia Dawson and Mark Bowles, September
6, 1995.

321

322 THE ADVANCED TURBOPROP PROJECT

 An advanced propeller swirl recovery model is shown in the NASA Lewis
Research Centers 8 x 6 foot supersonic wind tunnel. Propeller
efficiencies and noise are measured at cruise much numbers up to 0. 80
and at takeoff and approach conditions. Vane pitch angles and propfan-
to-vane axial spacings are varied. The testing was part of the Advanced
Turboprop Project, with the goal of providing the technology base to
enable the U.S. development of quieter, fuel efficient turboprop
engines with a comfortable aircraft interior environment. (NASA photo
no. 90-H-78).

FROM ENGINEERING SCIENCE TO BIG SCIENCE 323

for the new propeller technology. Initially the project involved only Hamilton
Standard, but the aircraft engine manufacturers, Pratt & Whitney, Allison, and
General Electric, and the giants of the airframe industry, Boeing, Lockheed, and
McDonnell Douglas joined the bandwagon as the turboprop appeared to become more
and more technically and socialIy feasible. The turboprop project became a large,
well-funded, "heterogeneous collection of human and material resources" that
contemporary historians refer to as "big science."6 At its height it involved over forty
industrial contracts, fifteen university grants, and work at the four NASA research
centers, Lewis, Langley, Dryden, and Ames. The progress of the advanced turboprop
development seemed to foreshadow its future dominance of commercial flight.

The project had four technical stages: "concept development" from 1976 to 1978;
"enabling technology" from 1978 to 1980; "large-scale integration" from 1981 to
1987; and finally "flight research" in 1987.7 During each of these stages, NASA's
engineers confronted and solved specific technical problems that were necessary for
the advanced turboprop project to meet the defined government objectives concerning
safety, efficiency at high speeds, and environmental protection. NASA Lewis
marshaled the resources and support of the United States aeronautical community to
bring the development of the new technology to the point of successful flight testing.
In 1987, these NASA engineers, along with a wide-ranging industry team, won the
coveted Collier Trophy for developing a new fuel efficient turboprop propulsion
system.8 The winning team included Hamilton Standard, General Electric, Lockheed,
the Allison Gas Turbine Division of General Motors, Pratt & Whitney, Rohr
Industries, Gulfstream, McDonnell Douglas, and Boeing —certainly the largest, most
diverse group, to be so honored in the history of the prize.

Despite this technical success, the predicted turboprop revolution never came, and no
commercial or military air fleet replaced their jets with propellers. The reason for this
failure was socio-economic, not technical. Throughout the project, social issues
influenced and defined the status of the advanced turboprop. From the beginning it
was the perception of an energy crisis, not a technological innovation, that spurred the
idea of the project itself. The Cold War and the existence of Soviet high-speed
turboprops played a key role in convincing Congress to fund the project. As the
project progressed, within each technological stage, the engineers used distinctive and
creative approaches to deal with the complex web of government, industry, and
academic contractors. More often than not, the main question was not does the
technology work, but how can we get government, industry, and the public to accept
this technology? In the end it was a socioeconomic issue again which shelved the
program. The reduction of fuel prices ended the necessity for fuel conservation in the
skies and today the advanced turboprop remains a neglected, or "archived"
technology.
This is not to imply that the technical achievements were unimportant. Each distinct
technical stage of the project determined a corresponding social action. During the
concept development stage, creative advocacy was necessary to sell the government
and industry on this radical idea. During the enabling technology stage, engineers
used complex project management skills to ensure that this massive team would
function effectively. During the large-scale integration stage, NASA had to deal with
a competitor that surprised them by introducing its own high-speed turboprop. Finally,
during the flight research stage, NASA became aware that no current airlines would
adopt the advanced turboprop and thus the

6. See James H. Capshew and Karen A. Rader, "Big Science: Price to the
Present," Osiris, 2nd ser., 7 (1992): 3-25.

7. Roy D. Hager and Deborah Vrabel, Advanced Turboprop Project (Washington, DC:
NASA SP-495, 1988), p. 610.

8. Citation for the Collier Trophy in Roy D. Hager and Deborah Vrabel, p. vi.

324 THE ADVANCED TURBOPROP PROJECT

engineers waged a battle to win the Collier Trophy to try and gain positive status and
recognition for their technical achievement.

The relationship between these technical and social spheres was never either a
simplistic story of social construction or technological determinism. Rather, the
relationship was one of interdependence. At times the project advanced on its
technical merits; at others, it progressed through political persuasion. At each stage,
only after NASA engineers and their industrial and academic partners solved both the
social and technical problems holding it back, was the advanced turboprop project
able to obtain funding and move forward. But ultimately, the socio-economic issue of
petroleum price and availability managed to scuttle NASA's technical success.

Thomas Hughes, a prominent historian of technology, has argued that the research and
development organizations of the twentieth century, no matter whether they are run by
a government, industry, or members of a university community, stifle technical
creativity.9 In these organizations there can be found "no trace of a flash of
genius."10 In contrast, the late 19th century for Hughes was the "golden era" of
invention —a time when the independent inventor flourished without institutional
constraints. Recently, David Hounshell has challenged Hughes's contention that
industrial research laboratories "exploit creative, inventive geniuses; they neither
produce nor nurture them."11 Not only can the industrial research laboratory nurture a
creative individual, but collectively, people engaged in research and development
contribute to making an invention a commercial reality. In his study of the
organization of research at Du Pont, Hounshell paid tribute to the individual brilliance
of the organic chemist Wallace H. Carothers, but he argued that the real "genius of
nylon was in the organization that developed it into one of the most successful and
profitable materials of the twentieth century."12 In our view, the NASA Advanced
Turboprop Project represents another case in which organizational capabilities, not
individual genius alone, create the opportunity for significant innovation. The
organization that supported the development of the turboprop was far more complex
than the research laboratory of an industrial firm, yet it responded to the energy crisis
to advance a radical idea. As Donald Nored, who headed the office at NASA Levis
Research Center that managed the three aircraft energy efficiency projects remarked,
"The climate made people do things that normally they'd be too conservative to
do."13 The history of the advanced turboprop demonstrates how a radical innovation
can emerge from a dense, conservative web of bureaucracy to nearly revolutionize the
world's aircraft propulsion systems.

The Conservative Team Environment

Although NASA won several Collier trophies for innovations related to the space
program, it had produced no winners in aeronautics since the founding of the agency
in 1958. NASA's predecessor organization, the National Advisory Committee for
Aeronautics (NACA), had received five Collier trophies for contributions to
aeronautics between 1929 and 1958. These trophies paid tribute to the individual
creativity and the unique research environment of the NACA's research laboratories.
James R. Hansen has described in this volume how engineer Fred E. Weick used the
NACA's unique wind tunnel facilities to develop

9. Thomas P. Hughes, American Genesis: p. 54.

10. Ibid., p. 183.

11. David A. Hounshell, "Hughesian History of Technology and Chandlerian Business
History: Parallels, Departures, and Critics," History and Technology 12 (1995): 217.

12. Ibid. See also, David A. Hounshell and John Kenly Smith, Jr., Science and
Corporate Strategy: Do Pont R&D, 1902-1980 (Cambridge, MA: Cambridge University
Press, 1988).

13. Interview with Donald Noted at Case Western Reserve University by Virginia
Dawson and Mark Bowles, August 15, 1995.

FROM ENGINEERING SCIENCE TO BIG SCIENCE 325

the NACA low-drag cowling. Succeeding Collier trophies awarded under the
institutional aegis of the NACA followed a similar pattern. Lewis A. Rodert won it for
developing a thermal ice prevention system for aircraft (see the essay by Glenn E.
Bugos, this volume), John Stack won it twice for his contributions to supersonic
theory and the development of the transonic wind tunnel, and Richard Whitcomb
carried off the prize for his discovery and empirical validation of the area rule. What
made the award garnered by the NASA/industry team in 1987 different was that it
recognized the collective talents of government engineers from four NASA research
centers, academic researchers, and contractors from the propeller, engine, airframe,
and airline industries.

The history of the turboprop project is interesting from an institutional standpoint

because it took root and flourished within NASA's conservative, bureaucratic
environment. It was modeled, not on NASA's small-scale aeronautical research
projects (typically carried on by former NACA laboratories), but on the large-scale
projects of the space program. The NASA Lewis Research Center adopted an
administratively complex team approach that depended on input not simply from
other NASA Centers, but also from numerous industrial and university contractors.
Essentially, NASA Lewis Research Center became the center of an extensive
government-industry-academic complex. At each stage in the project, the
management team determined what needed to be done and sought the appropriate help
both from within and outside NASA.

With its expertise in propulsion technology, the NASA Lewis Research Center was
ideally suited to manage the turboprop project. Set up in Cleveland, Ohio, during
World War II as an aircraft engine research laboratory, Lewis became the third
laboratory of the National Advisory Committee for Aeronautics. Lewis engineers
pursued aircraft engine research in the national interest—often over the objection of
the engine companies who perceived the government as interfering with the normal
forces of supply and demand. During the early years of the Cold War, the laboratory
participated in engine research and testing to assist the engine companies in
developing the turbojet engine. After the launch of Sputnik, the laboratory focused on
a new national priority-rocket propulsion research and development. Almost all work
on air-breathing engines ceased for nearly ten years.

The return to aircraft engine research coincided with drastic reductions in staff,
mandated by cuts in NASA's large-scale space programs.14 The mass exodus of nearly
800 personnel in 1972 sparked an effort to redefine the center's mission and find new
sources of funding. The following year, OPEC's oil embargo galvanized the Center's
director, Bruce Lundin, to look for ways to use its propulsion expertise to help solve
the energy crisis. In 1974, Lewis received $1.5 million for a wind-energy program
from the National Science Foundation and the Energy Research and Development
Administration (ERDA). A program in solar cell technology development followed on
its heels with increasing funding of various energy-related programs by ERDA and its
successor, the Department of Energy. The changing focus of the Center's activities
prompted rumors-emphatically denied-that it would become part of ERDA. The new
emphasis on energy efficient aircraft, unlike the ERDA projects, promised to keep
Lewis strongly in NASA's fold.15 Moreover, it brought high visibility to the
aeronautics side of NASA, long overshadowed by the enormous budgets and prestige
of the space program.

Although it shared similarities in management with NASA's space projects, the

turboprop project differed in significant ways. First, although the advanced turboprop
was the reincarnation of an old idea, it involved the creation of cutting-edge
technology. Space

14. Virginia P. Dawson, Engines and Innovation: Lewis Laboratory and American
Propulsion Technology (Washington, DC: NASA SP-4306, 1991).

15. Ibid.

326 THE ADVANCED TURBOPROP PROJECT

projects involved rigorous oversight, but generally relied on existing technology.
When necessary, NASA contracted with industry to produce whatever new
technology was needed for a particular mission. The turboprop project tapped the
creative talents of engineers at NASA in ways that were reminiscent of the NACA
tradition of in-house research, though in management scope it transcended the narrow
institutional boundaries of NASA's research centers. Second, though all NASA
projects of the early 1970s needed to be "sold" to an increasingly tight-fisted
Congress, the controversial nature of the turboprop meant that NASA Lewis had to
build support both at Headquarters and within the aviation community. What NASA
referred to as "advocacy" needed to be vigorous and continuous throughout the life of
the project.

The Energy Crisis and the Politics of Funding

The OPEC oil embargo of 1973 awakened the United States to the degree of control
outside nations had over the lives of every American. The increased price of oil
affected all areas of the economy, but none more than the airlines industry.16 Earl
Cook, noted geographer and geologist, has argued, "Whoever controls the energy
systems can dominate the society." 17 An extension of this argument is, whoever
possesses the fuel supply controls the energy systems. Five sources of energy,
including petroleum, natural gas, coal, hydropower, and nuclear, accounted for all fuel
consumption in the United States during 1973. Of these five sources, America was
most dependent upon petroleum, consuming approximately seventeen million barrels
of oil a day.18 At no other time in American history was Cook's aphorism more
evident than in 1973 when the United States imported six million barrels of oil a day,
64 percent of which came from the Organization of Petroleum Exporting Countries
(OPEC).19 The concern in the United States was that since OPEC controlled the
petroleum, could they dominate American society?

In response to the energy crisis, in 1973 the airlines industry initiated its own fuel-
saving program which reduced fuel consumption by over one billion gallons per
year.20 But these measures were not enough. jet fuel prices jumped from twelve cents
to over one dollar per gallon and total yearly fuel expenditures increased by one
billion dollars, or triple the earnings of the airlines. Prior to 1972, fuel accounted for
one-quarter of the commercial airlines' total direct operating costs.21 During the crisis,
fuel represented over half of the airlines' operating costs. The result was a reduction in
the number of flights, the grounding of some aircraft, and the "furloughing" of some
10,000 employees. If the situation in the early 1970s seemed bad, prospects for the
future appeared even worse. Linking the fate of the airlines, the cost of jet fuel and the
prosperity of the nation as a whole, airlines industry lobbyists rushed to their
congressmen. The politicians, in turn, appealed to NASA.
16. The Israeli victory during the Six-Day War in 1967 resulted in retaliation by OPEC.
Seeking to force a pro-Arab stance from the United States (Israel's ally), Saudi Arabia
imposed an American oil embargo concurrent with the quadrupling of oil prices from
the other OPEC nations. See Don Peretz, The Middle East Today 5th ed. (New York,
NY. Praeger, 1988), 154. Gary B. Nash, et al. The American People: Creating a Nation
and a Society 2d ed. (New York, NY: Harper & Row, Publishers, 1990), p. 971.

17. Earl Cook, Man, Energy, Society (San Francisco, CA: W. H. Freeman, 1976), p.
208.

18. A barrel contains 42 gallons.

19. In 1973, total U.S. crude oil imports totaled 1,184 million barrels, 765 of which
came from OPEC. The OPEC nations at that time included Algeria, Ecuador,
Indonesia, Iran, Iraq, Kuwait, Libya, Nigeria, Qatar, Saudi Arabia, United Arab
Emirates, and Venezuela. Statistical Abstract of the United States (Washington, DC:
U.S. Department of Commerce, 1994), p. 593.

20. Clifton F. Von Kann, testimony before the U.S. Senate, Committee on
Aeronautical and Space Sciences, September 10, 1975, p. 4.

21. Donald L. Noted, John B. Whitlow, Jr., William C. Strack, "Status Update of the
NASA Advanced Turboprop Project," unpublished report, Nored private papers.

FROM ENGINEERING SCIENCE TO BIG SCIENCE 327

Why was jet fuel so important to our national interest? Clifton F. Von Kann, senior
vicepresident of the Air Transport Association of America, pointed out in a 1975
Senate statement that airlines were "more than just another means of transportation
."22 He asserted they played a major part in the economic and military success of the
nation. They also

22. Kann testimony, p. 3.

328 THE ADVANCED TURBOPROP PROJECT

provided the infrastructure for the mail system, the national export system, and the
$60 billion tourist industry. jet fuel was the "life-blood" of the airlines, but it was also
their Achilles heel. He warned a failure to control the rising cost of fuel might result
in either the nationalization or the withering away of the "basic building block in the
structure of the U.S. economy."23 Senator Barry Goldwater linked this crisis to the
possible "loss of a large part of our world supremacy."24 The fuel crisis created an
opportunity for NASA at a time when Congress had drastically cut funding for the
space program. Aeronautics, the first "A" in NASA, had long taken a back seat to the
spectacular space missions of the Apollo years. Now the agency was ready to reassert
its role as the nation's premier institution for research and development in civil
aeronautics.

In January 1975, James Fletcher, the NASA Administrator, received a letter from
Senators Barry Goldwater and Frank Moss. 25 The letter suggested a massive
technology project involving NASA and industry to help ease the burden on the
airlines caused by the energy crisis. Its goal was the realization of a new generation of
fuel-efficient aircraft. Goldwater and Moss asked NASA to propose a plan, develop
the technology, and facilitate the "technology transfer process" to
industry.26 Technology transfer later became a particularly thorny issue in the debate
over whether the government should carry development to the point of costly flight
testing, or leave that phase to the manufacturers who stood to benefit handsomely
from this government-generated technology.
In February 1975, NASA formed the Intercenter Aircraft Fuel Conservation
Technology Task Force to explore all potential options .27 Sixteen government
scientists and engineers from NASA, the Department of Transportation, the Federal
Aviation Administration, and the Department of Defense took part in the seven-month
study. 28 James Kramer, the task force leader, called for any new ideas that would
satisfy government criteria, even those that might be considered "unusual." The task
force defined six major areas with the potential for significant impact on aircraft fuel
efficiency. It recommended the creation within NASA of the Aircraft Energy
Efficiency (ACEE) Program, the administrative umbrella for six new aeronautics
projects —three related to the airframe and three to the propulsion system.29

NASA assigned management of the three propulsion projects to the NASA Lewis
Research Center. The first of these propulsion projects focused on improving existing
turbofan engines through the redesign of selected engine components. It was the least
technically challenging of the three projects and aimed for a five percent increase in
fuel efficiency within a few years. The second project, the Energy Efficient Engine
(E3), involved building 11 a brand new engine from scratch" and offered a far greater
payoff —an increase in fuel efficiency of ten to fifteen percent. In essence, NASA
proposed to assume the risk for developing an "all new technology in an all up
engine."30 With a new "recoupment program" in place, the government expected to
get back some of its investment out of the profits of the engine manufacturers,
General Electric and Pratt & Whitney.

23. Ibid., p. 5.

24. Senator Goldwater's response to Kahn, ibid., p. 8.

25. It was likely that the NASA staff drafted the letter.

26. Barry Goldwater and Frank Moss to James C. Fletcher, as found in, Aircraft Fuel
Conservation Technology, task force report, September 10, 1975, pp. 138-39.

27. Roy D. Hager and Deborah Vrabel, Advanced Turboprop Project (Washington, DC:
NASA, 1988), p. 4.

28. Aircraft Fuel Conservation Technology, task force report, September 10, 1975,
pp. 1 and 2,

29. The Aircraft Energy Efficiency (ACEE) Program airframe projects included: the
Fuel Conservative Transport to improve on aerodynamic design and potentially save
fifteen to twenty percent in fuel use; the Composite Primary Structures which would
decrease the weight of aircraft through the use of composite materials and save 10
to 15 percent in fuel costs compared to an all metal airplane; and Laminar Flow
Control to allow an aircraft to maintain low drag, thus creating a potential fuel
savings of twenty to forty percent.

30. Nored interview.

FROM ENGINEERING SCIENCE TO BIG SCIENCE 329

In contrast to these two relatively conservative projects, the advanced turboprop

offered dramatic increases in fuel efficiency. NASA planners believed that an
advanced turboprop could reduce fuel consumption by twenty to thirty percent over
existing turbofan engines with comparable performance and passenger comfort at
speeds up to Mach 0.8 and altitudes up to 30,000 feet. (It should be noted that
commuter turboprop-powered aircraft in current use fly at far slower speeds and lower
altitudes.) The ambitious goals of the turboprop project made it controversial and
challenging both from a technical and social point of view. Technically, studies by
Boeing, McDonnell Douglas, and Lockheed pointed to four areas of concern:
propeller efficiency at cruise speeds, both internal and external noise problems,
installation aerodynamics, and maintenance costs.31Socially, the turboprop also
presented daunting problems. Because of the "perception of turboprops as an old-
fashioned, troublesome device with no passenger appeal," the task force report noted,
"the airlines and the manufacturers have little motivation to work on this engine
type."32 Clifton Von Kann succinctly summed up these concerns to Barry Goldwater
during his Senate testimony when he said that of all the proposed projects, "the
propeller is the real controversial one."33

What made the government willing to assume the risk for such a difficult project?
Proposed fuel savings was one important factor. However, the task force report
indicated another significant and related issue —the Soviet Union had a high speed
"turboprop which could fly from Moscow to Havana."34 The continuing Cold War
prompted the United States to view any Soviet technical breakthrough as a potential
threat to American security. During the energy crisis, the knowledge that Soviet
turboprop transports had already achieved high propeller fuel efficiency at speeds
approaching those of jet-powered planes seemed grave indeed and gave impetus to the
NASA program. During the government hearings, NASA representatives displayed
several photos of Russian turboprop planes to win congressional backing for the
project.35 The Cold War helped to define the turboprop debate. No extensive
speculation on the implications of Russian air superiority for American national
security seemed necessary. The Soviet Union could not be allowed to maintain
technical superiority in an area as vital as aircraft fuel efficiency. Thus, the report
included the demanding Advanced Turboprop Project as part of the ten-year, $670
million Aircraft Energy Efficiency Program to improve fuel efficiency.

Concept Development and Early Advocacy

Industry resistance and NASA Headquarters' sensitivity to the public relations aspect
of this opposition were among the key reasons that of the six projects within the
Aircraft Energy Efficiency (ACEE) program, only the advanced turboprop failed to
receive funding in 1976. John Klineberg, later director of Lewis Research Center,
recalled that it was delayed "because it was considered too high risk and too
revolutionary to be accepted by the airlines."36

31. Hager and Vrabel, Advanced Turboprop Project, p. 5.

32. Aircraft Fuel Conservation Technology, task force report, September 10, 1975, p.
44.

33. Clifton F. Von Kann, testimony before the U.S. Senate, Committee on
Aeronautical and Space Sciences, September 10, 1975, p. 9.

34. Mikkelson interview.

35. Aircraft Fuel Conservation Technology, task force report, September 10, 1975, p.
48. These Soviet long-range turboprops included the Tupolev TU-95 "Bear" (which
weighed 340,000 pounds, had a maximum range of 7,800 miles, a propeller diameter
of 18.4 feet, and operated at a .75 mach cruise speed) and the Antonov AN-22
"Cock" (which weighed 550,000 pounds, had a maximum range of 6,800 miles, a
propeller diameter of 20.3 feet, and operated at a .69 mach cruise speed).

36. John Klineberg, quoted in "How the ATP Project Originated," Lewis News, July 22,
1988.
 330 THE ADVANCED TURBOPROP PROJECT

If the advanced turboprop was so important to the national welfare, why did it
encounter such opposition from the airframe and aircraft engine manufacturers?
Donald Nored, the division chief in charge of the three propulsion projects at Lewis,
remarked that his engineering peers in industry were "very conservative and they had
to be." They were "against propellers" because they had "completely switched over to
jets." Because of their commitment to the turbojet, they continually cited problems
that they believed resulted from propellers. This included noise, maintenance, and the
fear that the "blades would come apart." Nored recalled each problem had to be "taken
up one at a time and dealt with."37 It appears the government's revolutionary vision of
the future frightened the aircraft industry with its large investment in turbofan
technology. Aircraft structures and engines are improved in slow, conservative,
incremental steps. To change the propulsion system of the nation's entire commercial
fleet represented an investment of mind-boggling proportions. Even if the government
put several hundred million dollars into developing an advanced turboprop, the
airframe and aircraft engine industries would still need to invest several billion dollars
to commercialize it. Revolutionary change did not come easily to an established
industry so vital to the nation's economy.

Turboprop advocates encountered not only the opposition of industry representatives,

but the hesitation and timidity of NASA Headquarters. By default, the advocacy role
fell to NASA Lewis engineers, though the public relations aspect of technology
funding had never been the Cleveland laboratory's strong suit. Lewis had a reputation
for being more conservative and technical than the other NASA Centers.38 One Lewis
engineer remarked that when other Centers sent five representatives to important
meetings, Lewis sent one. Moreover, research engineers from the aeronautics side of
NASA had little experience managing major contracts. Yet the energy crisis and the
need for projects to sustain the Center's viability within NASA galvanized a small
cadre of Lewis engineers into action. They used their technical and new-found
managerial creativity to sell NASA Headquarters and industry on a revolutionary new
propulsion system —one that might forever ground all existing subsonic turbojets.

Technically, the entire future of the advanced turboprop project initially depended on
proving whether a model propfan could achieve the predicted fuel efficiency rates.39 If
this model yielded successful results, then project advocates would be able to lobby
for increased funding for a large research and development program. Thus, even
during its earliest phase, the technical and social aspects of the project worked in
tandem.
Lewis project managers awarded a small group of researchers at Lewis and Hamilton
Standard a contract for the development of a two-foot diameter model propfan, called
the SR-1 or single-rotating propfan. Single-rotating meant that the propfan had only
one row of blades, as opposed to a counter-rotating design with two rows of blades,
each moving in opposite directions. This model achieved high efficiency rates and
provided technical data that the small group of engineers could use as ammunition in
the fight to continue the program.

At the same time that they proved the technology using small-scale models, Lewis
engineers built a consensus for the project, defending it against objections of skeptical
segments of industry and government advisory committees. Advocacy is essentially
"marketing" or "selling" to gain government funding and industry backing for new
programs like the advanced turboprop. Funding government programs is neither
scientific nor entirely rational, but depends on people and how they navigate a
complex bureaucracy,

37. Nored interview.

38. Ibid.

39. Aircraft Fuel Conservation Technology, task force report, September 10, 1975, p.
46.

FROM ENGINEERING SCIENCE TO BIG SCIENCE 331

while avoiding numerous political entanglements. During the Apollo years, NASA
had what amounted to a blank check to land a human being on the Moon within a
decade. Not needing to spend time and energy fighting for funding, engineers had
greater freedom to focus on building and testing hardware and managing space
missions. But to keep the programs of the 1970s alive, even those that responded to a
national crisis, required effort in non-technological spheres of activity.
Lewis was fortunate that Donald Nored, a maestro of project management, played a
strong role in building a constituency in support of the project. Unlike most of the
other Lewis engineers involved in advanced turboprop development, he hailed from
the space side of NASA's house. He had worked on chemical rockets and high power
lasers prior to taking up his post as head of the Aircraft Energy Efficiency Program
Office at Lewis in 1975. He helped to show aeronautical engineers, more at ease with
in-house research, how to negotiate the system to win funding. In 1981, with Frank
Berkopec, Nored attempted to demystify the advocacy process by laying down
guidelines for others within the Aeronautics Directorate. They disabused their order-
seeking engineering colleagues of the notion that advocacy could be compressed into
a series of well-defined steps. Rather, they wrote, it is

332 THE ADVANCED TURBOPROP PROJECT

 FROM ENGINEERING SCIENCE TO BIG SCIENCE 333

"basically informal, unstructured, and quite often confusing."40 Since only a few of
the proposed NASA programs received funding each year, they argued, the advocacy
process had become essential and activities related to it should receive a "high
priority."41

The advocacy guidelines indicated that the interactions with "industry, advisory
groups, and especially Headquarters will often require rapid, comprehensive, and in-
depth respondents [sic] to requests."42 One early request of the turboprop project
centered on the aircraft industry's concern over the safety of propellers. An aircraft
accident advisor raised a question during a meeting of the Industrial Advisory Board
at NASA Headquarters concerning the "safety aspect of propellers breaking away
from the engine and the damage caused by their impingement into the
fuselage."43 Lewis engineers quickly launched their own study into propeller safety
and commissioned similar studies at Hamilton Standard and Detroit Diesel Allison.
The results were overwhelmingly positive. Lewis examined over 12,000 accident
reports from 1973 to 1975 and found no instance where a propeller blade broke away
from its engine.44 Hamilton Standard reported that after fifty million hours of
propeller flight time there had never been an instance of structural failure.45 While
after twenty million hours, Detroit Diesel Allison found one structural failure; they
were quick to point out that "the aircraft landed routinely without further incident and
no one was injured in the aircraft or on the ground."46 This example typifies not only
the early skepticism and resistance by industry to the idea of returning to propeller
aircraft but also the "rapid, comprehensive, and in-depth responses" of NASA to
industry's concerns. The advocacy process required to "market" and "sell" the radical
turboprop project was in full swing. It continued to effectively diffuse the concerns of
skeptics.

Enabling Technology and Project Management

Successful advocacy brought the formal establishment of the Advanced Turboprop

Project in 1978 and initiation of the enabling technology phase. As the lead Center for
the project, NASA Lewis had full responsibility for the management of its
increasingly far-flung and complicated pieces. Before this phase began, NASA
engineers devised a detailed "management approach" and the plan was approved in
1977. Officially, Lewis was to have "responsibility to execute all detailed project
planning documentation, develop and implement the procurement of components and
systems, provide technical direction to contractors, perform contract administration,
perform engineering functions, coordinate the related in-house research and
technology programs, and exercise the usual project review reporting and control
functions."47 These interrelated activities put Lewis in the middle of an intricate web
of government (other NASA Centers), industry, and academic contracts. Project
managers were responsible for assigning the technology contracts. They also had the
equally important function of ensuring that both the public and the government
viewed the ATP positively.

40. Donald L. Nored and Frank D. Berkopec, "Guidelines for Advocacy of the New
Programs in the Aeronautics Directorate," unpublished report, January 1981, Nored
papers, NASA-Lewis Research Center, file Nored/Berkopec, box 238, p. 1
.
41. Ibid., p. 10.

42. Ibid.

43. J. E. Wikete, Aircraft Accident Information, August 4,1976, Nored papers, NASA,
box 224.

44. Paul J. McKenna (Lewis Research Center) to Wikete, July 12,1976, Noted papers,
NASA, box 224.

45. R.M. Levintan (Hamilton Standard) to Wikete, July 27, 1976, Nored papers, NASA,
box 224.

46. P.C. Stolp (Detroit Diesel Allison) to P. Christman, July 14, 1976, Nored papers,
NASA, box 224.

47. Project Plan for the Advanced Turboprop Program, September 1977, NASA,
Noted papers, box 229, p. 26.
334 THE ADVANCED TURBOPROP PROJECT
 FROM ENGINEERING SCIENCE TO BIG SCIENCE 335

Once the management structure was in place, the technology studies could begin.
Technically, this phase dealt with four critical problems: modification of propeller
aerodynamics, cabin and community noise, installation aerodynamics, and drive
systems.48 Propeller aerodynamic work included extensive investigations of blade
sweep, twist, and thickness. The late 1970s was the first time that engineers used a
high speed computer to analyze the design of a propeller. Computers were not yet in
widespread use when the turbofan replaced propeller-powered planes in the 1950s.
Lewis programmers used their Cray supercomputers to develop the first three-
dimensional propeller aerodynamic analysis, A further structural and aerodynamic
achievement was to use thinner titanium blades to reduce the flutter problems
associated with the steel propeller blades used in the 1940s and 1950s.

The advantage of propellers to save fuel had to be balanced against the potential harm
to the environment their noise caused.49 New computer-generated design codes not
only contributed to improved propeller efficiency, but contributed to solving problems
associated with noise. Engineers closely monitored the effect of propeller noise on
both cabin occupants and people on the ground. To study propeller acoustics, they
mounted propeller models on a JetStar aircraft fuselage at the NASA Dryden facility.
Microphones located on the airframe and also on a Learjet chase plane provided data
at close range and at a distance. After reviewing the sound pattern data, they
concluded that substantial

48. Donald L. Nored, John B. Whitlow, Jr., William C. Strack, "Status Update of the
NASA Advanced Turboprop Project," unpublished report, Noted papers, pp. 4-10.

49. Aircraft Fuel Conservation Technology, task force report, September 10, 1975, p.
18.

336 THE ADVANCED TURBOPROP PROJECT

noise reduction technology was necessary to meet the established goals. Eventually,
they achieved a reduction of sixty to sixty-five decibels of noise through a
combination of structural advances and flight path modifications.

The final two technical problems of the enabling phase dealt with installation
aerodynamics and the drive system. Numerous installation arrangements were
possible for mounting the turboprop on the wing. Should the propeller operate by
"pushing" or "pulling" the aircraft? How should the propeller, nacelle, and the wing be
most effectively integrated to reduce drag and increase fuel efficiency? Wind tunnel
tests were able to reduce drag significantly by determining the most advantageous
wing placement for the propeller. Engineers also examined various drive train
problems, including the gearboxes.
Solutions to all the enabling phase technical problems was still not enough to
guarantee the continued funding of the program. Key social questions were still
associated with this controversial technology. A vital concern for the advanced
turboprop project managers was the social question concerning passengers: how
receptive would they be to propeller-driven aircraft? In 1975, a government panel
reported that they were "generally opposed to the turboprop aircraft, primarily
because they felt that there would be little or no public acceptance." 50 If the public
would not fly in a turboprop plane, all the potential fuel savings would be lost flying
empty planes across the country.

In response to this concern, NASA and United Airlines initiated an in-flight

questionnaire to determine customer reaction to propellers. Both NASA and industry
were aware of the disastrous consequences for the future of the program if this study
found that the public was against the return of propeller planes. As a result, the
questionnaire de-emphasized the propeller as old technology and emphasized the
turboprop as the continuation and advancement of flight technology. The first page of
the survey consisted of a letter from the United Airlines vice president of marketing to
the passenger asking for cooperation in a "joint industry-government study concerning
the application of new technology to future aircraft."51 This opening letter did not
mention the new turboprops. The turboprop, inconspicuously renamed the "prop-fan"
to give it a more positive connotation, did not make its well-disguised appearance
until page four of the survey where the passenger is finally told that "'prop-fan' planes
could fly as high, as safely, and almost as fast and smooth as jet aircraft." This was a
conscious rhetorical shift from the term "propeller" to "prop-fan" to disassociate it in
peoples' minds from the old piston engine technology of the pre-jet propulsion era.
Brian Rowe, a General Electric vice president with oversight of the advanced
propeller projects, explained this new labeling strategy. He said, "They're not
propellers. They're fans. People felt that modern was fans, and old technology was
propellers. So now we've got this modern propeller which we want to call a
fan."52 The questionnaire explained to the passenger that not only did the "'prop-fans'
... look more like fan blades than propellers," they would also use twenty to thirty
percent less fuel than jet aircraft.

The questionnaire then displayed three sketches of planes-two were propeller driven
and the third was a turbofan. The passenger had to choose which one he or she would
"prefer to travel in." Despite all the planes being in-flight, the sketches depicted the
propellers as simple circles (no blades present), while the individual blades of the
turbofan were visible. These were all subtle and effective hints to the passenger that
the "prop-fan" was nothing new and that they were already flying in planes powered
by engines with fan blades.
50. George M. Low to Alan Lovelace, April 28, 1975.

51. United Airlines Passenger Survey, NASA, Nored papers, box 224.

52. Quoted by Martha Hamilton, "Firms Give Propellers a New Spin: GE leads high-
stakes competition for aircraft engineers with its 'fan,"' The Washington
Post, February 8, 1987, p. H4, column 1.

FROM ENGINEERING SCIENCE TO BIG SCIENCE 337

Not surprisingly, the survey yielded favorable results for the turboprop. Of 4,069
passengers surveyed, fifty percent said that they "would fly prop-fan," thirty-eight
percent had "no preference," and only twelve percent preferred a jet.53 If the airlines
could avoid fare increases due to the implementation of the turboprop, eighty-seven
percent of the respondents stated they would prefer to fly in the new turboprop.
Relieved and buoyed by the results, NASA engineers liked to point out that most of
the passengers did not even know what propulsion system was currently on the wing
of their aircraft.54 According to Mikkelson, all the passengers wanted to know was
"how much were the drinks, and how much was the ticket."55 Equally relieved was
Robert Collins, vice president of engineering for United Airlines, who concluded that
this "carefully constructedpassenger survey ... indicated that a prop-fan with
equivalent passenger comfort levels would not be negatively viewed, especially if it
were recognized for its efficiency in reducing fuel consumption and holding fares
down."56

At times project management also involved informing and changing government

opinion. Aeronautics programs within NASA, because of the low levels at which they
were traditionally funded, had never required close oversight by the General
Accounting Office (GAO). The large budget and greater visibility of the Aircraft
Energy Efficiency Program (ACEE) suddenly brought it unwanted attention. The first
draft of the General Accounting Office's 1979 review, though generally favorable
toward the ACEE program, was highly critical of the advanced turboprop project. It
concluded with the statement that the "GAO believes that much of the fuel savings
under ACEE attributed to the turboprop will not be realized. "57

The draft's "negative tone" and "misleading and distorted view of the program" deeply
concerned NASA Lewis project managers who feared the repercussions it would have
on funding decisions.58 They quickly went on the attack. Center Director John
Klineberg heatedly responded that the GAO had treated the turboprop project unfairly
in comparison with the other aircraft efficiency projects, calling the GAO ignorant of
the project's "inherent uncertainties."59

NASA Lewis project managers prevailed in the battle against the negativity of the
GAO draft report. The final publication specifically contained a retraction. The "GAO
carefully reevaluated its presentation and made appropriate adjustments where it
might be construed that the tone was unnecessarily negative or the data misleading."
An example of these "appropriate adjustments" is apparent in a comparison of how
one sentence changed from the draft to the final version. In the draft, the sentence
appeared as: "The Task Force Report shows that in 1975 there was considerable
disagreement on the ultimate likelihood of a turboprop engine being used on
commercial airliners."60 In the final publication, the GAO amended the same sentence
to: "The possible use of turboprop

53. Prop-Fan, survey results, December 1978, NASA, Nored papers, box 231.

54. Interview with Keith Sievers, August 17, 1995, and telephone interview with
Raymond Colladay, August 17 , 1995, by Virginia Dawson and Mark Bowles.

55. Mikkelson interview.

56 - Authors' italics. Robert C. Collins statement submitted to subcommittee on

transportation, aviation, and materials, House of Representatives Committee on
Science and Technology, February 26, 1981.

57. Preliminary draft of a proposed report, review of NASA's Aircraft Energy

Efficiency Project, GAO office, August 1979, Nored papers, box 182, p. 36.

58. Unknown NASA Headquarters administrator to J. H. Stolarow, January 24, 1980,

NASA, Nored papers, box 182, file GAO report.

59. John M. Klineberg to NASA Headquarters, December 21, 1979, NASA-Lewis

Research Center, Nored papers, box 182, file GAO report.
60. Preliminary draft of a proposed report, review of NASA's Aircraft Energy
Efficiency Project, GAO office, August 1979, Nored papers, box 182, p. 37.

338 THE ADVANCED TURBOPROP PROJECT

engines on 1995 commercial aircraft is still uncertain, but has gained support since
1975."61 These editorial changes giving the report a positive spin indicate the
effectiveness of project managers in changing public opinion. Everyone, it seemed,
had begun to associate the advanced turboprop technology with the possibility of
bringing about an aeronautical "revolution," a paradigm shift, or as Forbes magazine
headlined in 1984, "The Next Step." As surely as 'jets drove propellers from the
skies," the new "radical designs" could bring a new propeller age to the world.62

It is important to underscore how important the interpersonal skills of the project

managers were to continuation of the program throughout this enabling technology
phase. They were responsible not only for managing the project's technology, but also
for enabling, proving, maintaining, and adjusting support for the turboprop. They
continued to push this controversial technology against the conservative interests of
the government, industry, and the public. Their consistent success paved the way for
the third stage.

Large-Scale Integration and Competition

After two years of work, the advanced turboprop idea began to attract greater
commercial interest. As a result of NASA's advocacy efforts, news articles began to
predict the coming propeller "revolution." All indicators pointed to the introduction of
the new turboprops on commercial aircraft by the 1990s. With the small-scale model
testing complete, a data base, and an acceptable design methodology established, the
project moved into its most labor and cost intensive phase —that of large-scale
integration. The project still had serious uncertainties and problems associated with
transferring the designs from a small-scale model to a large-scale prop-fan. Could
engineers maintain propulsion efficiency, low noise levels, and structural integrity
with an increase in size? The Large-Scale Advanced Prop-fan (LAP) project initiated
in 1980 would answer these scalability questions and provide a database for the
development and production of full-size turbofans.

As a first step, NASA had to establish the structural integrity of the advanced
turboprop.63 Project managers initially believed that in the development hierarchy
performance came first, then noise, and finally structure. As the project advanced, it
became clear that structural integrity was the key technical problem.64 Without the
correct blade structure, performance could never achieve predicted fuel savings.
NASA awarded Hamilton Standard the contract for the structural blade studies that
were so crucial to the success of the whole program. In 1981, they began to design a
large-scale, single-rotating prop-fan made of composite material. Five years later they
completed construction on a 9-foot-diameter design very close to the size of a
commercial model. The model was so large that no wind tunnel in the United States
could accommodate it. The turboprop managers decided to risk the possibility that the
European aviation community might benefit from the technology that NASA had so
arduously perfected. They shipped the large-scale propeller, called the SR-7L, to a
wind tunnel in Modane, France, for testing. In early 1986, researchers subjected the
model to speeds up to Mach 0.8 with simulated altitudes of 12,000 feet. The results
confirmed the data obtained from the small model propeller designs. The large-scale
model was a success.

61. "A look at NASA's Aircraft Energy Efficiency Program," by the Comptroller
General of the United States, July 28, 1980, Nored papers, box 182, file GAO report,
p. 45.

62. Howard Banks, "The Next Step," Forbes, May 7, 1984, p. 31.

63. "Large-Scale Advanced Prop-Fan Program (LAP)," technical proposal by Lewis

Research Center, January 11, 1982, NASA, Noted papers, box 229.

64. Noted interview.

FROM ENGINEERING SCIENCE TO BIG SCIENCE 339

Success spawns imitators. While NASA continued to work with Allison, Pratt &
Whitney —and Hamilton Standard to develop its advanced turboprop, General
Electric (GE)—Pratt & Whitney's main competitor— was quietly developing an
alternative propeller system. A feature of radical inventions is that competitors often
introduce alternative forms of a similar technology before one form can prevail over
another. Historians of technology have shown many cases of "interpretative
flexibility" when "two or even more social groups with clearly developed
technological frames [artifacts] are striving for dominance in the field."65 This
happened when General Electric introduced its own radical alternative to NASA's
advanced turboprop project-the Unducted Fan (UDF). GE sprang the unducted fan on
NASA completely by surprise.
In NASA's design, the propeller rotated in one direction. This was called a single
rotation tractor system and included a relatively complicated gearbox. Since one of
the criticisms against the turboprop planes of the 1950s (the Electra, for example),
was that their gearboxes required heavy maintenance, GE took a different approach to
prop-fan design. Beginning in 1982, GE engineers spent five years developing a
gearless, counter-rotating, pusher system. They mounted two propellers (or fans) on
the rear of the plane that literally pushed it in flight, as opposed to the "pulling" of
conventional propellers. In 1983, the aircraft engine division of General Electric
released the unducted fan design to NASA shortly before flight tests of the NASA
industry design were scheduled. Suddenly there were two turboprop projects
competing for the same funds. Nored recalled: "They wanted us to

65. Wiebe E. Bijker, "The Social Construction of Bakelite: Toward a Theory of

Invention," The Social Construction of Technological Systems, 182.

340 THE ADVANCED TURBOPROP PROJECT

drop everything and give them all our money and we couldn't do that."66 NASA
Headquarters endorsed the "novel" unducted fan proposal and told NASA Lewis to
cooperate with General Electric on the unducted fan development and testing.

Despite NASA's initial reluctance to support two projects, the unducted fan proved
highly successful. In 1985, ground tests demonstrated a fuel conservation rate of
twenty percent.67 Development of the unducted fan leapt ahead of NASA's original
geared design. One year later, on August 20, 1986, GE installed its unducted fan on
the right wing of a Boeing 727. Thus, to many NASA engineers' dismay, the first
flight of an advanced turboprop system demonstrated the technical feasibility of the
unducted fan system —a proprietary engine belonging entirely to General Electric,
rather than the product of the joint NASA/industry team. Nevertheless, the
competition between the two systems, and the willingness of private industry to invest
its own development funds, helped build even greater momentum for acceptance of
the turboprop concept.

NASA engineers continued to perfect their single-rotating turboprop system through

preliminary stationary flight testing.68 The first step was to take the Hamilton
Standard SR-7A propfan and combine it with the Allison turboshaft engine and
gearbox housed within a special tilt nacelle. NASA engineers conducted a static or
stationary test at Rohr's Brown Field at Chula Vista, California, where they mounted
the nacelle, gearbox, engine, and propeller on a small tower. The stationary test met
all performance objectives after fifty hours of testing in May and June 1986. This
success cleared the way for an actual flight test of the turboprop system. In July 1986
engineers dismantled the static assembly and shipped the parts to Savannah, Georgia,
for reassembly on a modified Gulfstream II with an eight-blade, single-rotation,
turboprop on its left wing.69 The radical dreams of the NASA engineers for fuel
efficient propellers were finally close to becoming reality. The plane contained over
600 sensors to monitor everything from acoustics to vibration. Flight testing-the final
stage of advanced turboprop development-took place in 1987 when a modified
Gulfstream II took flight in the Georgia skies. These flight tests proved the predictions
of a twenty to thirty percent fuel savings (made by NASA in the early 1970s) were
indeed correct.

On the heels of the successful tests, of both the GE and the NASA-industry team
designs, came not only increasing support for propeller systems themselves, but also
high visibility from media reports forecasting the next propulsion revolution. The New
York Times predicted the "Return of the propellers" while a Washington
Times headline read, "Turboprops are back!"70 Further testing indicated that this
propulsion technology was ready for commercial development. As late as 1989, the
U.S. aviation industry was "considering the development of several new engines and
aircraft that may incorporate advanced turboprop propulsion systems."71 But the
economic realities of 1987 were far different from those predicted in the early 1970s.
Though all the technology and social problems standing in the way of
commercialization were resolved, the advanced turboprop never reached production, a
casualty of the one contingency that NASA engineers never anticipated —that fuel
prices would go down. (See figure 5) Once the energy crisis passed, the need for the
advanced turboprop vanished.

66. Noted interview.

67. James, J. Haggerty, "Propfan Update," p. 11.

68. Hagar and Vrabel, Advanced Turboprop Project, pp. 49-74. This stage was called
the Propfan Test Assessment (PTA) Project.
69. Mary Sandy and Linda S. Ellis, "NASA Final Propfan Program Flight Tests
Conducted," NASA News May 1, 1989.

70. Andrew Pollack, "The Return of Propellers," The New York Times, October 10,
1985, D2. Hugh Vickery, "Turboprops are back!," The Washington Times, November
1, 1984, p. 5B.

71. Mary Sandy and Linda S. Ellis, "NASA Final Propfan Program Flight Tests
Conducted," NASA News May 1, 1989.

FROM ENGINEERING SCIENCE TO BIG SCIENCE 341

Environmental Contingency and Insufficient Momentum

One of the main difficulties in the development of a radical new technology is the
potential project threatening problems that arise. If they are left unsolved they can
destroy an entire project. Historian of technology Thomas Hughes called these
problems "reverse salients." Hughes argues that all large technological systems (of
which the turboprop is an example) include political, economic, social, and
technological components.72 These system components are interrelated so that if one
of the components is changed or altered in any way, the rest of the system will also be
affected. The systems themselves grow and gain momentum by the process of
removing "reverse salients," which arise and could potentially cause the system to fail.
An example will help clarify the importance of solving these critical problems. In
1878, Thomas Edison encountered a technological reverse salient in his attempt to
develop his electric-lighting system. This problem was the short-lived filament of the
incandescent bulb. Edison realized that even if he solved this problem, a further
economic reverse salient remained. The expense of the copper wire needed to link the
entire system together was cost prohibitive for potential wide-scale acceptance. If
Edison could not reduce the amount of copper needed for his electric system, then
gas-lighting systems would become the more attractive alternative to the problem of
street lighting. What is important to understand is that either the technological or the
economic reverse salient could have caused the Edison system of electric-lighting to
fail.73

Like Edison, the managers of the turboprop project also confronted a variety of
critical problems. These problems included economic (the necessity of maintaining a
favorable ratio of cost to implement turboprop technology versus savings in fuel
efficiency), political (how to receive funding for a long-term project), social (how to
implement a technology which the public could perceive as a "step backward"),
institutional (how to successfully manage the government, industry, and academic
relations), and technical (how to actually build a turboprop that improved fuel
efficiency by twenty to thirty percent). Each of these problems had the potential to
sabotage the entire system. NASA engineers had their own, more practical and direct
term for "reverse salient" —a "showstopper." In 1984, engineers listed a number of
technical show-stoppers that threatened to derail the project if left unsolved—for
example, unacceptable levels of cabin noise.74

As system-builders solve critical problems, the system itself generates momentum.

This momentum continues to increase and build until, according to Hughes, either a
conversion, a catastrophe, or a contingency occurs. Conversions and catastrophes
break momentum through either a change in societal belief, like a religious
conversion, or a massive technological failure, like a nuclear-reactor catastrophe. But,
it is the role of contingency which interests us here as the key factor in the current
neglect of the advanced turboprop technology. Hughes identified one particular
"contingent environmental change" that altered the course of the entire automobile
industry —the energy crisis. He argues, "The oil embargo of 1973 and the subsequent
rise in gasoline prices ultimately compelled U.S. automobile manufacturers to change
substantially an automobile design that had been singularly appropriate to a low-cost-
energy environment."75

72. Thomas P. Hughes, Networks of Power Electrification in Western Society, 1880-

1930 (Baltimore, MD: The Johns Hopkins University Press, 1983).

73. Hughes, American Genesis: A Century of Invention and Technological

Enthusiasm (London, England: Penguin Books, 1989), pp. 71-74.

74. Chart entitled, "Potential show-stoppers," February 6, 1984, NASA-Lewis

Research Center, Noted papers, box 239, file ATP memos.

75. Ibid., p. 462.

342 THE ADVANCED TURBOPROP PROJECT

The development and subsequent neglect of advanced turboprop technology is the
result of this same environmental contingency. In the early 1970s, the energy crisis
created a situation which made it a national necessity for the government to explore
new ways to conserve fuel. What the managers of the Advanced Turboprop Project
(ATP) did not anticipate and could not control was a decrease in the cost of fuel. As
the energy crisis subsided in the 1980s and the fuel prices decreased, there was no
longer a favorable ratio of cost to implement turboprop technology versus savings in
fuel efficiency. As John R. Facey, advanced turboprop program manager at NASA
Headquarters, wrote, "An all new aircraft with advanced avionics, structures, and
aerodynamics along with high-speed turboprops would be much more expensive than
current turbofan-powered aircraft, and fuel savings would not be enough to offset the
higher initial cost."76 In the case of the ATP, its managers overcame all of their critical
problems. However, when contingent economic conditions changed so that fuel cost
was no longer a critical problem, regardless of the technical success of the project, the
advanced turboprop lost its potential market in the industrial world.

Yet Keith Sievers, at that time the manager of the ATP, along with a handful of
project staff, was convinced that the NASA industry team had made a significant
contribution to aviation that ought to receive recognition. To win the Collier Trophy,
he again summoned up the advocacy skills that had proved so valuable in bringing the
controversial advanced turboprop to the point of technical feasibility. He used them to
lobby for the prestigious Collier Trophy among the wide aeronautical constituency
that had participated in advanced turboprop development. NASA Headquarters
initially expressed some reluctance to lobby for awarding a prize for technology that
was unlikely to be used —at least in the near future. But the timing was perfect. There
was little competition from NASA's space endeavors since staff in the space
directorate were still in the midst of recovering from the tragic Challenger explosion.
As a result, the National Aeronautic Association awarded NASA Lewis and the
NASA Industry Advanced Turboprop Team the Collier Trophy at ceremonies in
Washington, DC. Today, the technology remains "on the shelf," or "archived,"
awaiting the time when fuel conservation again becomes a necessity.77

Despite the current neglect of the advanced turboprop, this case study demonstrates
how radical innovation can emerge from within a conservative, bureaucratic
government agency. The government—not industry—assumed the risk for developing
the new technology. It used taxpayers' money to advance a radical idea to the point of
technical feasibility. Engineers involved in the project used advocacy to build a
consensus among the members of the aeronautical community that the advanced
turboprop would prove a viable alternative to the far less energy efficient turbofan
technology. Indeed, the technical and social achievements of the project were
convincing enough to drive General Electric to invest its own funds to develop a
competing design. This competition was evidence of wide acceptance for the
turboprop concept.

The Collier Trophy in 1987 was presented to the "Lewis and the NASA Industry
Advanced Turboprop Team.'' The team, defined in its widest possible context,
included General Electric's independent contribution of the UDF and its subsequent
flight testing by NASA. In contrast to previous Collier trophies in aeronautics won by
the NACA, no individual received special mention. Certainly, throughout the eleven
years of its existence the project had encouraged inventiveness of individuals in a
variety of disciplines, from highly theoretical contributions in blade design and
acoustics to more routine testing.

76. John R. Facey, "Return of the Turboprops," Aerospace America (October 1988):
15.

77. Some specific technologies generated by the ATP project are in use today. These
include noise reduction advances, gearboxes that use the ATP design, and certain
structural advancements, for example, how to keep the blades stable, used in engine
designs today, including large, high by-pass ratio turbofans of the 1990s. Sievers
interview.

FROM ENGINEERING SCIENCE TO BIG SCIENCE 343

Participants in the project ran the gamut from government, university, and industry
researchers. But what the prize recognized above all was the project's management
genius. NASA Lewis managers did not simply manage contracts. They kept the
project alive. They used advocacy to win industry participation and cooperation, as
well as stimulate competition. They pushed both the technical and the social aspects
of the project to create the system's momentum. Yet once the energy crisis passed, this
momentum was insufficient to dislodge the massive technological momentum of the
existing turbofan system.

NASA engineers involved in the ATP project still remain confident that the future
economic conditions will make the turboprop attractive again. When fuel becomes
scarce and fuel prices begin to rise, the turboprop's designs will be "on the shelf' ready
to respond with tremendous fuel-efficient savings. But, technological neglect is not
the enthusiastic success on which NASA engineers built their careers. Donald Nored
wistfully reflected on the project and said, "We almost made it. Almost made it."78

78. Noted interview.

This article was downloaded by: [University of Calgary]
On: 02 October 2013, At: 12:51
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Technology Analysis & Strategic

Management
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/ctas20

Determining reverse salient types and

evolutionary dynamics of technology
systems with performance disparities
a a
Ozgur Dedehayir & Saku J. Mäkinen
a
Department of Industrial Management, Tampere University of
Technology, Tampere, Finland
Published online: 01 Nov 2011.

To cite this article: Ozgur Dedehayir & Saku J. Mäkinen (2011) Determining reverse salient types
and evolutionary dynamics of technology systems with performance disparities, Technology Analysis
& Strategic Management, 23:10, 1095-1114, DOI: 10.1080/09537325.2011.621308

To link to this article: http://dx.doi.org/10.1080/09537325.2011.621308

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
 Technology Analysis & Strategic Management
Vol. 23, No. 10, November 2011, 1095–1114

Determining reverse salient types and

evolutionary dynamics of technology systems
with performance disparities

Ozgur Dedehayir∗ and Saku J. Mäkinen

Downloaded by [University of Calgary] at 12:51 02 October 2013

Department of Industrial Management, Tampere University of Technology, Tampere, Finland

Technological system evolution is marked by the uneven evolution of constituent sub-systems.

Subsequently, system evolution is hampered by the resulting state of unevenness, or reverse
salience, which results from the presence of the sub-system that delivers the lowest level
of performance with respect to other sub-systems, namely, the reverse salient. In this paper,
we develop absolute and proportional performance gap measures of reverse salience and, in
turn, derive a typology of reverse salients that distinguishes alternative dynamics of change in
the evolving system. We subsequently demonstrate the applicability of the measures and the
typology through an illustrative empirical study of the PC (personal computer) technological
system that functions as a gaming platform. Our empirical analysis demonstrates that patterns
of temporal dynamics can be distinguished with the measurement of reverse salience, and that
distinct paths of technological system evolution can be identified as different types of reverse
salients emerge over time.

Keywords: technological system; system dynamics; reverse salient; computer games

Introduction
The evolution of a given technological system is hindered by performance imbalances that exist
among the sub-systems which comprise that system (Hughes 1983; Rosenberg 1976; Sahal 1981;
Constant 1980; den Hond 1998). In this state of affairs, the level of technological delivered by
the technological system is compromised by the sub-system that delivers the lowest level of
performance among all sub-systems, namely, the reverse salient (Hughes 1983, 1987). Because
the performance of the system can be further enhanced by improving the reverse salient through
greater utilisation of its performance potential, the reverse salient emerges as a focusing device
for technological development (Rosenberg 1969). The dynamics of system evolution is, in turn,
reflected in the evolution of the disparity that exists between the performance of the reverse salient
and its performance potential.
This performance disparity, in other words, the magnitude of reverse salience, emerges as an
informative parameter in understanding the evolution of technological systems, because it signifies

∗ Corresponding author. Email: [email protected]

ISSN 0953-7325 print/ISSN 1465-3990 online

© 2011 Taylor & Francis
http://dx.doi.org/10.1080/09537325.2011.621308
http://www.tandfonline.com
 1096 O. Dedehayir and S.J. Mäkinen

the system’s limited level of performance. The literature on technological system evolution is nev-
ertheless limited with respect to analytical tools that measure the state of reverse salience (Hugh,
Roche, and Bennett 2007). In one of the few studies considering measurement issues, Dedehayir
and Mäkinen (2008) propose an absolute performance gap measure of reverse salience magnitude.
However, such a measure considers only the disparity between sub-systems and remains inade-
quate in taking into account their heterogeneous levels of technological development. In addition
to the absolute measure, this paper therefore proposes a proportional measure of reverse salience,
which determines the ratio of the absolute performance gap with respect to the highest level
of performance attainable by the technological system. Together, the absolute and proportional
measures inform of the temporal change in the performance differential between sub-systems,
and at the same time denote the dynamics of change of the system as a whole. Furthermore, by
analysing the changes in these measures it becomes possible to derive different types of reverse
salients, which denote particular dynamics in technological system evolution.
Downloaded by [University of Calgary] at 12:51 02 October 2013

The structure of the paper is as follows. We first discuss technological systems and their general
properties, and reverse salients as inherent elements of these systems. Further, we develop absolute
and proportional performance gap measures of reverse salience, and then derive a typology of
reverse salients with respect to these measures. The empirical illustration, in turn, applies the
measures of reverse salience and the typology of reverse salients in the analysis of the PC (personal
computer) technological system. The empirical study considers two systemic contexts: the first
studying the evolution of the PC game and CPU (central processing unit) sub-systems, and the
second studying the PC game and GPU (graphics processing unit) sub-systems. Finally, we
conclude the paper with a discussion offering both scholarly and managerial implications along
with thoughts on avenues for future research.

Theoretical background
Technological systems
A number of different studies have adopted a systemic view of technology to reveal the complex-
ities that underlie the development of these technologies. For example, studies of the aeroplane
(Vincenti 1994; Constant 1987; Tushman and Murmann 1998), electricity supply (Hughes 1983),
the automobile (Abernathy and Clark 1985; Clark 1985) and the PC (Christensen 1997; Wade
1995; Christensen, Suarez, and Utterback 1998; Esposito and Mastroianni 1998), all illustrate the
technological system as a structured composition of technological parts or sub-systems. More-
over, all of these technological systems exhibit a number of properties that stem from general
systems theory (Skyttner 2001; Simon 1962; von Bertalanffy 1969).
First, the technological system has a hierarchically nested structure, whereby the system is seen
as a composition of smaller sub-systems (i.e. the system level technology comprises lower level,
sub-system technologies) that are themselves systems comprising further sub-systems, and so on
(Tushman and Murmann 1998; Murmann and Frenken 2006). For instance, the aeroplane techno-
logical system has some major sub-systems including the engine, which is composed of smaller
sub-systems, such as the compressor, combustion system and turbine, which are subsequently
divisible yet again into smaller sub-systems (e.g. Constant 1987).
Second, the sub-systems that are specialised for particular functions are interdependent both
within the same, as well as across, different levels of the system hierarchy (e.g. Tushman and
Murmann 1998). In this manner, the PC system’s HDD (hard disc drive), CPU and software
sub-systems, for example, are interconnected and influence one another (e.g. Christensen 1997).
 Determining reverse salient types and evolutionary dynamics of technology systems 1097

Third, the system and its properties are synthesised through the sub-systems that constitute it
(e.g. Tushman and Murmann 1998). Subsequently, the electricity supply system may be analysed
through the study of the capacitors, transmission lines and electric generator sub-systems, the
performances of which are synthesised at the system level to enable the distribution of electricity
(Hughes 1983).
Finally, the technological system as a whole is goal-seeking (Hughes 1983, 1987; Sahal 1981).
In this light, the electric supply system has been shown to develop over time to satisfy the objective
of increasing the area of electricity distribution (e.g. Hughes 1983), the PC system to increase
computational power (e.g. Christensen 1997) and the aeroplane to increase flight velocity (e.g.
Sahal 1981).
Since the technological system is goal-seeking, it evolves to improve its existing performance
output over time. The evolution of the system is nevertheless dependent on the development
of all the nested sub-systems (Hughes 1983; Rosenberg 1976; Tushman and Murmann 1998;
Downloaded by [University of Calgary] at 12:51 02 October 2013

Murmann and Frenken 2006). In this collective evolution, differences in the rates of development
of these interdependent sub-systems, however, lead to the appearance of technological imbalances
(Rosenberg 1976; Krell 1992). In the resulting state of technological disequilibrium, the system
level performance is significantly limited because of the presence of a sub-system that delivers
the lowest level of performance with respect to all other sub-systems, referred to as the reverse
salient (Hughes 1983).

Reverse salients
According to Hughes (1983), in its military origin, the reverse salient pinpoints that section of
a moving battle line which trails behind all other sections with respect to the overall direction
of movement, and subsequently hinders collective forward movement. In the context of evolving
technological systems, analogous to the military depiction, the reverse salient denotes a sub-
system that delivers the lowest level of performance with respect to all other sub-systems, and at
the same time hinders the system level performance delivery. This circumstance arises, first, from
the interdependence between sub-systems, whereby the inability of the reverse salient to utilise
the performance potential provisioned by other sub-systems creates a technological imbalance. In
turn, the synthesis of the sub-systems’ performances, in other words the system level performance,
is inhibited by the imbalance brought about by the reverse salient.
A number of studies in the literature illustrate reverse salients which have impacted the sys-
tem level performance in this manner. These include, for example, the motor and capacitor
sub-systems of the early direct-current electric system that inhibited the efficient distribution
of electricity beyond a certain range (Hughes 1983), the gyroscope sub-system of the ballistic
missile technological system that limited the accuracy of missiles (MacKenzie 1987), and the
method of transferring digitally created text and image data between different processes that con-
strained the development of the computer-integrated manufacturing (CIM) system (Hardstone
2004). However, examples, such as the music copyright managing institutions of the mobile
music technological system, which curbed the proliferation of mobile music throughout the end-
user markets in Japan and Korea (Takeishi and Lee 2005) also show that reverse salients may be
more than technical sub-systems. The heterogeneity of these examples, therefore, underlines the
diversity of reverse salients, which can be identified in a variety of technological systems.
Furthermore, these illustrations highlight not only the impact of reverse salients on the sys-
tem level performance, but also the incapacity of the reverse salient to utilise the performance
potential provisioned by interdependent sub-systems. More specifically, the reverse salient strays
 1098 O. Dedehayir and S.J. Mäkinen

behind the sub-system that provisions the highest level of technological performance, namely the
technological performance frontier (Dosi 1982), failing to fully utilise the performance potential
that it provides. As a result, the presence of a reverse salient gives rise to the unevenness formed
by this sub-system’s backward protrusion. As illustrated by Dedehayir and Mäkinen (2008), the
magnitude of unevenness, or reverse salience, can be measured as the technological performance
disparity between the technological performance frontier and the reverse salient.
This absolute performance gap measure, together with the level of performance delivered by the
reverse salient, informs of the level of performance that is delivered, as well as the performance
potential that remains unutilised by the technological system. While this absolute measure can
denote the system level dynamics, it nevertheless ignores the levels of performance attained by
the sub-systems. In this paper, we propose that reverse salience can additionally be measured
as the proportion of the absolute performance gap with respect to the performance delivered
by the technological performance frontier. This proportional measure, importantly, allows the
Downloaded by [University of Calgary] at 12:51 02 October 2013

comparison of system development at different points in time, taking into account the level of
technological evolution across sub-systems. As a consequence, the absolute and proportional
performance gap measures, taken together, provide more comprehensive information about the
system level development.

A typology of reverse salients

Because the performance of the technological system can be further enhanced by improving the
reverse salient through greater utilisation of its performance potential, the reverse salient emerges
as a focusing device for technological development (Rosenberg 1969). Hence, the system’s tech-
nological evolution is driven by a compulsive necessity to reduce the technological imbalance
(Rosenberg 1976; Sahal 1985; Nelson and Winter 1977), where the most fruitful reduction is
brought about by the improvement of the reverse salient.
The dynamics of change of the technological system can therefore be observed through the
temporal change in reverse salience. As a consequence, system dynamics is reliant on the rates
of development of both the reverse salient and the technological performance frontier. In this
manner, the technological performance frontier can undergo performance increase, experience no
performance change at all, or in certain peculiar circumstances, undergo performance decrease,
caused, for example, by excessive system complexity or shock from the institutional environment.
At the same time, the reverse salient experiences a rate of performance development that is
faster, the same, or slower than the rate of change of the technological performance frontier.
Consequently, the magnitude of reverse salience, when measured as absolute performance gap,
can increase or decrease in relation to the technological performance frontier. At the same time,
the magnitude of proportional performance gap can also increase or decrease in relation to the
technological performance frontier. We can, as a result, identify changes in reverse salience
according to the absolute and proportional measures, as illustrated in Figure 1.
From the changes in the two measures of reverse salience, we can derive four alternative types of
reverse salients, as depicted in Figure 1. Of these, the prohibitive reverse salient, first, experiences
a substantially lower rate of technological development in comparison with the development of
the technological performance frontier. It therefore brings about an increase in reverse salience,
measured both in absolute and proportional terms. A prohibitive reverse salient can, for example,
denote a constraint that remains unrecognised as a hindrance and therefore lies dormant and unno-
ticed for a period of time (Moors and Vergragt 2002; Moors 2006). Moreover, the continuation of
 Determining reverse salient types and evolutionary dynamics of technology systems 1099
Downloaded by [University of Calgary] at 12:51 02 October 2013

Figure 1. A typology of reverse salients.

its limited development can result in significant limitations to the technological system’s progress
toward existing system level objectives.
Second, the intermediating reverse salient is characterised by a rate of development that is
slower than the rate of change at the technological performance frontier, thus resulting in absolute
performance gap increase. Nevertheless, the reverse salient evolves at a sufficient pace to reduce the
proportional performance gap, even though the absolute gap persists over time. The simultaneous
occurrence of absolute performance gap increase and proportional performance gap decrease,
however, represents a narrow range of evolutionary possibilities, rendering this state volatile.
At the same time, the intermediating reverse salient denotes the system dynamics in between
relatively more stable states, such as when there is decrease or increase in both the absolute and
proportional measures. As a consequence, this reverse salient type may be observed when some
instability is induced into the system, for example, through exogenous factors such as regulatory
restrictions or initiatives concerning technological development (Pilkington 1998). Nevertheless,
the intermediating reverse salient type promotes system development within the existing regime.
The progressive reverse salient, in turn, demonstrates a higher rate of technological develop-
ment in comparison to the development of the technological performance frontier, thereby leading
to a decrease in both the absolute and proportional performance gaps. The reverse salient’s pace of
technological change may be sponsored by a sub-system industry in which organisations competi-
tively utilise the available level of technology in developing products that offer higher performance
(Zahra and Covin 1993; Das and van de Ven 2000). Product competition may subsequently push
the delivered technological performance level of the reverse salient even higher, also in rela-
tion to the technological performance frontier. As a result, the progressive change of this reverse
salient leads to closure of the technological performance disparity and therefore enhanced level
of utilisation of the reverse salient’s performance potential.
Finally, the reorienting reverse salient is characterised by a rate of development that is faster than
the rate of change at the technological performance frontier, thus resulting in absolute performance
gap decrease. Nevertheless, the reverse salient evolves at a pace that is slow enough to increase the
proportional performance gap concurrently. This temporal change in reverse salience may arise
from discontinuities in the technological performance frontier, sourced, for example, from disrup-
tive changes (Christensen 1997) within the system or shocks from the institutional environment
 1100 O. Dedehayir and S.J. Mäkinen

(Anderson and Tushman 1990). This may also represent new performance parameters rendering
the current parameter obsolete or less meaningful. In comparison with the intermediating reverse
salient type, the emergence of a reorienting reverse salient represents a fundamental change within
the system, which may cause the reorientation of system level objectives and therefore a shift in
the critical parameters for technological development.

Method
Our method of system analysis measures the magnitude of reverse salience between central sub-
systems which contribute to the system level performance. The evolution of the technological
system is in turn studied through the changing magnitude of reverse salience between sub-systems.
We therefore begin our method of analysis by developing the analytical measures of reverse
salience. In this effort, we first consider the evolution of any technological sub-system, that, when
Downloaded by [University of Calgary] at 12:51 02 October 2013

measured quantitatively by the changes in its technological performance indicator over time
(Sahal 1981), produces an S-curve (e.g. Christensen 1992; Foster 1986; Fisher and Pry 1971;
Andersen 1999). Second, we derive the measure of reverse salience magnitude by superimposing
the technological evolution curves of sub-systems upon a common set of axes, thus creating the
basis for two – absolute and proportional – measures of reverse salience (see Figure 2).
The dashed line in Figure 2 represents the actual level of technological performance delivered
by the reverse salient sub-system over time. The solid line, in comparison, denotes the evolution
of the technological performance frontier. Thus, the solid line represents the potential level of
technological performance provisioned to the reverse salient across this timeframe. The gap
between the two technological development lines represents the changing magnitude of reverse
salience over time. First, at any point in time, the magnitude of reverse salience can be measured
as absolute performance gap by calculating the vertical separation of the lines at that point in
time, denoted in Figure 2 by pt1 and pt2 , at time t1 and t2 , respectively.
The absolute performance gap measure indicates how far the reverse salient is behind the
technological performance frontier, although it does not reflect the reverse salient’s impact on
system level performance. In particular, the absolute measure does not take into account the level
of technological evolution across different sub-systems, at different points in time. On the contrary,

Figure 2. Illustrative S-curves of the technological performance frontier and reverse salient, and reverse
salience measured by absolute and proportional performance gap.
 Determining reverse salient types and evolutionary dynamics of technology systems 1101

the technological system performance is better reflected in the proportion of the technological
frontier performance that is yet to be utilised by the reverse salient. Therefore, the magnitude
of reverse salience is, second, measured as proportional performance gap, which is evaluated as
the ratio of the absolute performance gap with respect to the performance at the technological
frontier at a particular point in time. The proportional measures of reverse salience are indicated
in Figure 2 as pt1 /p3 at time t1 , and pt2 /p4 at time t2 , respectively.
It is, in turn, possible to measure the changes in reverse salience magnitude as absolute per-
formance gap (pt2 – pt1 in Figure 2) and proportional performance gap (pt2 /p4 – pt1 /p3
in Figure 2), over a particular period of time (t2 – t1 in Figure 2). Moreover, the analysis of the
temporal change in reverse salience allows the evaluation of an increase or decrease in the absolute
and proportional performance gap measures. Subsequently, by comparing the changes (whether
increase or decrease) in the absolute and proportional performance gaps within the given period
of time, it is possible to derive the type of reverse salient (see Figure 1) that persists across this
Downloaded by [University of Calgary] at 12:51 02 October 2013

particular time span. Further, the shifts from one reverse salient type to another can inform us of
the developments at the system level and its overall dynamics.

The technological system

Our empirical work illustrates the applicability of the absolute and proportional measures of
reverse salience in the analysis of an evolving technological system, and the identification of
reverse salient types from the observed changes in reverse salience. In our empirical illustration
we have analysed the PC technological system, focusing specifically on its function as a computer
gaming platform. We have selected this empirical setting, first, on the basis of the PC system’s
continuous endeavour to deliver higher levels of computer gaming performance. Second, while
computer gaming on the PC technological system has evolved over the years, disparities have
been observed in the performance levels of its sub-systems, namely, the PC hardware and PC
game software technologies. Third, the hardware and software sub-systems are interdependent,
as illustrated by the minimum hardware performance required by the PC game software in order
to deliver gaming performance on the PC technological system, whereby higher minimum hard-
ware requirements denote higher levels of overall system performance for gaming purposes. The
significance of the PC as a platform for computer gaming since the 1990s (Hayes and Dinsey
1995; Poole 2004), together with the highly dynamic nature of the sub-systems’ technological
evolutions, additionally provides a rich setting with continuously changing performance dispari-
ties. The combination of these factors, therefore, makes the PC technological system informative
for illustrating our measures of reverse salience and typology of reverse salients.
In our study, the PC represents the system level of a hierarchically structured technological
system, which evolves over time to deliver better gaming performance to the end-user. By gaming
performance, we refer, for example, to the video qualities (e.g. polygon rendering) that materialise
on the PC during game play, although in our study we do not attempt to measure these aspects.
Rather, we consider the technical performance of the PC’s hardware and software sub-system
technologies, which enables the delivery of system level qualities. We additionally recognise that
the evolution of these technologies is brought about by the strategic actions of firms in their
respective industries. Game developers, in particular, consider factors such as potential markets
and the compatibility of their technologies with complementary hardware technologies. In this
manner, these firms attempt to reconcile the highest level of technology that can be embedded in
the game design and the level of game technology that can actually be supported by the hardware
that is available in the market.
 1102 O. Dedehayir and S.J. Mäkinen

Rather than considering all sub-systems that constitute the highly complex PC system, our
analysis has concentrated on the most central technical sub-systems for gaming performance. In
this manner, we have studied the technological evolution of sub-systems, whereby one central
sub-system utilises the technological performance provisioned by another, to reflect the evolution
of gaming performance that is delivered at the system level. The illustrative empirical study has
therefore analysed the PC game software sub-system and its co-evolution with two important
and interdependent hardware sub-systems: the CPU and GPU. However, in order to limit our
study we have defined the system under analysis to include only two sub-systems. In our first
system definition, we have included and analysed the CPU and PC game sub-systems, and in
the second, we have included and analysed the GPU and PC game sub-systems. The system
definitions are important because the reverse salient can only be identified with respect to a
technological system that is defined a priori. Consequently, the reverse salient must be a sub-
system that lies within the system’s boundary and not one that is excluded from the system as an
Downloaded by [University of Calgary] at 12:51 02 October 2013

object under study. In this manner, exogenous factors may be antecedents of reverse salients, but
not reverse salients themselves. This further suggests that the reverse salient can be identified only
relative to a specific context, such that different reverse salients may be identified in alternative
contexts.1
In our analysis, the CPU is a central hardware component of the PC which is specialised to
interact with other components (e.g. hardware and software), to process data and enable the com-
putational performance of the computer. The GPU, in comparison, is specialised for the graphics
performance (qualitative aspects of the image on a display) of the PC in software applications,
and takes some of the processing responsibility away from the CPU for this purpose. The PC
game software, in turn, provides the instruction set that is processed by the CPU and GPU to
translate the data into the gaming performance made available on the PC. The centrality and
interdependence of these sub-systems for computer gaming is, further, made apparent from the
software developers’ stipulation of performance requirements, specifically for the CPU and GPU
components, to ensure the functionality of their software.

Technological performance indicators

For each of these nested sub-systems we selected technological performance indicators that allow
the analysis of performance evolution and the measurement of reverse salience. While the perfor-
mance of the CPU can be measured along different dimensions (for example, MIPS – millions of
instructions per second – number of transistors, bus bandwidth), we have selected the performance
indicator of processing speed, operationalised in this paper as the clock speed of the processor (in
Hertz [Hz]). The processor speed indicates the CPU’s speed of operation, governing the compu-
tational performance of the PC through its interaction with software programs such as PC games.
Higher speeds mean faster data manipulation and increased computer performance, resulting in
enhanced gaming performance at the system level. This is the most appropriate performance indi-
cator used in our analysis, because it is the central performance parameter stipulated by game
developers and publishers for game functionality. Similarly, while the performance of the GPU
can be measured along different dimensions, the graphics memory, measured in megabytes (MB),
is the central indicator stipulated by game developers and publishers for game functionality, and
is for this reason considered in our study. The graphics memory denotes the GPU’s ability to
store and make available graphics data to the PC. Larger memory capacity means increased com-
puter graphics performance, resulting in enhanced gaming performance delivered at the system
level.
 Determining reverse salient types and evolutionary dynamics of technology systems 1103

For the PC game sub-system we have considered that the software is designed to utilise a certain
level of hardware performance. For this reason, the game software stipulates a set of minimum
hardware performance requirements with which the software will function as designed. However,
the game software can additionally be designed to deliver a higher level of gaming experience,
for example by making additional game features accessible during game play (i.e. recommended
specifications). Nevertheless, in our study we have focused only on the functional performance
of the hardware and software sub-systems, and not on the additional gaming experience that may
be delivered by these sub-systems. We have therefore used the minimum hardware requirement
stipulated by the software as the technological performance indicator to study the PC game sub-
system’s evolution. In this manner, we identify a particular PC game software as bearing a higher
level of technological performance than another software, when its stipulated minimum hardware
requirement is higher than that of the latter. This is because a PC game that requires a higher level
of hardware performance will utilise a greater amount of hardware capacity to function, and, in
Downloaded by [University of Calgary] at 12:51 02 October 2013

turn, enable the delivery of a higher level of gaming performance at the system level.

The data
We collected data on CPU processor speeds from processor performance databases found
on the Intel and AMD (Advanced Micro Devices) corporate web sites (www.intel.com and
www.amd.com, respectively), the two primary manufacturers of CPUs. In the generation of this
data we omitted CPUs that were designed specifically for server, workstation, or similar applica-
tions, because our study focuses on games that are played on personal computers. Additionally,
we omitted multi-core CPUs from the data because this new platform represents a significant shift
in the technological paradigm of the CPU, whereby processing speed may no longer be the most
prevalent indicator of CPU performance. Similarly, we accessed GPU graphics memory data from
the corporate web sites of NVIDIA (www.nvidia.com) and ATI (acquired by AMD on October
2006), the two dominant players in the graphics processor industry.
The data on PC game minimum processor speed requirements and minimum graphics memory
requirements were, in turn, collected from the web sites of game publishers and game developers,
as well as the gaming community Gamespot.com and a major on-line PC game vendor, Ama-
zon.com. The list of PC games used in our empirical analysis was generated as a result of filtering
steps that ensured the reliability of the data. First, we limited the list of PC games to only those
that had been reviewed and rated by either one reputable online source, Gamespot.com, or one
reputable printed source, PC Gamer magazine (the reputation of these sources was identified by
experts from the game development industry). A game which had not been reviewed by either of
these sources was discarded on the basis that it did not meet the quality standards of the gaming
industry at large, and that its inclusion could therefore reduce the reliability of our data. Second,
we limited the list of PC games to those which had been launched in the USA.
All in all, we gathered data representing the technological performance levels of these sub-
systems over a period stretching fromAugust 1995 until the end of 2008. We selected the beginning
of the analysed timeframe to correspond with the launch of the Windows 95 operating system
that established a new technological platform upon which hardware and PC game software could
be developed, thereby producing a significant change in the gaming industry (Hayes and Dinsey
1995). Altogether, 603 CPU, 123 GPU and 3064 PC game related data points were collected and
evaluated to reveal the technological evolution curves of the sub-systems and, subsequently, the
temporal change in reverse salience as measured by absolute and proportional performance gap.
 1104 O. Dedehayir and S.J. Mäkinen

Results and discussion

Figures 3 and 4 display the technological performance evolutions that have taken place in two
different systemic contexts: the first comprising the CPU and PC game sub-systems, and the
second comprising the GPU and PC game sub-systems. In these figures, the produced development
curves depict the successively higher levels of technological performance that are delivered by
the sub-systems over time.
Figures 3 and 4 indicate that the CPU and GPU sub-systems form technological performance
frontiers, thereby defining the technological performance potentials for the reverse salient in
each technological system context. Further, the figures show that the PC game sub-system trails
Downloaded by [University of Calgary] at 12:51 02 October 2013

Figure 3. CPU and PC game sub-system technological performance evolutions with respect to processor
speed.

Figure 4. GPU and PC game sub-system technological performance evolutions with respect to graphics
memory.
 Determining reverse salient types and evolutionary dynamics of technology systems 1105

the technological performance of the CPU as well as the GPU sub-systems over time. These
results suggest that the PC game sub-system has the potential to attain higher levels of minimum
hardware requirement and increase the level of overall system performance. However, the PC
game sub-system does not fully utilise this potential provisioned by the CPU and GPU sub-
systems. As a result, the gaming performance delivered on the PC system is hampered by the
software performance. We therefore ascribe the PC game sub-system as the reverse salient within
each systemic context, because it delivers the lowest level of technological performance when
compared to the performance of the interdependent hardware sub-system, and at the same time
hinders the system level performance delivery.
Since the reverse salient limits the level of performance delivered by the technological system,
developments in reverse salient performance can indicate overall system evolution. However, to
understand system dynamics through the level of performance potential that remains unutilised by
the reverse salient, and therefore the system, it is additionally necessary to measure the magnitude
Downloaded by [University of Calgary] at 12:51 02 October 2013

of reverse salience. Figures 5 and 6 subsequently plot the absolute performance gap measure of
reverse salience and its temporal change, derived from the performance evolutions of the CPU
and PC game sub-systems, and the GPU and PC game sub-systems, respectively.
Figure 5 highlights two distinct periods: the first denoting absolute performance gap increase
(to a value of 2127 MHz) and the second denoting absolute performance decrease, with a turning
point in the year 2002. In contrast, Figure 6 illustrates an ever-increasing absolute performance
gap between the GPU and PC game sub-systems, particularly emphasizing a rapid incline in the
year 2004. This development underlines the PC game sub-system’s growing inability to exploit
the graphics memory performance that is available.
While these findings denote system level dynamics through the absolute performance gap mea-
sure, they are nevertheless impartial to the levels of performance attained by the sub-systems.
To compare system development at different points in time, taking into consideration the level
of technological evolution across sub-systems, it is necessary to evaluate the proportional per-
formance gap measure of reverse salience. Figures 7 and 8 present the temporal changes in the
proportional performance gap measures, derived from the performance evolutions of the CPU
and PC game sub-systems, and the GPU and PC game sub-systems, respectively.

Figure 5. Absolute performance gap measures over time between CPU and PC game sub-systems.
 1106 O. Dedehayir and S.J. Mäkinen
Downloaded by [University of Calgary] at 12:51 02 October 2013

Figure 6. Absolute performance gap measures over time between GPU and PC game sub-systems.

Figure 7. Proportional performance gap measures over time between CPU and PC game sub-systems.

Figure 7 demonstrates two distinct periods of evolution, similar to that observed in the absolute
performance gap evolution shown in Figure 5. The initial period signifies an increasing propor-
tional performance gap until the year 2003, followed by a period of proportional performance gap
decrease. Interestingly, the latter period coincides in particular with the technical limitations that
materialised in the single-core CPU design. Although the introduction of the multi-core CPU2
circumvented the technical reverse salients in the CPU sub-system in 2005, further gains in pro-
cessor speed performance for the single-core CPU could not be observed after this time (see
Figure 3). This is because the transition from the single-core to the multi-core design represents
a shift in the CPU technology paradigm, thereby elevating other performance indicators (e.g. the
number of cores) as more central to overall CPU performance than processor speed. The resulting
plateau of development in the CPU sub-system has subsequently allowed the PC game sub-system
 Determining reverse salient types and evolutionary dynamics of technology systems 1107
Downloaded by [University of Calgary] at 12:51 02 October 2013

Figure 8. Proportional performance gap measures over time between GPU and PC game sub-systems.

to close the performance gap, both in absolute and proportional terms, through a higher rate of
development. This finding, moreover, suggests that the PC game sub-system may experience a
significant shift in time as well. This is because the discontinuous change from single to multi-core
CPU design provides the interdependent PC game sub-system with an alternative performance
parameter, namely the number of cores, which emerges as being central for software function.
In contrast, the proportional performance gap between the GPU and PC game sub-systems
shown in Figure 8 oscillates, predominantly between the values of 50% and 75%. This outcome
can be attributed to the consecutive levels of graphics performance that increase in a stepwise
fashion, specifically in powers of two (e.g. 8–16 MB and 16–32 MB), rather than through gradual
increments. As a result, the evaluated proportional performance gaps between the technological
performance frontier and the reverse salient essentially represent technological generation gaps.
Thus, a proportional performance gap of 50% (e.g. when the technological performance frontier is
at 16MB and reverse salient at 8MB) represents a difference of one generation, while a proportional
performance gap of 75% (e.g. when the technological performance frontier is at 32MB and the
reverse salient is at 8MB) represents a difference of two generations.
To further study the dynamics of change in the two system contexts, we analyse the relation-
ship between the absolute and proportional performance gap measures of reverse salience, as
represented in Figures 9 and 10.
Figure 9 plots the absolute and proportional measures against one another and reaffirms the
earlier observed temporal change in reverse salience pertaining to processor speed performance.
The figure emphasises a period of reverse salience increase, in both absolute and proportional
terms, followed by a period of decrease in both measures, separated by a short, intermediate
duration of time. In contrast, the development of the GPU and PC game sub-systems in Figure 10
shows a highly erratic pattern, leading up to 2004. This sequence of sudden changes is particularly
pronounced in the proportional performance gap, and is likely to be caused by successive gener-
ation leaps in the PC game sub-system performance, following the leaps in the GPU sub-system
performance.
Both Figures 9 and 10 illustrate the changes in reverse salience magnitude across time, as well
as the intermediate turning points between increasing and decreasing magnitudes. Consequently,
 1108 O. Dedehayir and S.J. Mäkinen
Downloaded by [University of Calgary] at 12:51 02 October 2013

Figure 9. Plot of absolute and proportional performance gap measures for CPU and PC game sub-systems.

Figure 10. Plot of absolute and proportional performance gap measures for GPU and PC game sub-systems.

it becomes possible to trace the types of reverse salients that emerge across the timeframe of
analysis using the typology of reverse salients depicted in Figure 1. Figures 11 and 12 utilise this
typology in identifying reverse salient types with respect to the evolution of the CPU and PC
game sub-systems, and the evolution of the GPU and PC game sub-systems, respectively.
Figure 11 highlights three periods of development which align with the observations made in
Figure 9. The first is a period of alternation between the prohibitive and intermediating reverse
salient types, between August 1995 and August 2002 (points A–E in Figure 11). This span of time,
overall, coincides with the PC game sub-system falling increasingly behind the technological
performance frontier. The second period, between August 2002 and December 2002 (point F in
Figure 11), while short-lived, underlines a point of transition in the system dynamics. In turn,
during the following period that stretches from December 2002 until December 2008 (points
G–K in Figure 11), the PC game sub-system reduces the absolute as well as the proportional
 Determining reverse salient types and evolutionary dynamics of technology systems 1109
Downloaded by [University of Calgary] at 12:51 02 October 2013

Figure 11. Reverse salient types across successive time spans for CPU and PC game sub-systems.

Figure 12. Reverse salient types across successive time spans for GPU and PC game sub-systems.

performance gap of reverse salience. Subsequently, the PC game sub-system alternates between
the progressive and intermediating reverse salient types during this time. The shift in dynamics
indicates that the rate of technological development promoted by firms in the game development
industry is increased in relation to the rate of technological development in the CPU industry.
This result may additionally point to the greater aptitude of game developing firms to reconcile
the level of technology embedded in the game design and the level of game technology that can
be supported by the hardware available in the market.
The continuation of performance gap closure, as also evidenced in Figure 7, may suggest
that the PC game sub-system is approaching a discontinuous change. A shift from the existing
performance parameter of minimum processor speed requirement, to a new central performance
parameter, for example the minimum number of processor cores, may therefore materialise in time.
 1110 O. Dedehayir and S.J. Mäkinen

This shift in the central performance parameter can simultaneously lead to an abrupt increase in
the proportional performance gap and a decrease in absolute performance gap. As a result, the PC
game sub-system may shift, in time, from the progressive to the reorienting reverse salient type.
Such a shift concurrently denotes the reorientation of the technological focus of game developing
firms, who must, in turn, consider the installed base of the hardware technology with respect to
the new central performance parameter.
Figure 12, conversely, illustrates the oscillating change in reverse salience magnitude also
observed in Figure 10. In this manner, the PC game sub-system generally alternates between the
progressive and prohibitive reverse salient types over successive periods of time. However, the
figure also shows occasions when the reverse salient type can be categorised either as interme-
diating or prohibitive (points C and H in Figure 12). This finding coincides with the stepwise
development that is inherent to graphics related performance (see Figure 8), which increases
the probability of observing unchanging proportional performance gap, despite changing absolute
Downloaded by [University of Calgary] at 12:51 02 October 2013

performance gap. Overall, however, the sequence of erratic changes indicates the PC game sub-
system’s responsive utilisation of available GPU performance, thereby signifying the continuing
importance of graphics related technological performance to the game development industry.

Conclusion and implications

The objective of the paper was to develop measures of reverse salience and a typology of reverse
salients that can be applied in the analysis of technological systems. To this end, we have first
developed a proportional performance gap measure that quantifies the state of reverse salience and
complements the absolute performance gap measure developed earlier by Dedehayir and Mäkinen
(2008). This proportional measure importantly allows the comparison of system development at
different points in time, taking into account the level of technological evolution across sub-
systems. Moreover, the absolute and proportional performance gap measures, taken together,
inform us of the system level development and at the same time help identify different types of
reverse salients. We have, subsequently, demonstrated the applicability of these measures and the
typology of reverse salients in an illustrative empirical study of an evolving technological system.
Our empirical illustration studied two essential systemic contexts in PC gaming as determined
from the PC game manufacturers’ software requirements. We analysed the CPU and PC game
sub-systems in the first systemic context and the GPU and PC game sub-systems in the second.
In our analyses, we studied the technological evolutions of these sub-systems and evaluated the
magnitudes of reverse salience. Our findings showed that reverse salience evolved through periods
of increasing followed by decreasing absolute and proportional performance gaps for the CPU
and PC game sub-systems. In contrast, with respect to the evolution of the GPU and PC game
sub-systems, we observed continuous increase in the absolute performance gap, and oscillation
in the proportional performance gap measures. From these results, we determined the types of
reverse salients that emerged across time. Our findings from the analysis of the CPU and PC
game sub-systems illustrated shifts from prohibitive to intermediating, and from intermediating
to progressive reverse salient types. In comparison, the study of the GPU and PC game sub-
systems demonstrated alternation between the progressive and prohibitive reverse salient types
over consecutive periods of time. In summary, these findings illustrate that the magnitude, direction
and types of dynamics in system level evolution can be identified through the reverse salience
measures.
Overall, our empirical analyses showed that the PC game sub-system delivered the lowest
level of performance and therefore hampered the technological system’s delivery of gaming
 Determining reverse salient types and evolutionary dynamics of technology systems 1111

performance. Consequently, the PC game sub-system was determined to be the reverse salient
within each systemic context. This finding rests fundamentally on the premise that the reverse
salient must be a sub-system which lies within the a priori defined system. However, reverse
salients may be caused by factors that are exogenous to the system (Hughes 1983). For example,
a reverse salient may emerge when there is excessive resistance from the system’s environment
(analogous to enemy forces that oppose a battle line’s forward movement), thereby limiting
further performance improvement inside the system. In this manner, external forces may also
act as antecedents to the PC game sub-system’s emergence as a reverse salient, given that the
development of PC game technology is driven by the strategic actions of firms inside the game
development industry and that these actions are contingent on environmental considerations. Such
external factors have been demonstrated to influence systemic development in earlier literature; for
example, the intervention of institutional agents and policy makers (Araujo and Harrison 2002),
changing market behaviour (Adner and Levinthal 2001), as well as network externalities (Katz
Downloaded by [University of Calgary] at 12:51 02 October 2013

and Shapiro 1985; Schilling 1998) and the installed base (Katz and Shapiro 1986). Consequently,
delimiting the system from its environment emerges as an imperative first step in the analysis of
reverse salience.
Furthermore, our developed typology of reverse salients allows us to understand the state
of system dynamics as well as future trends of system evolution. Hence, a period marked by
a progressive reverse salient, may, for example, signify a time of intense competition in the
industry to improve central technological performance parameters, resulting in improved benefits
to customers and therefore the gaining of market share (e.g. Sheremata 2004).A prohibitive reverse
salient, by contrast, may build-up pressure to remove the development blockage, where this reverse
salient emerges as a significant hindrance. This may, in turn, induce a radical innovation that leads
to accelerated technology-based competition in the industry and renders the current prohibitive
reverse salient obsolete (Dewar and Dutton 1986). The prohibitive reverse salient may alternatively
prove to be inconsequential to overall system performance, in which case this sub-system may
be excluded from the system altogether in time, in what could be described as a ‘blitzkrieg’
resolution. Third, the period marked by an intermediating reverse salient is likely to take place in
the transition from one reverse salient type to another. Consequently, the intermediating reverse
salient may bring about a change in competitive dynamics and, in turn, open up windows of
opportunity for strategic manoeuvring (e.g. Lieberman and Montgomery 1988). Conversely, the
reorienting reverse salient represents an abrupt change, as the reverse salient sub-system reorients
its development focus to a new technological performance frontier. This shift may alter the basis of
competition as the performance parameter that is central to the industry is supplanted by another
(Christensen 1997).
While our developed typology of reverse salients has been illustrated in this paper with limited
scope, the method is applicable to larger systemic contexts, as well as to a variety of technolog-
ical systems. These represent natural extensions of our present work. Our proposed framework
focuses on interlinks between different sub-systems and studies the provisioning of technological
performance in one sub-system and the utilisation of this performance by another, interdepen-
dent sub-system. However, our method does not differentiate between the types of interplay that
may exist among sub-systems. A fruitful extension of our present work is therefore to distin-
guish the types of system dynamics, such as purely evolutionary or co-evolutionary dynamics,
that may result from the different types of sub-system interconnections, such as between sim-
ply interlinked and between complementary sub-systems (Teece 1986). Furthermore, while our
framework derives the dynamics of change from the respective provisioning and utilisation of
technological performance among sub-systems, it does not inform of the length of time required
 1112 O. Dedehayir and S.J. Mäkinen

for a particular sub-system to utilise the performance that is made available. Nevertheless, such
time-based measurement can reveal the temporal dynamics, in addition to the pure performance
based dynamics, in the evolution of technological systems. The development of an additional,
time-based measure of reverse salience is subsequently an important avenue for future research.
Finally, our proposed typology of reverse salients offers a useful tool for understanding future
changes in system dynamics and the competitive landscape (Prahalad and Hamel 1994; Bettis
and Hitt 1995). Therefore, by tracing the sequence of reverse salient types over time, subsequent
research can, with further empirical studies, increase our understanding of the sequence and nature
of shifts between different types of reverse salients. The typology of reverse salients additionally
facilitates the recognition and possibly the forecasting of turning points in technological system
evolution, as witnessed in the intermediating reverse salient shift with respect to the CPU and PC
game sub-system evolutions, and therefore changes in the competitive landscape.
Downloaded by [University of Calgary] at 12:51 02 October 2013

Notes
1. We are grateful for the reviewer’s contribution to this important aspect of system definition.
2. http://www.intel.com/pressroom/kits/core2duo/pdf/microprocessor_timeline.pdf.

Notes on contributors
Ozgur Dedehayir, PhD, is a Research Fellow at the Tampere University of Technology (TUT), Finland. He is a member
of CITER (Center for Innovation and Technology Research), conducting his research on the evolution of technological
systems. He received a BSc in Mechanical Engineering from the University of Melbourne, Australia, and an MSc in
Industrial Management from TUT, Finland.

Saku J. Mäkinen, PhD, is Professor of Strategic Management of Technology and Innovation at Tampere University of
Technology (TUT), Finland. He has formerly been with the National University of Singapore. His research has appeared
or is forthcoming in leading forums such as Technological Forecasting and Social Change, Technovation and IEEE Trans-
actions on Engineering Management. His research interests include strategic technology and innovation management,
and technology and industry co-evolution.

References
Abernathy, W.J., and K.B. Clark. 1985. Innovation: Mapping the winds of creative destruction. Research Policy 14(1):
3–22.
Adner, R., and D. Levinthal. 2001. Demand heterogeneity and technology evolution: Implications for product and process
innovation. Management Science 47(5): 611–28.
Andersen, B., 1999. The hunt for S-shaped growth paths in technological innovation: A patent study. Journal of
Evolutionary Economics 9: 487–526.
Anderson, P., and M.L. Tushman. 1990. Technological discontinuities and dominan designs: A cyclical model of
technological change. Administrative Science Quarterly 35(4) 604–33.
Araujo, L., and D. Harrison. 2002. Path dependence, agency and technological evolution. Technology Analysis & Strategic
Management 14(1): 5–19.
Bettis, R.A., and M.A. Hitt. 1995. The new competitive landscape. Strategic Management Journal 16(Special issue:
Technological transformation and the new competitive landscape): 7–19.
Christensen, C.M. 1992. Exploring the limits of the technology S-curve. Part 1: Component technologies. Production and
Operations Management 1(4): 334–57.
Christensen, C.M. 1997. The innovator’s dilemma: When new technologies cause great firms to fail. Boston, MA: Harvard
Business School Press.
Christensen, C.M., F.F. Suarez, and J.M. Utterback. 1998. Strategies for survival in fast-changing industries. Management
Science 44(12): S207–S220.
 Determining reverse salient types and evolutionary dynamics of technology systems 1113

Clark, K.B. 1985. The interaction of design hierarchies and market concepts in technological evolution. Research Policy
14(5): 235–51.
Constant, E.W. 1987. The social locus of technological practice: Community, system, or organization? In The Social
construction of technological systems, ed. W.E. Bijker, T.P. Hughes and T.P. Pinch, 233–42. Cambridge, MA: The
MIT Press.
Constant, E.W.I. 1980. The origins of the turboject revolution. Baltimore, MD: John Hopkins University Press.
Das, S.S., and A.H. van de Ven. 2000. Competing with new product technologies: A process model of strategy. Management
Science 46(10): 1300–16.
Dedehayir, O., and S.J. Mäkinen. 2008. Dynamics of reverse salience as technological performance gap: An empirical
study of the personal computer technology system. Journal of Technology Management and Innovation 3(3): 55–66.
den Hond, F. 1998. The ‘similarity’ and ‘heterogeneity’ theses in studying innovation: Evidence from the end-of-life
vehicle case. Technology Analysis & Strategic Management 10(4): 529–43.
Dewar, R.D., and J.E. Dutton. 1986. The adoption of radical and incremental innovations: An empirical analysis.
Management Science 32: 1422–33.
Dosi, G. 1982. Technological paradigms and technological trajectories. Research Policy 11: 147–62.
Downloaded by [University of Calgary] at 12:51 02 October 2013

Esposito, E., and M. Mastroianni. 1998. Technological evolution of personal computers and market implications.
Technological Forecasting and Social Change 59: 235–54.
Fisher, J.C., and R.H. Pry. 1971. A simple substitution model of technological change. Technological Forecasting and
Social Change 3(1): 75–88.
Foster, R.N. 1986. Innovation: The attacker’s advantage. New York: Summit Books.
Hardstone, G.A.P. 2004. Capabilities, structures and strategies re-examined: Incumbent firms and the emergence of
complex product systems (CoPS) in mature industries. Technology Analysis and Strategic Management 16(2):
173–96.
Hayes, M., and S. Dinsey. 1995. Games War. London: Bowerdean.
Hugh, M.J., M.Y. Roche, and S.J. Bennett. 2007. A structured and qualitative systems approach to analyzing hydrogen
transitions: Key changes and actor mapping. International Journal of Hydrogen Energy 32: 1314–23.
Hughes, T.P. 1983. Networks of power: Electrification in Western Society, 1880–1930. Baltimore, MD: The Johns Hopkins
University Press.
Hughes, T.P. 1987. The evolution of large technological systems. In The Social Construction of Technological Systems,
ed. W.E. Bijker, T.P. Hughes and T.P. Pinch, 51–82. Cambridge, MA: The MIT Press.
Katz, M.L., and C. Shapiro. 1985. Network externalities, competition, and compatibility. The American Economic Review
75(3): 424–40.
Katz, M., and C. Shapiro. 1986. Technology adoption in the presence of network externalities. Journal of Political Economy
94: 822–41.
Krell, G. 1992. The innovation impact model: A tool to study the impact of technological change. Technology Analysis &
Strategic Management 4(3): 211–26.
Lieberman, M.B., and D.B. Montgomery. 1988. First-mover advantages. Strategic Management Journal 9: 41–58.
MacKenzie, D. 1987. Missile accuracy: A case study in the social processes of technological change. In The Social
Construction of Technological Systems, ed. W.E. Bijker, T.P. Hughes and T.J. Pinch, 195–222. Cambridge, MA: The
MIT Press.
Moors, E. 2006. Technology strategies for sustainable metals production systems: A case study of primary aluminium
production in the Netherlands and Norway. Journal of Cleaner Production 14(12): 1121–38.
Moors, E.H.M., and P.J. Vergragt. 2002. Technology choices for sustainable industrial production: transitions in metal
making. International Journal of Innovation Management 6(3): 277–99.
Murmann, J.P., and K. Frenken. 2006. Toward a systematic framework for research on dominant designs, technological
innovations, and industrial change. Research Policy 35: 925–52.
Nelson, R.R., and S.G. Winter. 1977. In search of useful theory of innovation. Research Policy 6: 36–76.
Pilkington, A. 1998. The fit and misfit of technological capability: Responses to vehicle emission refulation in the US.
Technology Analysis & Strategic Management 10(2): 211–24.
Poole, S. 2004. Trigger happy: Video games and the entertainment revolution. New York: Arcade Publishing.
Prahalad, C.K., and G. Hamel. 1994. Strategy as a field of study: Why search for a new paradigm? Strategic Management
Journal 15: 5–16.
Rosenberg, N. 1969. The direction of technological change: Inducement mechanisms and focusing devices. Economic
Development and Cultural Change 18: 1–24.
Rosenberg, N. 1976. Perspectives on technology. Cambridge: Cambridge University Press.
Sahal, D. 1981. Patterns of technological evolution. London: Addison-Wesley.
 1114 O. Dedehayir and S.J. Mäkinen

Sahal, D. 1985. Technological guideposts and innovation avenues. Research Policy 14: 61–82.
Schilling, M.A. 1998. Technological lockout:An integrative model of the economic and strategic factors driving technology
success and failure. Academy of Management Review 23(2): 267–84.
Sheremata, W.A. 2004. Competing through innovation in network markets: Strategies for challengers. Academy of
Management Review 29(3): 359–77.
Simon, H. 1962. The architecture of complexity. Proceedings of the American Philosophical Society 106(6): 467–82.
Skyttner, L. 2001. General systems theory: Ideas and applications. Singapore: World Scientific.
Takeishi, A., and K. Lee. 2005. Mobile music business in Japan and Korea: Copyright management institutions as a reverse
salient. Journal of Strategic Information Systems14: 291–306.
Teece, D.J. 1986. Profiting from technological innovation: Implications for integration, collaboration, licensing and public
policy. Research Policy 15(6): 285–305.
Tushman, M.L., and J.P. Murmann. 1998. Dominant designs, technology cycles, and organizational outcomes. Research
in Organizational Behavior 20: 231–66.
Vincenti, W. 1994. The retractable landing gear and the Northrup ‘Anomaly’: Variation-selection and the shaping of
technology. Technology and Culture 35(1): 1–33.
Downloaded by [University of Calgary] at 12:51 02 October 2013

von Bertalanffy, L. 1969. General system theory. New York: George Braziller.
Wade, J. 1995. Dynamics of organizational communities and technological bandwagons: An empirical investigation of
community evolution in the microprocessor market. Strategic Management Journal 16(Special issue, Summer):
111–33.
Zahra, S.A., and J.G. Covin. 1993. Business strategy, technology policy and firm performance. Strategic Management
Journal 14(6): 451–78.
J Knowl Econ
DOI 10.1007/s13132-016-0390-8

Involving Consumers: The Role of Digital Technologies

in Promoting ‘Prosumption’ and User Innovation

Thierry Rayna 1 & Ludmila Striukova 2

Received: 2 November 2015 / Accepted: 17 June 2016

# The Author(s) 2016. This article is published with open access at Springerlink.com

Abstract Since the First Industrial Revolution, consumers have been mainly consid-
ered as playing a passive role with regard to production. However, the recent decades
have seen a progressive growth in consumer involvement in production processes, for
instance, in the form of user innovation, DIY and mass customisation. Yet, it was not
until the advent of digital technologies that consumers’ input in production processes
could become really significant in all dimensions of production (design, manufacturing
and distribution). This increased (sometimes leading) role of consumers in production
has been referred to as ‘prosumption’. While prosumption has so far been mainly
significant online (where consumers have arguably taken over the creation and distri-
bution of content), recent advances in digital technologies (mobile networks, 3D
printing) have enabled prosumption to reach to world of physical objects, as illustrated
by the increased importance of consumer-made goods (‘makers’ movement) and of the
‘sharing economy’. Although prosumption can be highly disruptive to markets, there
has not been so far a thorough investigation of the nature of this phenomenon. Yet,
understanding this nature is critical to apprehend disruptions. Furthermore, while digital
technologies have been acknowledged as a key factor in the rise of prosumption, the
exact role of each set of technologies has not yet been studied, which is a problem, as
different technologies enable different types of prosumption. The aim of this article is to
fill these two gaps. To do so, a framework of prosumption, which encompasses both
inputs and outputs of consumer contribution, is introduced. This framework is then
applied to multiple case studies, which enables to show how prosumption has evolved
and what changes lie ahead.

* Ludmila Striukova
[email protected]

Thierry Rayna
[email protected]

1
Novancia Business School Paris, 3, rue Armand Moisant, 75015 Paris, France
2
University College London, Gower Street, London WC1E 6BT, UK
 J Knowl Econ

Keywords Prosumption . User innovation . Digital technologies . Sharing economy . 3D

printing

Introduction

As a form of Open Innovation, User Innovation has had a growing importance over the
past few decades, both in the literature and as a part of firms’ strategies. While firms
have always used to some extent user input as a source of innovation (Rayna and
Striukova 2010), the concepts of ‘User Innovation’ and ‘Lead Users’ have gained a
widespread acceptance following the seminal works of Eric von Hippel (von Hippel
1976, 1977, 1978, 1986). Yet, both in these works and in the abundant literature on
User Innovation that has ensued, user input is generally only considered at the
conception stage. Regardless of how inventive they are, users are seen as bound to
the firm, because they lack the manufacturing, distribution and commercialisation
capabilities that only firms have.
While this might have indeed been the case until the late 1990s, the progressive
diffusion of digital technologies (in particular, following the Internet ‘boom’ of the mid
1990s) has radically and durably changed the relationship between users and firms.
Indeed, a common trait of all digital technologies is that that they empower users, and
in particular consumers, with the ability to create, replicate, distribute and (in some
instances) commercialise (Rayna 2008).
The wide availability of digital technologies means that user innovation is no longer
restricted to professional end-users and, indeed, consumers have been increasingly
engaged in co-creation activities with firms (Rayna and Striukova 2015). In fact,
looking at the massive quantity of User Generated Content available online, one could
even argue that consumers have in some cases superseded firms, as they now play a
leading role in design, ‘manufacturing’ and distribution.
As a result of their increasingly active role in the production processes, consumers
have progressively become ‘prosumers’. Since innovative users are no longer bound to
the firms, countless user innovations have taken place independently from the firms
and, sometimes, even against the firms’ will (Zwass 2010; Rayna and Striukova 2015).
As a matter of fact, whether legitimately or not, the number of cases where consumers
directly compete with firms (for instance by diffusing innovations that have been
rejected by the firm) has sharply increased over the past few years. 1 Hence, as a
consequence of the diffusion of digital technologies, the relationship between firms and
users has turned from simple cooperation to (potential) coopetition.
While coopetition between prosumers and firms only used to be prevalent online,
the recent advances in digital technologies have led this phenomenon to spread to the
rest of the economy. For instance, digital manufacturing technologies, such as 3D
printing and additive manufacturing, provide end-users (and progressively consumers)
with the ability to manufacture objects. In fact, there have already been numerous cases
of users manufacturing and selling improved versions of existing commercial products
or items complementing such products. 2 Such technologies have even given rise to
consumer-borne ecosystems around popular products (e.g. Legos, Star Wars). Besides

1
See, for instance, the example of the ‘Game of Thrones’ smartphone dock, p. 25.
2
See, for instance, the case of the Square Helper, p. 17.
J Knowl Econ

manufacturing, recent advances in digital technologies have also began to considerably

disrupt the service industries (e.g. taxis, holiday rentals), with prosumers starting to
supply services in direct competition with businesses, by renting out their cars (or
driving people around), their flats, their expertise, or any other asset they might own;
the inescapable ‘sharing economy’ is indeed yet another form of prosumption.
As the prosumer phenomenon is arguably spreading to the whole economy, it is
critical to understand better the roots and evolution of this phenomenon. To do so, one
has to consider how, and to which extent, digital technologies have given means to
consumers to contribute to production processes, and whether such contributions are
related to design (when user innovation traditionally takes place), manufacturing or
distribution.
To do so, this article introduces an ‘input-output’ prosumption framework that
enables to analyse consumers’ involvement in the production process with regard to
both the type of input they contribute and the type of contribution (relative to the
production process) that is made (‘Towards a prosumption framework’ section). Then,
based on multiple case studies and using a historical approach, this prosumption
framework is used to investigate the evolution of the impact of digital technologies
on prosumption and user innovation: Sect. 3 discusses user innovation and prosumption
in the pre-digital era; Sect. 4 investigates user innovation and prosumption in the early
(mid 1990s-mid 2000s) digital era; Sect. 5 studies the impact of Web 2.0 (e.g. social
media) technologies on prosumption and user innovation; and Sect. 6 considers the
most recent impacts of digital technologies and explores the consequences of user
manufacturing (through 3D printers) and of the sharing economy. Finally, Sect. 7
tackles the issue of leveraging prosumption from a firm’s perceptive and outlines
related good practices.

Towards a prosumption framework

While the word ‘consumer’ appeared in the English language some time during the late
Middle Ages,3 the modern concept of consumer as someone who purchases goods and
services for personal use unsurprisingly came into existence in the late eighteenth
century, at the time of the First Industrial Revolution. Indeed, mechanisation, and the
resulting concentration of means of production it necessarily entailed, meant that for the
first time in history, production and consumption of goods and services became largely
disconnected.4
From then on, the perception of the importance of the role played by consumers in
the economy progressively grew, up to a point where after World War II consumption
started to be considered as the focal point and driving force of modern economies,
which had by then become ‘mass consumption societies’ (Matsuyama 2002). Yet,
while recognised as critically important, the role of consumers was still seen as
essentially passive, and even increasingly so, as both consumption choices and volumes

3
At the time, as its French root consumer (i.e., ‘burn’) implies, it referred to someone who spent recklessly
and wasted resources.
4
As opposed, for instance, to the work on a medieval farm, where people would first produce for themselves
and then trade what they did not consume in order to obtain goods or services they did not produce.
 J Knowl Econ

of consumption were thought to be triggered by corporations, through advertising

(Galbraith 1958, 1967).
Hence, the traditional view of mass consumption society is that firms are in charge
of deciding, designing, manufacturing (producing), marketing and distributing goods
and services, while the role of consumers is, plainly, to consume. Yet, this ultra-passive
view of consumers began progressively to change in the late 1970s because of the
success of ‘new’ business models, which involved some degree of consumer partici-
pation (e.g. fast-food restaurants, flat packed furniture).
Taking notice the increasing role of consumers in the production processes through
DIY (‘do-it-yourself’) activities (such as self-service, self-help and self-repair), Toffler
(1981) coined the term prosumer, a portmanteau word of ‘producer’ and consumer.
Besides the consumer involvement in production processes through DIY activities
(which was already happening at the time), Toffler (1981), considering advances in
computing, also foresaw a consumer involvement of in the design of customised
products (albeit to a fairly minor extent, i.e. providing measurement for customised
clothing or objects). Kotler (1986) developed this idea further and emphasised that
progresses in information and communication technologies would soon enable con-
sumers to creatively engage in production processes, through mass customisation (i.e.
choosing amongst a range—often wide—of options set by the producers, who then
manufacture the products on demand).
While ‘DIY’ business models certainly picked up during the 1980s, with more firms
integrating self-service into their products (e.g. factory outlets, home repairs), it took
over 3 decades for the mass customisation foreseen by both Toffler (1981) and Kotler
(1986) to take place at a significant scale. Consequently, it is not surprising that, for a
while, the prosumer concept seemed all but forgotten in the literature.5 Yet, this concept
began to regain popularity during the 2000s, following the advent of Web 2.0 technol-
ogies (and, more specifically, social media) in particular.
As noted in Rayna and Striukova (2015), Web 2.0 technologies enable consumers to
contribute and share textual and multimedia content over the Internet, with almost no
technical knowledge being required. Services such as Facebook, Flickr, YouTube, and
Wikipedia, led to a surge of consumer (or user)-generated content available online. As
highlighted in Tapscott and Williams (2006) and Ritzer and Jurgenson (2010), it
became rapidly clear that in all these services, consumers, and not firms, were actually
doing the ‘bulk of the work’. Indeed, as noted in Ritzer and Jurgenson (2010):

Prosumption was clearly not invented on Web 2.0, but given the massive
involvement in, and popularity of, many of these developments (e.g. social
networking sites), it can be argued that it is currently both the most prevalent
location of prosumption and its most important facilitator as a ‘means of
prosumption’.

However, as will be shown in the following sections, prosumption is nowadays far

from being restricted to online content and technologies, such as 3D printing and smart

5
In contrast, there was a keener interest during the 1980s and 19990s in the role played by clients—generally
in a business to business setting—in service and innovation processes.
J Knowl Econ

electricity grid, have enabled consumers to engage in the production (in the sense of
manufacturing) of physical products.
Looking at the literature, it is clear that prosumption can take many different forms,
some complex, other much simpler. To fully apprehend this phenomenon, it is thus
critical to be able to categorise prosumption activities. Yet, when doing so, a further
difficulty arises because two different aspects have to be considered: (1) what the nature
of what is contributed by the consumers to the production process is and (2) to which
aspect of the production process it is contributed. Hence, the prosumption framework
introduced in this article consists of two taxonomies: one related to the prosumption
inputs in the production process (Fig. 1), and the other related to its output (Fig. 2).
With regard to inputs, a thorough examination of the various cases of prosumption
activities shows that consumers either provide labour (in the sense of work force, as in
the case of flat packed furniture), intellectual capital (for those activities where
consumer creativity and expertise are required; for instance, customisation), or physical
capital (in the case where consumers’ own physical assets are used as a part of the
production process). Interestingly, some prosumption activities require consumers to
combine several types of input. For instance, factory outlets require the consumer to
provide labour (i.e. to carry the goods), but also to have (or rent) a car (physical
capital). Likewise, contributing to Web 2.0/social media requires both knowledge and
creativity (intellectual capital), but also to have an Internet connected device or
computer (physical capital).

Labour

Flat packed furniture

Factory outlet
Music downloads

Bespoke clothing

Physical Intellectual
capital Web 2.0 capital
Fig. 1 Taxonomy of prosumer input in the production process
 J Knowl Econ

Fig. 2 Taxonomy of prosumer output in the production process

Are there prosumption activities that require all three types of inputs? Indeed, there
are. For instance, this is typically the case for user innovation when it relates to physical
goods. User innovation requires consumers to provide intellectual capital. Consumers
engaging in user innovation also need to have tools (physical capital), and building a
prototype may also require hard labour, as no machinery is available at prototyping
stage.6
It could be argued that when consumers take part in production processes, they
either contribute ‘time’ or ‘money’ (or a mix of both). This is actually reflected in the
taxonomy presented in Fig. 1. Indeed, the two axes, labour and intellectual capital,
correspond to a contribution related to time (whether ‘force’ or ‘brains’), whereas the
third axe, physical capital, relates to money. In this respect, it can be noted that, indeed,
in many cases, prosumers face a trade-off between providing time and money (for

6
Though, as will be discussed in the following sections, additive manufacturing/3D Printing technologies
enable user innovators to mechanise the prototyping stage.
J Knowl Econ

instance, a wealthier prosumer could pay someone to mount flat packed furniture,
whereas a less wealthy prosumer engaging in user innovation would probably build a
prototype by hand instead of using a 3D printer). 7
As shown in the output taxonomy (Fig. 2), the inputs provided by prosumers can
contribute to different parts of the production process. Indeed, prosumers can contribute
to design (in its broadest sense, i.e. all that pertains to the conception phase of a product
or service), manufacturing (understood here as the reproduction, whether tangible or
intangible, of a product or service) and distribution.
Whereas ‘traditional’ production usually involves fairly little participation from the
consumers (i.e. no participation in design, no or fairly little assembly required, involve-
ment in distribution is limited to visiting a local store), other forms of production entail
a more significant contribution (Fig. 2). Factory outlets, for instance, require a signif-
icant involvement of consumers in distribution, since they need to drive (presumably)
further to the factory instead of going to a nearby store. Likewise, DIY (e.g. flat packed
furniture) requires consumers to contribute to manufacturing, as they carry out the final
assembly of the product. Finally, mass customisation leads to consumers playing a role
in design, a role that can range from minor to significant, depending on the number of
customisation options offered and the extent of the modification of the product. In some
cases (e.g. bespoke clothing), because customisation options available are not set (and
possibilities are ‘endless’), consumers can even take a leading role in design.
While many examples have been given in the literature, there is currently no clear
definition of what prosumption actually is. This is certainly due to the fact that
prosumption can take very different forms. Thus, the output taxonomy in Fig. 2 enables
to better define what prosumption is.
In the literature, prosumption is considered as a situation where consumers take an
active role in productive processes. Yet, even in the most traditional and ‘passive’ view
of consumption, consumers have always played an active role, albeit a minor one, in
distribution (e.g. going to the store) and, to an even lesser extent, in manufacturing (e.g.
unpacking the product, putting batteries in an electronic device). Thus, based on the
taxonomy in Fig. 2, prosumption can be defined as follows:

Proposition 1 Prosumption is defined as a situation where consumers are involved in

productive processes in either of the following ways:

& Consumers are involved to any extent in design.

& Consumers are involved significantly in manufacturing.
& Consumers are involved significantly in distribution.

User innovation is defined in von Hippel (1976) as a situation where users take a
leading role at the design stage of a product (or service). Users perceive the need for an

7
It could also be argued that all three types of input are always involved. For instance, music downloads
require some labour (to click on the ‘download’ button) as well as some intellectual capital (to be able to
choose which track to download). Likewise, while all necessary tools are provided with Ikea flat packed
furniture, scissors (i.e. physical capital) might be required to open the packaging, also while a detailed manual
is supplied, some elements of intellectual capital (e.g., being able to read) are also needed. However, in those
cases, it is clear that quantity of such inputs used is negligible.
 J Knowl Econ

innovation, which leads to an invention. Users then build and test a prototype, which is
then commercialised (if successful) by the company. Looking at Proposition 1, it is
clear that this definition corresponds very well to one particular case of prosumption:
when consumers are (heavily) involved in design, but not significantly in manufactur-
ing and distribution.

Thus, when carried out by consumers (there are, of course, cases when user innova-
tors are not consumers), user innovation is, in fact, a component of prosumption.8

Proposition 2 User innovation, when carried out by consumers (i.e. for final con-
sumption purpose), is a form of prosumption related to the design phase of the
productive process. In such a case, user innovators are prosumers.

The two frameworks (Figs. 1 and 2) and the two propositions introduced in this
section will be used in the following sections to analyse the impact of digital techno-
logies on prosumption and user innovation.

Prosumption in the pre-digital era

Prosumption in the pre-digital era has been largely touched upon in the previous
section. Examples drawn from the literature, in particular Toffler (1981), Kotler
(1986) and von Hippel (1976), as well as the related prosumer inputs and outputs they
entail, are summarised in Table 1. Looking at these examples, it is clear that they are
rather unidimensional, both from the input and the output perspective. The only
exception is user innovation, as in the pre-digital era, it generally involved all three
types of input.
Indeed, innovating consumers had not only to think but also to carve, solder,
cut, etc. with their own tools in order to build a prototype. For instance, when
mountain biking began in the early 1970s, existing commercial bikes were not
suited to this type usage. As a result, early users created their own bikes out of
strong old bike frames (or built new frames by soldering thicker tubes), with
balloon tires and motorcycle lever-operated drum brakes (Lüthje et al. 2005).
They were therefore using their design (intellectual capital), spare parts and
tools (physical capital), as well as labour.
However, the relative level of contribution related to each type of input did
essentially depend on the case at hand (for instance, building the prototype
might require a significant amount of intellectual capital but little labour or
investment in tools). With regard to output, user innovators were in some cases
leading the development of new products, while in other cases, they only made
fairly minor contributions to design. In all cases, however, pre-digital user
innovation was on the output side almost always restricted to design, as users
had no means to manufacture (on a large scale) and distribute their innovation
and, therefore, had to rely on companies to do so.

8
In contrast, User Innovation carried out by businesses (e.g. the traditional example of scientific instruments
used by universities and R&D labs) is not prosumption.
J Knowl Econ

Table 1 Pre-digital forms of prosumption: prosumers’ input and output

Input Output

Labour Intellectual Physical Design Manufacturing Distribution

capital capital

Flat-packed furniture Significant – – – Significant (Minor)

Factory outlet (Minor) – Significant – – Significant
(simple) Mass – Minor – Minor – –
customisation
Bespoke clothing – M/S/L M/S/L – –
User Innovation M/S/L M/S/L M/S/L M/S/L – –

Prosumption in the early digital era

For most consumers in the ‘more economically developed countries’, the early digital
era spanned from the mid-1990s to the mid-2000s. In the mid-1990s, the Internet, but
also other forms of digital replication and distribution (e.g. CD/DVD burners), started
to become widely available. Yet, as opposed to the Web 2.0 era that ensued, the early
digital era was still characterised by rather passive consumption (e.g. consumer would
subscribe and read online the encyclopaedia Microsoft Encarta), as the tools enabling to
create and distribute digital goods, while available, were rather difficult and tedious to
use.9
As a matter of fact, in the early part of this digital era, it is even rather hard to find
any significant example of prosumption. This began to change in the late 1990s, first
with the advent of peer-to-peer network technologies, and then with the availability of
(legal) digital downloads.
Interestingly, it is a (consumer) user innovation that unleashed the first wave of
digital prosumption. Frustrated by the lack of instant music download (back then,
buying music online meant receiving a CD by the mail a couple of days after you
ordered it) and taking advantage of the newly released ‘MP3’ audio compression
format, three college teenagers, Shawn Fanning, John Fanning and Sean Parker,
developed and released the first largely adopted peer-to-peer network. Napster, which
enabled consumers to easily share and download music files from one another, became
an instant and massive success. Within a matter of months, Napster reached 80 million
users, who arguably set up the first (albeit non-commercial and non-IP compliant)
online music distribution service (Gowan 2012).10
While Napster was shut down by court order in July 2001 after 2 years of operation,
other peer-to-peer sharing protocols (e.g. Kazaa, eMule, BitTorrent), also borne by

9
Publishing a picture online alone would require using a picture editing software (to compress the picture), an
HTML editor to write the code of the webpage on which the picture would be displayed, and an FTP client to
upload the files to a remote web server.
10
As a comparison, 15 years later and despite a heavy worldwide advertising campaign, Apple Music only
reached 11 million users during its initial 3-month free trial (Byford 2015). Spotify, the music streaming
market leader, had as of January 2015 60 million users, including 15 million paying users (https://news.spotify.
com/us/2015/01/12/15-million-subscribers/).
 J Knowl Econ

large user communities, rapidly emerged. Unlike Napster, which was restricted to
music files, these new protocols could be used to share any kind of digital media
(e.g. movies, software, books, pictures) and were far more resilient to court orders.
The rapid emergence and diffusion of the peer-to-peer sharing protocols, carried out
in total independence from the industry (and in spite of it), are hallmarks of the impact
of digital technologies on user innovation. Unlike in the case of pre-digital user
innovation (Table 1), digital technologies enable users to replicate (manufacture) and
distribute digital goods without the help of any firm. Quite strikingly, the first signif-
icant user innovation of the digital era has been to provide tools (peer-to-peer network-
ing protocols) enabling consumers to independently (and in a totally decentralised
manner) replicate and distribute digital goods amongst themselves (whereas, until then,
they had to rely on central servers owned by companies).
A couple of years after peer-to-peer networks emerged, the industry, finally perceiv-
ing the needs for a legal offering of digital content, began to supply music, films, TV
shows, and software over the Internet. This new mode of content delivery made
consumers de facto prosumers. Indeed, unlike before, when the goods (e.g. CDs,
DVDs), even if ordered online, were manufactured and distributed by the firms
(sometimes to the consumer’s very door), online digital delivery (whether download
or streaming) necessarily means that consumers take part in both manufacturing and
distribution. Indeed, their Internet bandwidth is used to ‘transport’ the goods and their
hard drive (or other storage media) to ‘fabricate’ and store the goods.
Table 2 summarises, with regard to both inputs and output, the three new modes of
prosumption enabled by the advent of digital technologies. In particular, in relation to
user innovation—‘Napster (software)’—the difference with pre-digital user innovation
(Table 1) is striking. The (legal) ‘Music Downloads’ example shows new mode of
consumption where prosumption becomes the norm and where consumers play a
significant role in manufacturing and distribution. However, as illustrated by the
‘Napster (usage)’ example, firms are no longer necessary. In the case of peer-to-peer
networks (which are used for both illegal and legal content), prosumers can even play a
leading role and completely bypass traditional manufacturing and distribution channels.
11

Prosumption in the Web 2.0 era

Hence, digital technologies have transformed consumers into prosumers for digital
goods. Yet, in the early years of the digital era, prosumption was mainly restricted to
manufacturing (in the sense of reproduction) and distribution, because the tools
enabling to ‘design’ (create) digital content were still expensive and rather complex
to operate (a legitimately acquired software suite enabling to publish content online
would range from around hundred of euros to a couple of thousands).
This changed significantly in the mid-2000s with the advent of Web 2.0 technolo-
gies. Unlike their forebears, Web 2.0 online services were such that virtually no
additional tool was required to create and publish content online. Whether social media

11
For this particular example, the table displays ‘Minor’ for labour because of the original and repetitive task
of ‘ripping’ (importing in the computer and converting to MP3) CDs and vinyls.
J Knowl Econ

Table 2 Examples of early digital forms of prosumption: prosumers’ input and output

Input Output

Labour Intellectual capital Physical capital Design Manufacturing Distribution

Napster (software) – Leading Leading Leading Leading Leading

Napster (usage) (Minor) – Leading – Leading Leading
Music downloads – – Significant – Significant Significant

(Facebook, Twitter), content sharing platforms (YouTube, Instagram, Flickr), informa-

tion aggregation and sharing (Wikipedia), these platforms ‘moved’ the tools of creation
from the desktop computers to the Web, thereby making them free of access and much
easier to use (complex software could simply not be deployed online, so it had to be
simplified and only retain essential features).
As noted in Rayna and Striukova (2015), Web 2.0 technologies have considerably
lowered the barriers to user creation and, in fact, many Web 2.0 services derive their
value almost exclusively from the content created by users. Yet, while the ‘early digital
era’, discussed in the previous section, was characterised by consumers taking a leading
role in manufacturing and distribution, the ‘modern’ digital era is significantly differ-
ent. Indeed, far from taking advantage of the prosumption potential provided to
consumers by digital technologies, all Web 2.0 services have adopted a central mode
of distribution. Of course, consumers still contribute (they need an Internet connected
device to consume), but this contribution is far smaller than what could have been done
if social media or Wikipedia, for instance, had been hosted on a fully decentralised
consumer network.
While this may seem a fairly minor issue, this is a clear example of a deliberate
decision not to take advantage of prosumption capabilities. As a result of this decision,
consumers in many countries have been experiencing Internet bandwidth issues,
because everyone is pulling digital content from a handful of centralised servers. Alas,
overcoming this problem requires significant investment: When thousands or millions
are watching the same high-definition YouTube video hosted on centralised servers, the
bandwidth needed is simply enormous. Yet, at the same time, all these consumers make
little (if any) use of their upstream bandwidth, which could be used to share with other
consumers the bits of the video that have already been downloaded. This would
certainly be a much more efficient and a far less costly way to distribute videos.
This is something that the British BBC understood perfectly well when it launched
iPlayer, its Internet streaming catch-up television and radio service, late 2007. Daunted
by the enormous investment required to host such a service using centralised servers,
BBC decided instead to develop its own peer-to-peer network as the backbone of
iPlayer. Hence, when watching a replay of a BBC programme, consumers would, at the
same time, share this video with other consumers willing to watch it. Videos were of
course encrypted using a Digital Rights Management (DRM) system, so that users,
even though they were distributing the videos, could not save them on their hard drive
or share them outside of the iPlayer service. Unfortunately, because of the bad
reputation of peer-to-peer networks (irremediably associated with consumers piracy),
and despite not a single case of IP infringement taking place with iPlayer, BBC
 J Knowl Econ

removed this feature late 2008 and moved on to a less efficient (and more costly)
centralised way of distribution.
Table 3 illustrates the respective prosumers’ contributions for these three cases. In
the case of social media, the matter is nonetheless slightly more complex. Indeed, while
consumers do not physically take part in reproduction (manufacturing) and distribution
of content, which is hosted on the companies’ servers, they do play a leading role with
regard to which content is ultimately consumed (and distributed). Indeed, the more a
particular content is ‘liked’ or shared, the more it is viewed and distributed. Hence, one
could consider that, in spite of the centralised hosting, consumers play a leading role,
with regard to manufacturing and distribution on social media.
Comparing Tables 2 and 3, one can notice not only a relative ‘decrease’ in
prosumption related to manufacturing and distribution but also greater levels of
prosumption at the design stage. In fact, the table could have been filed with examples
of online platforms where consumers essentially produce most of the content. Of
course, this does not mean that there are no other forms of prosumption, as the BBC
example shows. Yet, while technically possible, it does not appear that there are, as yet,
examples of (legal) prosumption where consumers are leading with regard to all three
aspects of the output.

Prosumption goes offline: 3D printing and sharing economy

Until recently, and aside from the handful of special cases presented in ‘Prosumption in
the pre-digital era’ section, prosumption was largely thought to be mainly an online
phenomenon. Yet, nowadays, prosumption increasingly also happens ‘offline’.
One of the reasons for that is the advent of digital manufacturing technologies, such
as Additive Manufacturing/3D printing. While these technologies have now existed for
a couple of decades, their very high cost (generally tens or hundreds of thousands of
euros) has kept them out of hands of consumers. However, the cost of 3D printers has
rapidly decreased over the past few years (prices have been cut tenfold in just 5 years).
Nowadays, home 3D printers can be purchased for less than a thousand euros and
professional-grade printers for less than 2500 euros (Rayna and Striukova 2016).
These new manufacturing technologies enable consumers to take a leading role of
both manufacturing and distribution of physical objects, and there are already countless
of examples of prosumption activities that have been enabled by 3D printing technol-
ogies. One of such cases is the Square Helper.

Table 3 Examples of modern digital forms of prosumption: prosumers’ input and output

Input Output

Labour Intellectual capital Physical capital Design Manufacturing Distribution

Wikipedia – Leading Significant Leading Significant Significant

Social media – Leading Significant Leading Significant Significant
(Leading) (Leading)
BBC iPlayer – – Significant – Leading Leading
J Knowl Econ

Chris Milnes, creator of Square Helper, noticed during his ‘consumption activities’
an issue with the Square smartphone bank card reader that an increasing number of
coffee shops and food carts in his neighbourhood were using. When the employees
would attempt to swipe the card, the Square card reader (connected to the jack audio
plug of the smartphone) would spin, preventing the card’s magnetic strip to be read
correctly and even potentially damaging the card (Martin 2013). Frustrated by the extra
waiting time when paying because of this flaw in the card reader design, Chris Milnes
decided to design and build a prototype of a widget that placed between the smartphone
and the reader, which would prevent the latter from swivelling.
This looks, so far, like a typical case of prosumption related to user/consumer
innovation. Where it gets interesting is that the advent of digital manufacturing
technologies has provided this prosumer with manufacturing capabilities. Before the
advent of (cheap) 3D printing, prosumers like Chris Milnes would have had no choice
but to find a company to manufacture his widget (most likely, in the case of Chris
Milnes, he would have contacted the company that makes the Square reader, hopping
for his innovation to be integrated in the product line, which it may have not).
Digital manufacturing technologies considerably increase consumers’ prosumption
capabilities and, instead, Chris Milnes purchased a 3D printer and started to manufac-
ture his widget (which he named ‘Square Helper’) at home (he calculated that he could
build up to 700 U a week), built a simple website with a PayPal shopping cart and
started to sell and distribute the widget himself.12
However, this kind of ‘full-range’ prosumption, where consumers are leading in all
inputs and outputs, does not necessarily require home manufacturing, as the latter may
also be outsourced (just like a firm would). For instance, a consumer (wishing to remain
anonymous, but going by the alias ‘Autoparis and Me’), user of the Autolib’ French car
sharing service, got frustrated by the lack of docking station in the car enabling him to
use his phone as navigator. This consumer built a prototype, specifically designed for
the Autolib’ car (the dock clips to the dashboard vents of the car). Once this was done,
he uploaded the digital blueprint of the dock he designed on two online 3D printing
platforms (Shapeways and Sculpteo), where he outsources the manufacturing and
distribution. When someone purchases the smartphone dock on either platform, the
dock is then 3D printed on demand by the platform, and then packaged and shipped by
the platform to the client.13,14
There are also cases when prosumers outsource the 3D printing manufacturing of
physical products they have designed, but do the distribution themselves (for instance,
because they wish to include a manual or other complementary items). In all cases,
however, when consumers outsource both manufacturing and distribution, this is still
prosumption, because consumers are leading manufacturing and distribution (albeit by
commission), just like a regular producer would.
The impact of these new digital manufacturing technologies is, however, not limited
to the user innovation cases presented above. Indeed, 3D printing technologies also
empower consumer to provide services to other consumers (or even firms). 3D Hubs is
a perfect example of that. This online platform enables people who would like to 3D

12
http://www.squarehelper.com/
13
http://www.shapeways.com/shops/autoparisandme
14
http://www.sculpteo.com/en/s/fgqnt/
 J Knowl Econ

print and do not have a printer to find people around them who have 3D printers and are
willing to print for them for a fee. At this time, over 23,500 ‘providers’ have signed up
with 3D Hubs. While some of these providers are professionals equipped with very
expensive machines, the vast majority of them are individuals (often arts, design and
architecture students, as well as hobbyists) who use the service either to recover the cost
of the printer or to supplement their income. 3D Hubs is, in fact, a direct competitor to
the leading online 3D printing platforms, such as Shapeways, Sculpteo or i.Materialise
(Rayna et al. 2015), and it is a clear example of prosumption where consumers are
leading with regard to manufacturing, but also provide the key input (physical capital).
This last example, where owners of 3D printers ‘share’ (for a fee) their spare
capacity with other users, is, in fact, quite reminiscent of what is generally referred to
as the sharing economy. Also, occasionally called ‘collaborative economy’, this con-
cept encompasses platforms, such as Airbnb,15 Blablacar,16 Uber,17 which enable users
(generally consumers) to rent out something they own or provide services to each other.
As emphasised in Belk (2014) and Hamari et al. (2015), it is the rise of digital
technologies (in particular ICTs) that has enabled the growth of ‘collaborative con-
sumption’. While it is indeed undeniable that digital technologies have enabled new
forms of consumption and services, there is still an on-going debate as to, precisely,
what part of this recent phenomenon is actually new. In particular, while this is certainly
the most widely used term, there has been a lot of criticism of the sharing economy
denomination, which is considered by some as a misnomer, considering that actual
sharing seldom happens. Instead, the vast majority of these services do imply financial
transactions (Eckhardt and Bardhi 2015).
Consequently, there have been attempts to find an appropriate denomination for this
new phenomenon. Eckhardt and Bardhi (2015), for instance, suggested the designation
‘access economy’, because platforms, such as Airbnb, Uber and Blablacar, provide
access to spare capacities, whether of physical goods or people (or both, in the case of
Uber). However, it has been argued, rightfully, that providing ‘access’ to resources is
what businesses have been doing all along and, therefore, access economy is not an
accurate description of the new phenomenon.
Indeed, if one has a thorough look at the various services that are generally
considered as representative of the sharing economy, one cannot help but notice that
there is, indeed, nothing really new about the services offered. Often, there is even
nothing really new about the way there are offered (e.g. holiday accommodations or
taxi rides have been ‘rented’ through intermediaries before). What is really new is by
whom these services are offered: consumers.18
As a matter of fact, a close examination of the key examples of the sharing economy
reveals that they all mostly relate to prosumption. And this is what is actually new:

15
Airbnb is a platform that enables people to let (often sublet, in the case of a flat) a couch, a room in their
home, or even the whole flat).
16
Blablacar is a platform that connects drivers and passengers willing to travel together between cities and
share the cost of the journey.
17
Uber is a mobile app that enables consumers with smartphones to submit a trip request that is then routed to
Uber drivers who use their own car.
18
Uber, in many countries, offers different types of services: taxi rides operated by professionals (e.g., Uber
X) and transportation services offered by non-professional individuals (e.g., Uber Pop). It is the latter that is
generally thought as being part of the ‘sharing economy’.
J Knowl Econ

Digital technologies have enabled to expand offline (i.e. with physical goods) the
opportunities of prosumption. In this respect, Belk (2014) considers the sharing
economy as the offline extension of the online Web 2.0 phenomenon. Consequently,
it would only seem sensible that the phenomenon usually referred to as sharing
economy should in fact be named ‘prosumer economy’, since generalised prosumption
is the defining feature of this new phenomenon.
The various examples of offline prosumption mentioned in this section are
summarised in Table 4. In the case of consumer innovation, such as Square Helper,
consumers have a leading role in all aspects of the output, as well as most of the inputs.
In this respect, it can be noted that digital manufacturing technologies enable to
substitute capital (intellectual or physical) for labour. Unlike in the ‘old fashioned’ user
innovation described in ‘Prosumption in the pre-digital era’ section, user innovators
nowadays seldom need work force in order to innovate (though they may still use it, if
they prefer to).
Platform such as 3D Hubs enable consumers to provide on-demand manufacturing
services, in which case they have a leading role in providing the physical capital (i.e.
the printer) and, of course, in manufacturing (they might occasionally have to supply
intellectual capital to fulfil the service, for instance if the client failed to supply a good
enough digital blueprint).
Finally, the prosumption framework introduced in ‘Towards a prosumption frame-
work’ section can be used to describe contributions to inputs and outputs in the sharing
economy. In that case, manufacturing relates to the production of the service (e.g.
renting a room, a car drive) and distribution relates to the connection between the
supplier of the service and the client. Thus, prosumers who supply services as a part of
the sharing economy play a leading role in providing physical capital (their flat, their
car). In some cases, the service involves labour (e.g. driving people) and might also
involve intellectual capital (e.g. providing an appealing description of a flat for
Airbnb). With regard to the output, consumers have a leading role in delivering the
service (manufacturing), but only a minor role in distribution, since it is the platform
(Airbnb, Uber, etc.) that plays the main role in connecting people.

Overcoming the challenges of leveraging prosumers

Based on the previous section, it is quite clear that digital technologies have turned
prosumption and user innovation from something that was totally under the firms’

Table 4 Examples of physical forms of prosumption: prosumers’ input and output

Input Output

Labour Intellectual capital Physical capital Design Manufacturing Distribution

Square Helper – Leading Leading Leading Leading Leading

3DHubs – (Minor) Leading – Leading –
Blablacar, Uber Pop Leading – Leading – Leading Minor
Airbnb – Minor Leading Minor Leading Minor
 J Knowl Econ

control into a phenomenon that can take place, both online and offline, in total
independence from the firms. In fact, nowadays, digital technologies enable almost
anyone to become an entrepreneur and start competing with existing businesses.
Consumers have therefore become competitors, and whether they end up cooperating
or competing with a firm depends largely on the firm’s strategy.
In such a context, ‘hearing the voice of the consumers’ has turned from something
mainly optional into something fundamental. Because they refused to do so, several
industries have been heavily disrupted by prosumption (whether ‘legal’ or not). The
recording industry, which was pushing for more expensive higher quality physical
formats (e.g. DVD-A, SACD), while most consumers wanted cheaper, lower quality
music directly distributed online, is a good example of that. By not listening to
consumers, the industry triggered a pandemic of illegal prosumption (‘piracy’), which
was only curbed (despite numerous attempts) once adequate offering (e.g. Spotify,
Deezer, Apple Music) had appeared (Aguiar and Waldfogel 2015). Unfortunately for
the recording industry, overlooking (actual or potential) prosumption led to the entry of
new competitors that understood much better the needs of the consumers and took
consumers’ ability to obtain what they needed by themselves as a starting point for their
new business models.19
Likewise, it can be argued that the so-called ‘uberisation’ of the economy is in fact
caused by a lack of understanding of consumer needs and by a gross underestimation of
the prosumption potential. Indeed, it is clear that digital technologies have made,
through prosumption, most markets contestable. Contestable markets (Baumol 1982)
are characterised by a small number of firms that are forced to operate at near-
competitive levels (where profits are close to zero) because of the threat of short-term
entrants. This occurs when (1) market entry and exit costs are very low, (2) there are no
sunk costs and (3) entrants have access to the same technology as incumbents.
Thanks to digital technologies, consumers are often placed in this situation. Unlike a
business, they do not need to invest, as they already own the goods (e.g. car, couch or
flat, 3D printer). They have virtually no cost of entry or exit (it takes just a couple of
minutes to sign up to Blablacar or Airbnb). Finally, they have now access to the same
technology as the ‘big boys’ (Uber, with Uber Pop, provides everyone with central
booking platform, just like professional taxis have; Shapeways enables everyone to 3D
print with machines that cost hundreds of thousands of euros). Because digital tech-
nologies make markets contestable, this renders value capture more challenging, as any
attempt to leverage value to a large extent might trigger the entry of prosumers on the
market.
However, this is not the only issue, as there are also strong implications for user
innovation. In the pre-digital era (and, still, until fairly recently for physical products),
firms were entirely in control of the fate of user innovation. If an innovation suggested
by users did not fit the firm’s strategy, it would simply be disregarded, as it was highly
unlikely that the innovator could do anything independently from the firm (especially
in the case of a consumer). As was shown in the previous sections, digital technologies

19
Early 2015, in a letter to its shareholders, movie/TV show streaming giant Netflix conceded that its main
competitor was in fact Popcorn Time, the peer-to-peer Netflix clone, which in addition to being free of charge
also had a greater catalogue than Netflix. http://files.shareholder.com/downloads/NFLX/3874203383x0
x804108/043a3015-36ec-49b9-907c-27960f1a7e57/Q4_14_Letter_to_shareholders.pdf
J Knowl Econ

have changed everything: Nowadays, consumer innovators are able to release and
diffuse their innovation without any help from the firm.
In the digital realm, a typical example of this is the iOS ‘Jailbreak’ that has existed
almost since the early days of the iOS platform (and despite continuous attempts of
Apple to prevent it) and enables users (generally consumers) to introduce features and
software that Apple did not accept to include in its operating system (but, occasionally,
decided to include at later stage).
Nowadays, similar phenomena do also take place in the ‘physical’ realm. For
instance, in 2013, Fernando Sosa, a fan of the TV show ‘Game of Thrones’, spent a
significant amount of time designing a smartphone charger dock in the shape of the
famous throne of the series. As in the case of the Square Helper discussed in
‘Prosumption goes offline: 3D printing and sharing economy’ section, Fernando Sosa
wanted to diffuse his innovation. After a contact with the right holders’ (HBO) lawyers,
Sosa asked for an official licence and offered to pay HBO a percentage on sales, which
was denied.20 While, as per HBO’s request, Sosa stopped selling the famed dock, the
digital blueprint of a very similar object surfaced very shortly afterwards on sharing
websites and peer-to-peer networks.
Interestingly, this shows that the market is not just contestable not only with regard
to what is produced by the firms but also with regard to all the potential complements
and derivatives that are not. If there is a need that is not fulfilled by the firm (e.g. Game
of Thrones merchandising), consumers will fulfil it themselves, regardless of whether
the firm wants it or not. Whereas before the digital era, a ‘void’ in the market, whether
deliberate or not, would remain void, nowadays, it is more than likely that such a void
will be filled by prosumption.
Yet, the challenges caused by prosumption can be overcome by embracing the
prosumption phenomenon. As mentioned in Rayna and Striukova (2016), Hasbro has
adopted a completely opposite strategy to HBO’s for its ‘My Little Pony’ franchise.
Instead of attempting to curb prosumption, Hasbro decided to encourage it, which
provided the opportunity to control it. Realising that consumer fans would eventually
design, manufacture and attempt to distribute objects inspired by the characters of the
franchise, Hasbro, instead of blocking them, gave them a platform to do so.
Hasbro signed a deal with the online 3D printing platform Shapeways, which
enables fans to upload and sell objects inspired by the My Little Pony franchise. 21
Fans can upload any object inspired by the franchise, for as long as it abides some
common sense rules (and is actually printable), and sell them at a price of their choice.
In return, Hasbro gets a (small) percentage of the sales price and controls what is put for
sell.
The resulting objects are arguably of very good quality. 22 By embracing
prosumption instead of fighting it, Hasbro has created an ecosystem that enables them
to benefit from the creativity and skills of consumers. And, of course, if any particular
of those prosumer objects become highly in demand on Shapeways, nothing prevents
Hasbro from making it a part of its regular product line and mass-manufacture it.
20
http://nuproto.com/iThrone.html
21
Hasbro also offers through Shapeways licences for a couple of other of its franchises, such as DragonVale,
Dungeons&Dragons, GIJoe, Monopoly, Scrabble or Transformers. http://www.shapeways.
com/engage/superfanart
22
See http://www.shapeways.com/superfanart/mylittlepony for examples.
 J Knowl Econ

While these are only a couple of examples, there is a clear trend towards a greater
(potential or actual) prosumption. Yet, in many cases, prosumers still constitute an
untapped resource, although many of them are skilled, educated, very much knowl-
edgeable of their needs and of the needs of others, and also have means (or have access
to means) of production (whether design, manufacturing or distribution). While some
companies have learned to use ‘prosumer power’ to their advantage, most do not.
Because of that, they face the risk that one day, a priori cooperative consumers become
their competitors or that a new entrant that understands how to leverage prosumers
wipes these firms out of their own market.
It is thus critical that firms stop considering consumers as a (somewhat) captive herd,
but, instead, as coopetitors in a joint ecosystem. Involving all the participants in the
ecosystem and ensuring that the benefits of the ecosystem are ‘fairly’ shared is certainly
a challenge that an increasing number of firms will have to face in the near future.

Conclusion

It is clear that digital technologies have had a major impact on prosumption and user
innovation. From a situation where prosumption and user innovation were (generally)
initiated by firms and carried out under their full control, digital technologies have
progressively made prosumption and consumer innovation ubiquitous and
unavoidable.
Yet, prosumption can and does take many forms and each different form comes with
different challenges for the firms. It was therefore necessary to understand precisely the
nature of each form of prosumption. The framework introduced in ‘Towards a
prosumption framework’ section enables to do so, as illustrated by its application to
multiple case studies throughout the article.
One of the insights provided by the use of this framework is that while
digital technologies have enabled consumers to be leading with regard to any
aspect, whether input (labour, intellectual capital, physical capital) or output
(design, manufacturing, distribution), of the production process, the actual form
of prosumption that prevails on the market often has a rather minimal involve-
ment of the prosumer (e.g. music is not distributed on peer-to-peer networks,
prosumer still needs a company like Blablacar to intermediate). In fact, it is
quite likely that in many situations, full-scale prosumption is neither achievable
nor desirable.
Yet, prosumption has become a reality and most successful business models inte-
grate, at least to some extent, some form of prosumption. As discussed in ‘Overcoming
the challenges of leveraging prosumers’ section, regardless of the extent to which it
happens, prosumption is a force to reckon with, as it not only makes most consumer
markets contestable but also enables consumers to fill in gaps left by firms in the
market.
For firms, involving prosumers has not only become an obligation, but it can
also be a significant source of competitive advantage. Nonetheless, there is a
fine line between prosumption benefiting the firm and prosumption undermining
the firm. A line that, furthermore, moves over time! As discussed in ‘Over-
coming the challenges of leveraging prosumers’ section, there is undoubtedly a
J Knowl Econ

shift needed towards ecosystem management, where consumers are no longer

seen as a (relatively) captive heard but as legitimate competitors.
Whether actual or potential, the importance and the effect of prosumption in the
economy have considerably increased over the past decades. With prosumption
reaching the realm of physical objects and services, one could argue that the industry
disruptions that have been observed online are about to reach the entire economy.
Hence, the question is: are we moving away from a ‘digital economy’ towards a
prosumer economy?
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International
License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and repro-
duction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a
link to the Creative Commons license, and indicate if changes were made.

References

Aguiar, L., & Waldfogel, J. (2015). Streaming reaches flood stage: does Spotify stimulate or depress music
sales? Institute for Prospective Technological Studies Digital Economy Working Paper 2015/05,
European Commission Joint Research Centre.
Baumol, W. J. (1982). Contestable markets: an uprising in the theory of industry structure. American
Economic Review, 72(1), 1–15.
Belk, R. (2014). You are what you can access: sharing and collaborative consumption online. Journal of
Business Research, 67(8), 1595–1600.
Byford, S. (2015). Apple Music has a third as many paying subscribers as Spotify. The Verge. Published 20
October 2015 (Retrieved 28 October 2015). http://www.theverge.com/2015/10/20/9572667/apple-music-
subscriber-count-number-paid.
Eckhardt, G. M. and Bardhi, F. (2015). The sharing economy isn’t about sharing at all. Harvard Business
Review. Published 28 January 2015 (Retrieved 29 October 2015). https://hbr.org/2015/01/the-sharing-
economy-isnt-about-sharing-at-all.
Galbraith, J. K. (1958). The Affluent Society. Houghton Mifflin.
Galbraith, J. K. (1967). The New Industrial State. Princeton University Press.
Gowan, M. (2012). Requiem for Napster. PC World. Published 18 May 2002 (Retrieved 28 October 2015).
http://www.pcworld.idg.com.au/article/22380/requiem_napster/.
Hamari, J., Sjöklint, M., & Ukkonen, A. (2015). The sharing economy: why people participate in collaborative
consumption. Journal of the Association for Information Science and Technology. doi:10.1002/asi.23552.
Kotler, P. (1986). The prosumer movement: a new challenge for marketers. Advances in Consumer Research,
13(1), 510–513.
Lüthje, C., Herstatt, C., & Von Hippel, E. (2005). User-innovators and Blocal^ information: the case of
mountain biking. Research Policy, 34(6), 951–965.
Martin, S. (2013). Square Helper: designed and mass manufactured in a home office with a MakerBot.
SolidSmack. Published 18 January 2013 (Retrieved 29 October 2015). http://www.solidsmack.
com/culture/square-helper-designed-and-manufactured-in-a-home-office-with-a-makerbot/.
Matsuyama, K. (2002). The rise of mass consumption societies. Journal of Political Economy, 110(5), 1035–
1070.
Rayna, T. (2008). Understanding the challenges of the digital economy: the nature of digital goods.
Communications & Strategies, 71, 13–26.
Rayna, T., & Striukova, L. (2010). Large-scale open innovation: open source vs. patent pools. International
Journal of Technology Management, 52(3/4), 477–496.
Rayna, T., & Striukova, L. (2015). Open innovation 2.0. Is co-creation the ultimate challenge? International
Journal of Technology Management, 69(1), 38–53.
Rayna, T., & Striukova, L. (2016). From rapid prototyping to home fabrication: how 3D Printing is changing
business model innovation. Technological Forecasting and Social Change, 102, 214–224.
Rayna, T., Striukova, L., & Darlington, J. (2015). Co-creation and user innovation: the role of online 3D
printing platforms. Journal of Engineering and Technology Management, 37, 90–102.
 J Knowl Econ

Ritzer, G., & Jurgenson, N. (2010). Production, consumption, prosumption: the nature of capitalism in the age
of the digital ‘prosumer’. Journal of Consumer Culture, 10(1), 13–36.
Tapscott, D., & Williams, A. D. (2006). Wikinomics: how mass collaboration changes everything. New York:
Portfolio.
Toffler, A. (1981). The third wave. New York: Bantam Books.
von Hippel, E. (1976). The dominant role of users in the scientific instrument innovation process. Research
Policy, 5(3), 212–239.
von Hippel, E. (1977). Transferring process equipment innovations from user-innovators to equipment
manufacturing firms. R&D Management, 8(1), 13–22.
von Hippel, E. (1978). A customer-active paradigm for industrial product idea generation. Research Policy,
7(3), 240–266.
von Hippel, E. (1986). Lead users: a source of novel product concepts. Management Science, 32(7), 791–805.
Zwass, V. (2010). Co-creation: toward a taxonomy and an integrated research perspective. International
Journal of Electronic Commerce, 15(1), 11–48.
 Technological Forecasting & Social Change 160 (2020) 120248

Contents lists available at ScienceDirect

Technological Forecasting & Social Change

journal homepage: www.elsevier.com/locate/techfore

Balancing innovation and exploitation in the fourth industrial revolution: T

Role of intellectual capital and technology absorptive capacity
⁎
Tarique Mahmooda, , Muhammad Shujaat Mubarikb
a
Mohammad Ali Jinnah University, Karachi, Pakistan
b
Faculty of Business Administration & Social Sciences, Mohammad Ali Jinnah University, Karachi, Pakistan

A R T I C LE I N FO A B S T R A C T

Keywords: Industry 4.0, which features the Internet of things (IoT), cloud computing, big-data, digitalization, and cyber-
Fourth industrial revolution (Industry 4.0) physical systems, is transforming the way businesses are being run. It is making the business processes more
Innovation autonomous, automated and intelligent, and is transmuting the organizational structures of businesses by di-
Technology absorptive capacity gitalizing their end-to-end business processes. In this context, balancing innovation and ex-
Intellectual capital
ploitation—organization's ambidexterity—while stepping into the fourth industrial revolution can be critical for
Organizational ambidexterity
Structural equation modeling
organizational capability. This study examines the role of intellectual capital (IC)—human capital, structural
capital and relational capital—in balancing the innovation and exploitation activities. It also examines the role of
technology's absorptive capacity in the relationship between IC and organizational ambidexterity (OA). Data
were collected from 217 small and medium enterprises from the manufacturing sector of Pakistan using a closed-
ended Likert scale-based questionnaire. The study employs partial least square-Structural Equation Modeling
(PLS-SEM) for data analysis. Findings indicate a profound influence of all dimensions of IC, both overall and by
dimensions on organizations’ ambidexterity. Findings also exhibit a significant partial mediating role of tech-
nology absorptive capacity (TAC) in the association of IC and ambidexterity. The findings of the study emphasize
the creation of specific policies aimed to develop IC of a firm, which in turn can enable a firm to maintain a
balance between innovation and market exploitation activities. The study integrates the TAC with the IC-OA
relationship, which is the novelty of the study.

1. Introduction old technologies, processes, and interactions with new ones. Further, in
the presence of staggering effects of COVID-19, the need to transform
The world is on the verge of a fourth industrial revolution (Industry businesses according to the paradigm of Industry 4.0 is ever-increasing
4.0), which is ready to transmute the way businesses exploit the mar- as the firms not only have to maintain a balance between their in-
kets, innovate and adopt technologies. The pace and momentum of this novation and market exploitation activities but also need to enhance
revolution is unmatched, turbulent and exponential. Both practitioners their technological absorption capacity (TAC) to cope with the chal-
and researchers envision that Industry 4.0 can enable firms to gain ef- lenges posed by the this pandemic. There is mounting pressure on al-
ficiency and faster innovation. It is worth mentioning that this re- most every industry across the globe (Vlačić et al., 2019; Kafouros et al.,
volution is transforming the global economic structure and is in- 2020). Researchers claim that the impacts of the fourth industrial re-
eluctable for any single country or organization (EEF- volution are not only limited to large businesses but can also have a
The Manufacturers’ Organization, 2016). In such situation, businesses profound influence on the small and medium enterprises (SMEs).
are required to upgrade their technological and innovative capabilities Hence, present day businesses, irrespective of their size and nature, are
according to the Industry 4.0 needs for competing in the dynamic en- entering a turbulent, challenging and dynamic business environment
vironment (Horváth, and Szabó, 2019; Frank et al., 2019). It is also (Ashton and Morton, 2005; Hitt et al., 1998). In this situation, the
essentially revolutionizing the way organizations interact with their survival and growth of a business greatly depends upon the way it
customers, employees and suppliers. It is worth mentioning that In- balances innovation and market exploitation—organizational ambi-
dustry 4.0 requires businesses to transform themselves by replacing the dexterity (OA)—and the extent to which it can absorb the technology-

This article belongs to the special section on Technological Innovations & Their Financial & Socio-economic Implications in the Era of Fourth Industrial Revolution.
⁎
Corresponding author e-mail: [email protected]
E-mail address: [email protected] (T. Mahmood).

https://doi.org/10.1016/j.techfore.2020.120248
Received 21 June 2020; Received in revised form 6 August 2020; Accepted 7 August 2020
0040-1625/ © 2020 Elsevier Inc. All rights reserved.
T. Mahmood and M.S. Mubarik Technological Forecasting & Social Change 160 (2020) 120248

related knowledge—technology absorptive capacity (TAC)—from the ex- (Mubarik et al., 2019; Mubarik, 2015). These IC dimensions will in-
ternal environment. TAC provides firms with the capability to adapt teract across different organizational levels, playing distinct roles when
and evolve in the era of Industry 4.0. It allows firms to sustain a creating ambidextrous capabilities (Fernández-Pérez et al., 2017;
competitive advantage by means of organizational innovation in the Kang and Snell, 2009) and leading to superior organizational perfor-
context of a dynamic industry. TAC can also foster the capacity to ab- mance (Andrews, 2010; Liu and Chen, 2009). Before discussing how the
sorb new technological knowledge—knowledge that is valuable be- IC interacts with OA and the role TMT plays in IC-ambidexterity dyad,
cause it leads to further organizational innovation. Researchers it is essential to review the various definitions of IC and how it is
(Ahmed et al., 2019;Mubarik, 2015) claim that the fourth industrial measures. To this effect, the following section briefly reviews the de-
revolution is profoundly changing the basis of competitive advantage finition of IC and its way of operationalization.
from tangible resources to intangible resources (Grant, 1996; Hitt et al.,
1998; Wong and Aspinwall, 2005; Secundo et al., al.,2020). In such 2.1. Intellectual capital
case, a firm's intellectual capital (IC) can play a fundamental role in
attaining OA, technology absorption and competitive advantage. The The fourth Industrial Revolution creates new relationships between
importance of IC can be traced back to the resource-based researchers humans and machines, along with changes in the work characteristics,
mentioned earlier (e.g. Penrose, 1959; Becker, 1962; Wernerfelt, 1984) organizational structure and relationships. This has brought IC into the
who pointed out that a business's external environment and its level of heart of fourth industrial revolution and is the reason that researchers
intangible resources determine its level of success. The resource-based claim that stepping into the fourth industrial revolution requires having
view (Wernerfelt 1984) claimed that competitive advantage of cor- strong IC. The concept of IC primarily incorporates activities of em-
poration are rooted through organizations that have VRIN resources ployees, directors, intelligent human beings and stakeholders of the
(valuable, rare, inimitable and non-substitutable) (Kor and company that create value. John Kenneth Galbraith (Itami, and Roehl.,
Mahoney, 2004). VRIN can be acquired through human capital (HC), 1991) first coined the term “intellectual capital” or IC in 1969. Ac-
relationships and processes collectively termed as IC ( Lev and Zambon cording to Edvinsson (1997) and Jørgensen and Boje (2010), the term
2003; Chu et al., 2006; Mubarik et al., 2019). By developing IC, a firm includes a degree of relative 'intellectual action'. Researchers mention
can attain OA—a fit between its firm's innovation and exploitation activities IC as the knowledge and relationships that can be transformed into
(Barney, 2001; ; Russo and Fouts, 1997). In an IC-ambidexterity dyad, a organizational performance. However the framing of IC as a subject of
firm's capacity to absorb outside technology related knowledge can play research is a fairly new trend. It started from after the seminal article of
an instrumental role (Mubarik 2015; Antonelli and Colombelli, 2011). Stewart (1997) published in Fortune magazine where he discussed IC as
A number of studies (e.g. Ashton and Morton 2005; Barathi 2007; being the knowledge, ability, and strength of employees, which could
Arenas, and Lavanderos, 2008; Hsu and Fang 2009; Asiaei, and Jusoh, strengthen the competitiveness of an organization. He also mentioned
2015) examine the role of OA in the association between IC and firm that the distinction between the market value and the book value of an
performance. However the scholastics work examining the role of TAC organization is IC. Various researchers (e.g. Barathi Kamath, 2007;
in IC- OA is missing (Mubarik et al., 2019 Carte, 2005; Mouritsen, 1998, Bontis and Fitz-Enz, 2002; Bontis and Nikitopoulos, 2001; Kang and
2005; Seleim and Khalil, 2011; Alvarez and Barney, 2001; Nonaka and Snell, 2009) from distinctive backgrounds explain the concept of IC
Konno, 1998). Against this backdrop, examining how TAC intervenes in differently; however three components of IC can be found in almost
the interplay of ICeOA is essential for devising strategies for IC led every definition: HC, RC and SC. The following lines briefly explain
ambidexterity. This study undertakes this task, and in doing so, it each of these components.
contributes to the literature in three ways. First, it integrates the dis- Human Capital (HC): Defined as the knowledge, skills, and cap-
persed variables of IC, ambidexterity and absorptive capacity in the abilities of an individual the notion of HC can be traced back to the
framework, and provides empirical evidence on this triad. Second, it early 1950s (Mubarik et al., 2018). However, the inception of HC
provides application of IC-ambidexterity models in the context of SMEs theory (Becker, 1962a) and theory of firm (Penrose, 1959) exposed the
as the majority of the studies conducted on this conception focused on importance of HC both at firm and country levels. These researchers
the larger firms. Third, for the first time, the study introduces and ex- defined HC as the knowledge, skills, and abilities of a person that can be
amines the role of top management team in IC, ambidexterity and ab- instrumental in improving his job-related performance. Becker (1964)
sorptive capacity triad. explained that improving HC through education and training can in
The paper has been divided into five sections. The following section turn improve an organization's performance. He also considered health
briefly discusses the theoretical and empirical literature on the asso- as an essential component of HC. Scholars in the early 1990s, (e.g.
ciation of IC, OA and absorptive capacity interplay. Section 3 entails the Grant, 1996; Russo and Fouts, 1997; Edvinsson, 1997; Mouritsen, 1998;
methodology employed for analysis whereas Section 4 discusses the Hitt et al., 1998) included qualities like attitude, creative thinking, and
findings of the study. Section 5, the last section, concludes the study. problem solving as being part of HC, and they developed several con-
structs to measure it at the firm level. Likewise, HC of an organization is
2. Review of literature also defined as the combined competencies of employees to resolve
problems of customers, suppliers, and the organization. The organiza-
Organizations in the modern era are existing in a complex en- tion-wide HC is the knowledge and institutional memory about prior-
vironment of ever-increasing dynamism and uncertainty. For the suc- itizing the importance of the organizational issues. This resource com-
cess of an organization, it is vital for it to develop and acquire tacit prises the individual skills, collective experience, general know-how
resources and knowledge (Hitt et al., 2001; Hitt et al., 1998), leading to and management expertise of all of the employees (Edvinsson, 1997).
a shift from physical resources to intangible (imitable) resources, and Gupta and Roos (2001) determine that HC includes skills and knowl-
for gaining both ambidexterity and competitiveness (Ashton and edge, so intellectual ability in employees work for quickly adopt
Morton, 2005; Wong and Aspinwall, 2005). Thus, management of change, innovation and effective solution to the problem etc. Various
knowledge-based resources has become a key driver for better perfor- studies (Arshad et al., 2015; Hershberg, 1996; Mubarik et al., 2018)
mance and sustainable competitive advantage (Grant, 1996; considered HC as being the repository of knowledge, abilities and skills
Mubarik et al., 2019; Sharkie, 2003; Teece, 2012). The literature on IC exemplified in labor for the production and economic value. Thus, HC
contains attention-grabbing opinions about the technique in which or- can be defined as the knowledge and skills achieved by the worker
ganizations adjust exploration and exploitation activities. In the same through experience and education (Sullivan and Sullivan, 2000). Diaz-
vein, HC, structural capital (SC) and relational capital (RC) represent Fernandez et al. (2017) supported this definition and pointed out that
different knowledge stocks at different organizational levels HC is comprised of personal attributes such as skills, knowledge, and

2
T. Mahmood and M.S. Mubarik Technological Forecasting & Social Change 160 (2020) 120248

Fig. 1. IC taxonomy.

Fig. 2. Conceptual framework.

expertise. It is HC that plays a substantial part in organizational per- stakeholders, and customers. In this regard, relational or social capital
formance as well as in the economic development of any country. is considered as the bond that is formed with organizational relation-
Drawing from the HC theory and resource-based theory, we define HC ships with stakeholders, and that impacts the lives of the organization.
as the knowledge, skills and abilities of the employees that are instru- Researchers (e.g., Eisenhardt and Sull, 2001;Kang et al., 2007;
mental in increasing organizational performance (Figs. 1 and 2). Mom et al., 2015; Mubarik et al., 2016) explain relational or social
Structural capital (SC): From the perspective of the organization, SC capital as being a blend of different relationships such as market re-
incorporates all non-human knowledge resources. SC represents an lationships, power relationships and cooperation. RC encompasses
organization's processes and structures through which it performs its trust, stronger understanding, collaboration and the relationship among
business transactions. These structures range from tangible to in- the strategic partners in this regard, and it characterizes interactions,
tangible stuff that the organization offers, for instance, copyrights, connection stocks, closeness, linkages, loyalty, and goodwill between
patents, software systems, databases, processes and trademarks, ac- an organization and downstream clients, upstream suppliers, external
countability, organizational culture, trust among employees and effi- stakeholders, and strategic partners (Lazzarotti et al., 2017; Mubarik
ciency ( Zameer et al., 2020). Structural or organizational capital ac- et al., 2019; Naghavi and Mubarak, 2019). It is also called external
cordingly comprises of internal value drivers of an organization, capital that comprises customers, brands, and the reputation of the
namely, routines, processes, customer files, database, manuals, and the company business collaboration; and channels of distribution, licensing
literature and structure of an organization (Reza et al., 2020)(). agreements and customer satisfaction.
Asiaei and Jusoh (2015) state that organizational capital comprises of
internal capital, which encompasses management philosophy, in- 2.2. Organizational ambidexterity
tellectual property, management processes, financial relations, in-
formation and networking systems, and corporate culture. Although OA was coined back in 1976 by Duncan, 1976 ) , the
Relational capital (RC): RC is characterized by the organization's concept gained heightened attention after the popularity of the fourth
reputation and the loyalty of customers. All of these resources are industrial revolution. Industry 4.0 requires a balanced yet dynamic
connected to the organization's external relations with its suppliers, approach to simultaneously drive the innovation and exploitation

3
T. Mahmood and M.S. Mubarik Technological Forecasting & Social Change 160 (2020) 120248

activities, which cannot be attained without having OA. This notion of exploitation, they make great use of its present organization resources
an ambidextrous organization was substantiated by March, 1991), when giving an explanation. Hence, organizations have to bring about
suggesting exploitation (market capitalization) and exploration (in- the right mix of both exploration and exploitation. It implies that all
novation) as the two learning activities carried out by organizations. three components of IC play a fundamental role in the adaptation of
Exploration entails innovation, experimentation, search, and discovery ambidexterity.
whereas exploitation, in contrast, is linked with efficiency, selection, First HC—operationalized as education, skills and capabilities of
refinement and implementation activities. Exploration and exploitation, employees—plays a significant role in promoting both innovation and
therefore, essentially require different strategies, organization struc- productivity. It promotes ambitexterous capacity as employees possess
tures and contexts. Various scholars agree that an organization faces the creativity and competence that is essential to refine existing
trade-off between properly exploiting existing competencies and ex- knowledge and develop new knowledge (Kang and Snell, 2009; ). It is
ploring new opportunities by aligning their functions (García- precisely through such HC that ambidexterity is able to access and utilize
Morales et al., 2007;Alänge and Steiber, 2018; Baškarada et al., 2016; knowledge from multiple domains, discover novel solutions to existing pro-
Junni et al., 2013; Mubarik et al., 2019). blems, and challenge assumptions behind prevailing knowledge and prac-
The most prevalent ambidextrous organizations are categorized as tices. Effectively managed HC contributes to the incremental and radical
contextual ambidexterity, innovative ambidexterity, structural ambi- innovations (Mubarik et al., 2016; Raisch et al., 2009).
dexterity, and sequential ambidexterity. Asiaei et al. (2018) argue that experience, education, skills and training
Contextual Ambidexterity: This denotes the behavioural and inter- are important dimensions of HC and they contribute to the ambidex-
active capacity to simultaneously determine adaptability and alignment trous behaviours. The higher levels of experience and knowledge enable
(Gibson and Birkinshaw, 2004). Meanwhile, the alignment includes the employees to resolve problems in a quick and effective matter. Like-
consistency among the activities of organization units, while adapt- wise, these HC constructs also enables employees to work smartly and
ability is related to the ability for quickly configuring the activities in effectively, thus increasing their ambidextrous characteristics
connection with variations in the task environment. In contrast, other (Mubarik et al., 2019). Considering these arguments, the author pro-
concepts of OA state that contextual ambidexterity is based on the in- poses that:
stantaneous pursuit of paradoxical agendas within the single unit of an
H1: Human capital (HC) has a positive influence on organizational
organization. Innovative ambidexterity can be considered as an ambi-
ambidexterity (OA).
dextrous result variable that captures the exploitation and exploration
performance of an organization (Simsek et al., 2009). Innovative am- Likewise, the RC, the second most important dimension of IC, in-
bidexterity indicates the firm's ability to instantaneously follow both fluences the firm's ability to manage exploitation and exploration si-
discontinuous [exploratory] and incremental [exploitative] innovation. multaneously. According to researchers (Mubarik et al., 2019, 2016),
Structural ambidexterity: Is associated with essential trade-off in- improving relationship with suppliers and customers helps the organi-
novative ambidexterity such as exploitative vs. exploratory innovation, zation not only in market exploitation but also in developing the new
it is not determined with one organization unit. Structural ambi- market trends. Consequently, relational systems will facilitate interac-
dexterity grounds on spatial separation of organizational units which tions between humans and processes to help them work together,
are furnished with paradoxical activities. (Jansen et al., 2012; creating ambidexterity. Further, proper social perspective will help the
Volberda and Lewin, 2003), in this regard, referred to enabling me- firm to overcome the tensions between exploitative and exploratory
chanism of structural differentiation as segmenting organization system activities (Lavie et al., 2010; Stettner and Lavie, 2014).
into subunits, each tends to establish specific attributes in line with Relational/social capital also establishes itself in the associated ties
requirements of the external environment and acknowledge this is a that ambidexterity can develop with external parties (e.g., alliance
possible avenue to organizations ambidexterity. partners, suppliers, customers and consumers). Connecting such ties
Sequential ambidexterity: In contrast to structural ambidexterity, se- enables ambidexterity to both improve and renew their knowledge base
quential ambidexterity involves a vibrant perception on an either ex- by having superior potential control and access to over a various range
ploratory or exploitative decision and is gained through the mechanism of perspectives, specialized knowledge and skills (Mom et al., 2015;
of temporal separation. Thus, sequential ambidexterity rises from dy- Tiwana, 2008). Accordingly, we theorize that a relational/social capital
namic, temporal sequencing of exploration and exploitation. This se- will promote the pursuit of ambidextrous learning activities by devel-
quential ambidexterity perception is supported by Good and oping a perspective that supports access to and combination of diverse
Michel, 2013; Güttel and Konlechner, 2009; Junni et al., 2013). The skills and knowledge both within and across the confines of ambi-
sequential ambidexterity is time paced sequence of exploitation and dexterity. Along these lines, the author proposes that:
exploration and stated that this is consistent with dynamic capability
H2: Relational capital (RC) has a positive influence on organiza-
view.
tional ambidexterity (OA).
2.3. Association between IC and organizational ambidexterity There is no ambidextrous organizations is based on individual
learning. No organization can grow unless the individuals of that or-
Researchers (Asiaei et al., 2018; Bontis and Nikitopoulos, 2001; ganization are well-skilled, but this is an inadequate condition for
Khalique et al., 2015; Mubarik et al., 2019) mention IC as being the key ambidextrous organizations. Structural/organizational capital is the
capability needed to acquire OA—a fit between exploitation and ex- knowledge that resides in organization system, patents, databases,
ploration. Mubarik et al. (2019) argue that organizational-level ambi- processes, and structures (Youndt et al., 2004). Hence, rather than
dexterity can only be adopted by improving the ICs (firms processes, different firm's possessing, structural/organizational capital to develop
HC, and relationships). They further explain that transforming the IC on exploration and exploitation as two disconnected blocks, a versatile and
the lines of ambidexterity may enable firms to effectively maintain a flexible structural/organizational capital is considered to attain OA.
balance between exploration and exploitation activities. Likewise, Crossan et al. (1999) suggests that structural/organizational capital is
Bontis (1998) also mentions the need for the IC in maximizing and understood as the set of rules, structures, routines and standardized
balancing both exploitation and exploration activities of a firm. When processes followed by the organization to help build an organizational
the organizations explore and exploit simultaneously, it is called the culture with a similar frame of reference for all employees. It is ex-
organization ambidextrous learning. Because the organization together ploitation in nature, and employees in the organization intend to solve
use exploration, that is, identifying and learning new knowledge along their problems based on the decisions that were previously proved
with their current knowledge to make of an organization and useful ( Chang et al., al.,2020). On the other hand, for Eisenhardt and

4
T. Mahmood and M.S. Mubarik Technological Forecasting & Social Change 160 (2020) 120248

Table 1 external technologies and technological knowledge, and then diffusing,

Sampling distribution. assimilating, communicating and absorbing these into their organiza-
S. no Industry Percentage Number tions. Identifying, transforming, acquiring and exploiting technological
knowhow can only be done based on the knowledge and learning
1 Textile 33% 72 processes that have already been mastered by the firm. As mentioned
2 Leather 28% 61
by Bontis (2019) IC is an essential mainstay of TAC. Ahmed et al.
3 Sports 16% 34
4 Food 14% 31
(2019) further argue that IC can play an instrumental role in improving
5 Metal 9% 19 the TAC of a firm. They claim that all three dimensions of IC can help a
Total 217 firm to interact with external environment and to identify, acquire and
assimilate the external knowledge related to the technology. A firm
with a strong TAC can have better capabilities to attain OA, whereas
Sull (2001), the notion of structural/organizational capital is less con- deficient TAC acts an obstacles for the transfer of better practices,
nected to rules, procedures and common style of work (Daft and technological knowledge, and OA. In short, enabling TAC can be in-
Weick, 1984) and provides new opportunities and autonomy for in- strumental in attaining OA.
dividual and group work in order to explore and experiment with the
new styles of work environment and the way in which they unify the H4: Technology absorptive capacity (TAC) mediates the relationship
work. Drawing on the aforementioned discussion, the last proposition between human capital (HC) and organizational ambidexterity
is: (OA).

H3: Structural capital (SC) has a positive influence on organizational H5: Technology absorptive capacity (TAC) mediates the relationship
ambidexterity (OA). between relational/social capital and organizational ambidexterity
(OA).
H6: Technology absorptive capacity (TAC) mediates the relationship
2.4. Technology absorptive capacity, IC and organizational ambidexterity
between structural/organizational capital and organizational am-
bidexterity (OA)
TAC is considered as being an essential organizational capability in
order to embrace the fourth industrial revolution. It appears to have a
profound influence on the activities related to technology and innova-
3. Methodology
tion in organizations. It is closely linked with OA and is considered to be
the prime enabler of OA (Barney, 1991; Prahalad and Hamel, 2007).
3.1. Population & sampling
Researchers (e.g. Lund Vinding, 2006; Zahra and George, 2002) argue
that organizations need to improve their capacity to absorb outside
For the study, the data was collected from the manufacturing sector
technological knowledge—TAC—for improving their ambidexterity
SMEs (Small Medium Enterprises) of Pakistan. According to the defi-
(Lund Vinding, 2006; Zahra and George, 2002). Vinding (2006) defines
nition provided by the SMEDA (2018), organizations with the staff
absorptive capacity as “the firm's ability to identify, assimilate and exploit
between 10 and 100 are small, and between 101 and 250 are cate-
knowledge from the environment”. Drawing from the premise of HC
gorized as medium organizations. We approached 450 SMEs with the
theory, resource-based view and IC theory, when IC can directly con-
help of Small and Medium Enterprises Development Authority
tribute to a firm's ambidexterity, it has the potential to improve the firm
(SMEDA) of Pakistan. A total of 221 SMEs responded to our ques-
absorptive capacity (Mubarik, 2015). Absorptive capacity has two
tionnaire by providing the required data. After screening 04 cases were
major types: First, potential absorptive capacity, which represents ac-
removed from the data because of unengaged responses and missing
quisition and assimilation of knowledge outside the firm. Second, rea-
values. Hence 217 cases were processed for analysis. Table 1 shows the
lized absorptive capacity, which includes assimilation and transforma-
industry-wide distribution of the final sample used in the study.
tion of knowledge outside the firm (Lund Vinding, 2006). García-
Morales et al., 2007, p.531) defines TAC as the process that “involves
acquisition (through which the firms obtains so called technological 3.2. Measurement
stock), assimilation and transformation (the capacity to develop and
refine routines to facilitate combining existing technological knowledge To collect the primary data from a large number of respondents, the
with that acquired, and to assimilate this knowledge and to exploit questionnaire proves to be a reliable tool. We developed a close-ended
technological knowledge)”. TAC helps a firm to upgrade, expand and questionnaire to collect the data. Table 2 exhibits the detail of the
utilize existing capabilities and technologies to innovate, incorporating constructs with sources. The questionnaire employed for the study had
the technological knowledge acquired and transforming the firm's op- two portions; first part containing the demographic information of the
erations to increase the productivity of the goods and capital employed. respondent organization and the second part containing the items re-
Firms can reinforce their technological competences by importing lated to our conceptual framework. All the constructs were measured

Table 2
Constructs and their sources.
Construct Sub-constructs Items Sources

Intellectual Capital Green Human Capital 12 Items Becker, 1962a, 1962b; Hershberg, 1996; Mubarik et al., 2018;Devadason, 2016
Green Structural 09 Items Burt, 2017;Kamall Khan, 2016)
capital
Green Relational 09 Items Lazzarotti et al., 2017; Lopes-Costa and Munoz-Canavate, 2015; Mom et al., 2015; Mubarik et al., 2019;
Capital Devadason, 2016
Technology Absorptive Capacity Realized 03 Items (Andrawina, 2009, 2008; Lund Vinding, 2006; Zahra and George, 2002)
Potential 03 Items (Andrawina et al., 2008; Lund Vinding, 2006; Zahra and George, 2002)
Organizational Ambidexterity Green Exploration 04 Items (Birkinshaw and Gupta, 2013; Fernández-Pérez de la Lastra et al., 2017b, 2017a, 2017c; Fu et al., 2016;
Mubarik et al., 2019; Raisch et al., 2009)
Green Exploitation 04 Items (Andriopoulos and Lewis, 2009; ; Lavie et al., 2010; Mubarik et al., 2019; Stettner and Lavie, 2014)

5
T. Mahmood and M.S. Mubarik Technological Forecasting & Social Change 160 (2020) 120248

on five points Likert scale where 1 (Strongly Disagree) 2 (Disagree) 3 Table 3

(Neutral) 4 (Agree) 5 (Strongly Agree). The description of the in- Reliability and validity.
dividual constructs is given below. Construct Item Loadings AVE CR CB Alpha
Human capital (HC) is defined as the knowledge skills and abilities
exploit by individual employees. This resource comprises education, Intellectual Capital Human HC1 0.68 0.61 0.87 0.91
Capital HC2 0.75
experience, training, and skills.
HC3 0.71
Relational capital (RC) is defined as the knowledge embedded with HC4 0.81
networks of interrelationships and their interactions among individuals. HC5 0.91
This resource includes strategic alliances, customer and supplier rela- HC6 0.72
tions, and customer knowledge. HC7 0.82
HC8 0.91
Structural capital (SC) is the institutional Knowledge utilized
HC9 0.69
through patents, databases, structures, processes, and systems. This HC10 0.71
resource encompasses the System and program, research and develop- HC11 0.72
ment, intellectual property rights (IPRs). HC12 0.73
Structural SC1 0.79 0.59 0.79 0.87
Organizational Ambidexterity (OA) has been operationalized as ex-
Capital SC2 0.72
ploration and exploitation are two matchless learning activities that SC3 0.78
organizations split their resources and attention. This resource en- SC4 0.81
compasses Exploration mentions to variation, experimentation, search, SC5 0.7
and discovery; In contrast, exploitation is linked with efficiency, se- SC6 0.72
SC7 0.79
lection, refinement and implementation activities. Exploration and ex-
SC8 0.81
ploitation, therefore, require radically different strategies, organization SC9 0.77
structures and contexts. Relational RC1 0.75 0.58 0.83 0.89
Technology Absorptive Capacity (AC), is defined as “the firm's ability Capital RC2 0.73
to identify, assimilate and exploit the technology related knowledge RC3 0.77
RC4 0.76
from the external environment”. This resource encompasses the RC5 0.79
Supplier development department, Face-to-face meetings with supplier, RC6 0.71
Intranet with knowledge management system and Cross-functional RC7 0.77
meetings. RC8 0.79
RC9 0.88
Technological AC1 0.81 0.59 0.81 0.91
3.3. Analytical method: PLS-SEM Absorptive AC2 0.82
capacity AC3 0.77
AC4 0.72
The study will apply PLS-structural equation modeling for testing AC5 0.71
the framework. PLS-SEM has been preferred because of its ability to AC6 0.81
model multiple relationships simultaneously. Similarly, the use of this AC7 0.73
technique was opted due to its ability to control the endogeneity pro- AC8 0.87
AC9 0.71
blem. PLS is also a preferred technique to model relationships when one
AC10 0.72
variable is dependent and independent at the same time. This technique AC11 0.81
well caters the problem of non-normality of data and is considered AC12 0.75
highly robust against non-normal data. AC13 0.72
AC14 0.73
Organizational Exploration OA1 0.7 0.57 0.86 0.94
4. Results and discussions Ambidexterity OA2 0.76
OA3 0.77
OA4 0.87
4.1. Demography of the respondents
Exploitation OA5 0.81 0.58 0.81 0.89
OA6 0.72
Table 3 shows the brief demography of the respondents’ organiza- OA7 0.83
tions. The unit of analysis of the study is organization. A total of 450 OA8 0.74
organizations were approached for data collection whereas 217 orga-
al., 2011; Hulland, 1999; Latan and Ghozali, 2012a)
nizations responded positively. It shows an average response rate of
b. Average Variance Extracted (AVE) > 0.5 indicates Convergent Validity
(Bagozzi and Yi, 1988; Fornell and Larcker, 1981)
Table 3
c. Composite Reliability (CR) > 0.7 indicates Internal Consistency (Gefen et al.,
Respondents demography.
2000)
Percentage Number (n = 217) d. CB alpha > 0.7 indicates indicator reliability (Dijkstra and Henseler, 2015).

Industry
48%. Among respondent firms 33% belong to textile, 28% leather, 16%
Textile 33% 72
Leather 28% 61 sports, 14% food and 9% from leather. The majority of the firms’ life
Sports 16% 34 spans from 7 to 18 years.
Food 14% 31
Metal 9% 19
4.2. Reliability and validity of the model
Size
Small 48% 104
Medium 52% 113 The study has examined the reliability and validity of the construct
Firm life and model by adopting the procedure recommended by Hair et al.
1–6 years 23% 49 (2016). For checking the reliability and internal consistency, we com-
7–12 years 30% 65
puted the values of CB alpha and CR. The threshold values of CR and CB
13–18 years 30% 66
> 19 years 17% 37 alpha of all constructs are higher than the threshold values of 0.70. The
validity of the construct has been ascertained by adopting the twofold

6
T. Mahmood and M.S. Mubarik Technological Forecasting & Social Change 160 (2020) 120248

Table 4 context of IC, thus expanding its applicability. Although majority of the
Fornel-Larcker criteria. studies explain how HC, RC and SC contribute to the absorptive capa-
HC RC SC TAC OA city of a firm, researchers disagree about the direct impact of these
three dimensions on OA. For example Mubarik et al. (2019) claim that
Human Capital (HC) 0.78 HC has a direct association with firm performance; however firms
Relational Capital (RC) 0.25 0.74
processes (operationalized as structural capital) and RC (oper-
Structural Capital (SC) 0.36 0.52 0.74
Technological Absorptive capacity (TAC) 0.30 0.54 0.42 0.76
ationalized as the relationship of a firm with its suppliers, employees
Green Organizational Ambidexterity (OA) 0.41 0.39 0.49 0.57 0.75 and customers) do not directly exert any effect on their OA. Never-
theless, according to Mubarik et al. (2019), these two dimensions
Note: Diagonal values are the square root of AVE. contribute in improving the capacity to absorb outside knowledge.

Table 5 5. Conclusion, implications, and limitations

Hypotheses testing.
Path Beta p-value Decision Industry 4.0, featuring digital technologies like the internet of things
(IoT), cloud computing, big-data, and cyber-physical systems, is radically
Hypothesis 1 Human Capital → OA 0.48 0.000 Supported changing the conventional business processes. It improvesthe business
Hypothesis2 Relational Capital → OA 0.31 0.000 Supported
Hypothesis 3 Structural Capital → OA 0.28 0.004 Supported
processes and makes them robust, autonomous, automated and in-
Hypothesis 4 Human Capital → TAC→ OA 0.13 0.008 Supported telligent. The change includes the incorporation of the latest technol-
Hypothesis 5 Relational Capital →TAC→ OA 0.21 0.014 Not Supported ogies (e.g. blockchain), machines and infrastructure to create auto-
Hypothesis 5 Structural Capital → TAC→ OA 0.05 0.142 Supported mated, seamless and interconnected networks of the firms. Further,
R Square 0.67
Industry 4.0 is also converting the organizational structures by digita-
Q Square 0.48
f Square 0.55 lizing their end-to-end business processes. In such context, balancing
GFI 0.93 innovation and exploitation—the organization's ambidexterity—while
RMSEA 0.051 stepping into the fourth industrial revolution can be critical for an or-
ganization's capability. Industry 4.0 is the apex case of digital trans-
*p ≤ 0.05 Hypothesis reject.
formation, and it is not limited to the application of advanced digital
R-square values of 0.75, 0.50, and 0.25 represent substantial, moderate and
gadgets and tools but is also about incorporating them into production
weak model respectively.
f- square, effect size are according to Cohen (1988), f 2 values 0.35 (large), 0.15
and business processes. It is important to note that Industry 4.0 is not
(medium), and 0.02 (small). limited to the manufacturing sector but is equally applicable to the
q-square, predictive relevance of predictor exogenous variables as according to service sector and thus has far-reaching economic and organizational
Henseler et al. (2009), q2 values 0.35 (large), 0.15 (medium), and 0.02 (small). implications. The present study aims to go beyond the surface level to
SRMR and CFI are the values used to ascertain the overall model fitness. For understand the performance parameters of Industry 4.0 especially in
model fitness SRMR <0.08, CFI>0.90. the context of Pakistan. The country is trying to leap-frog its transfor-
mation along the lines of Industry 4.0. Against this backdrop, the ob-
approach recommended by Hair et al. (2018). First we ascertained the jective of this study was to examine the influence of IC (overall and by
convergent validity by estimating the values of Average Variance dimensions) on the OA. Likewise, the study also investigated the role of
Extracted (AVE). The AVE values of all constructs are greater than 0.50 TAC between IC and OA. Our findings exhibit a profound influence of
confirming convergent validity. Further, discriminant validity of the overall IC and its components on OA. Likewise, our findings reveal a
construct has been ascertained by computing Fornell-Larcker criteria as significant mediating role of TAC in the relationship between IC and
exhibited in Table 4. The square rooted values of AVE, shown in di- OA. In short, the findings reveal an axial role of IC in maintaining a
agonal in Table 4, are greater than inter-construct correlations. balance between an organization's ability to explore and exploit. In this
context, the role of, TAC is critical. The TAC routes organization's IC
4.3. Hypotheses testing toward maintaining a balance between the activities of exploration and
exploitation.
The results of the hypotheses testing appear in Table 5. The results Our findings offer important policy implications. First and foremost,
show that IC (β=0.21, p = 0.001) has a profound positive impact on firms first have to introspect—evaluating their present position as
OA. Further, results show that TAC (β=0.32, p = 0.000) significantly compared to Industry 4.0 requirements. A successful transformation to
mediates the relationship between IC and OA. The results of IC di- Industry 4.0 requires a broad set of capabilities, pivotal among which is
mensions show that all three dimensions of IC, namely, HC (β=0.18, OA, which can be best attained by strengthening the IC and improving
p = 0.002), RC (β=0.31, p = 0.000), and SC (β=0.28, p = 0.004), the capacity to absorb technology related knowledge. It requires orga-
have a significant direct impact on OA. The results of mediation show nizations to re-examine their strategy for approaching IC.
that absorptive capacity significantly mediates the relationship between Conventionally, organizations do not properly devise any specific
two out of three dimensions of IC (HC and RC). The value R square strategy to maintain and improve IC, leaving it at the mercy of others.
shows that the strength of the model is moderate. Further, the results of In such case, the IC, specific to the OA, may not be improved but rather
blindfolding (Q square 0.48) show the predictability of the model. In compromised. Our findings imply that organization should develop a
condensed form, the findings reveal a significant impact of IC (overall comprehensive IC development (ICD) strategy in synch with their other
and by dimensions) on OA both directly and indirectly. The results of business strategies. Execution of the ICD strategy can help organizations
the study concur with those of Stettner and Lavie (2014), to improve their productivity and innovation simultaneously. As a
Mubarik et al. (2019), and Mubarik et al. (2016). These studies show matter of fact, incorporating a sense of ambidexterity in IC can provide
that improving knowledge, skills and abilities of firms' human resource firms with a competitive edge. It implies that organizations should
improves their capacity to absorb knowledge outside the firms, which develop HC, which possesses both the ability to ramp up productivity
further contributes to both strands of performance that is, exploration and innovation. Likewise, firms should develop internal organizational
and exploitation. Likewise, Rasiah (2018) argues that both improving processes in a way that help these organizations to perform on both
HC and maintaining a collaborative relationship with customers and market exploitation and exploration. The same is the case with RC.
suppliers enable firms to expand their horizons and their absorptive Diffusing the ability of ambidexterity in relationships of an organization
capacity. These results explain the application of HC theory in the with its suppliers, customers and employees could greatly help to attain

7
T. Mahmood and M.S. Mubarik Technological Forecasting & Social Change 160 (2020) 120248

a balance between innovation and productivity. Acad. Manag. Rev. 9 (2), 284–295.
The study has some limitations, the first and foremost being that the Diaz-Fernandez, M., Pasamar-Reyes, S., Valle-Cabrera, R., 2017. Human capital and
human resource management to achieve ambidextrous learning: a structural per-
study is static in nature. As we have collected cross-sectional data, the spective. Bus. Res. Q. 20 (1), 63–77.
collection of longitudinal data on the same concepts can provide a Edvinsson, L. (1997). Developing Intellectual Capital at Skandia. 30, 8.
better understanding of the IC-ambidexterity association. Second, the Duncan, R.B., 1976. The ambidextrous organization: Designing dual structures for in-
novation. The management of organization 1 (1), 167–188.
study focuses on selected SMEs from the manufacturing sector of EEF - The Manufacturers’ Organization, 2016. The 4th Industrial Revolution (4IR): A
Pakistan. Hence the generalizability of the results needs to be done with Primer For Manufacturers. EEF. https://www.eef.org.uk/resources-and-knowledge/
care. researchand-intelligence/industry-reports/the-4th-industrial-revolution-a-primerfor-
manufacturers Accessed on: March 12, 2020.
Eisenhardt, K.M., Sull, D.N., 2001. Strategy as simple rules. Harvard business review 79
Acknowledgment (1), 106–119.
Fernández-Pérez de la Lastra, S., García-Carbonell, N., Martín-Alcázar, F., Sánchez-
Gardey, G., 2017a. Building ambidextrous organizations through intellectual capital:
None.
a proposal for a multilevel model. Intang. Cap. 13 (3), 668.
Fernández-Pérez de la Lastra, S., García-Carbonell, N., Martín-Alcázar, F., Sánchez-
Supplementary materials Gardey, G., 2017b. Intellectual capital role in ambidexterity emergence: a proposal of
a multilevel model and research agenda. J. Intell. Cap. 18 (4), 733–744.
Fernández-Pérez de la Lastra, S., García-Carbonell, N., Martín-Alcázar, F., Sánchez-
Supplementary material associated with this article can be found, in Gardey, G., 2017c. Intellectual capital role in ambidexterity emergence: a proposal of
the online version, at doi:10.1016/j.techfore.2020.120248. a multilevel model and research agenda. J. Intell. Cap. 18 (4), 733–744.
Frank, A.G., Mendes, G.H., Ayala, N.F., Ghezzi, A., 2019. Servitization and Industry 4.0
convergence in the digital transformation of product firms: a business model in-
References novation perspective. Technol. Forecast. Soc. Change 141, 341–351.
Fu, N., Ma, Q., Bosak, J., Flood, P., 2016. Intellectual capital and organizational ambi-
Ahmed, S.S., Guozhu, J., Mubarik, S., Khan, M., Khan, E., 2019. Intellectual capital and dexterity in Chinese and Irish professional service firms. J. Organ. Eff. People
business performance: the role of dimensions of absorptive capacity. Journal of Perform. 3 (2), 94–114.
Intellectual Capital 21 (1), 23–39. García-Morales, V.J., Ruiz-Moreno, A., Llorens-Montes, F.J., 2007. Effects of technology
Alänge, S., Steiber, A., 2018. Three operational models for ambidexterity in large cor- absorptive capacity and technology proactivity on organizational learning, innova-
porations. Triple Helix 5 (1), 1–25. tion and performance: An empirical examination. Technology Analysis & Strategic
Alvarez, S.A., Barney, J.B., 2001. How entrepreneurial firms can benefit from alliances Management 19 (4), 527–558.
with large partners. Acad. Manag. Perspect. 15 (1), 139–148. Gibson, C.B., Birkinshaw, J., 2004. The antecedents, consequences, and mediating role of
Andrawina, L. (2009). Relationship Between Knowledge Sharing and Absorptive Capacity organizational ambidexterity. Acad. Manag. J. 19.
Moderated by Organizational and Technology Factors: A Conceptual Model. Good, D., Michel, E.J., 2013. Individual ambidexterity: exploring and exploiting in dy-
1865–1869. namic contexts. J. Psychol. 147 (5), 435–453.
Andrawina, L., Govindaraju, R., Samadhi, T.A., Sudirman, I., 2008. Absorptive capacity Grant, R.M., 1996. Toward a knowledge-based theory of the firm: knowledge-based
moderates the relationship between knowledge sharing capability and innovation theory of the firm. Strateg. Manag. J. 17 (S2), 109–122. https://doi.org/10.1002/
capability. In: Proceedings of the IEEE International Conference on Industrial smj.4250171110.
Engineering and Engineering Management. IEEE, pp. 944–948. Gupta, O., Roos, G., 2001. Mergers and acquisitions through an intellectual capital per-
Andrews, R., 2010. Organizational social capital, structure and performance. Hum. Relat. spective. J. Intell. Cap. 2 (3), 297–309.
63 (5), 583–608. Güttel, W.H., Konlechner, S.W., 2009. Continuously hanging by a thread: managing
Andriopoulos, C., Lewis, M.W., 2009. Exploitation-exploration tensions and organiza- contextually ambidextrous organizations. Schmalenbach Bus. Rev. 61 (2), 150–172.
tional ambidexterity: managing paradoxes of innovation. Organ. Sci. 20 (4), Hershberg, T., 1996. Human capital development: America's greatest challenge. Ann. Am.
696–717. Acad. Pol. Soc. Sci. 544 (1), 43–51.
Antonelli, C., Colombelli, A., 2011. The generation and exploitation of technological Hitt, M.A., Bierman, L., Shimizu, K., Kochhar, R., 2001. Direct and moderating effects of
change: market value and total factor productivity. J. Technol. Transf. 36 (4), human capital on strategy and performance in professional service firms: a resource-
353–382. based perspective. Acad. Manag. J. 44 (1), 13–28.
Arenas, T., Lavanderos, L., 2008. Intellectual capital: object or process? J. Intell. Cap. 9 Hitt, M.A., Keats, B.W., DeMarie, S.M., 1998. Navigating in the new competitive land-
(1), 77–85. scape: building strategic flexibility and competitive advantage in the 21st century.
Arshad, R., Noor, A.H.M., Yahya, A., 2015. Human capital and Islamic-based social im- Acad. Manag. Perspect. 12 (4), 22–42.
pact model: small enterprise perspective. Procedia Econ. Financ. 31, 510–519. Horváth, D., Szabó, R.Z., 2019. Driving forces and barriers of Industry 4.0: do multi-
Ashton, C., Morton, L., 2005. Managing talent for competitive advantage: taking a sys- national and small and medium-sized companies have equal opportunities? Technol.
temic approach to talent management. Strateg. HR Rev. 4 (5), 28–31. Forecast. Soc. Change 146, 119–132.
Asiaei, K., Jusoh, R., 2015. A multidimensional view of intellectual capital: the impact on Hsu, Y.H., Fang, W., 2009. Intellectual capital and new product development perfor-
organizational performance. Manag. Decis. 53 (3), 668–697. mance: the mediating role of organizational learning capability. Technol. Forecast.
Asiaei, K., Jusoh, R., Bontis, N., 2018. Intellectual capital and performance measurement Soc. Change 76 (5), 664–677.
systems in Iran. J. Intell. Cap. 19 (2), 294–320. Itami, H., Roehl, T.W., 1991. Mobilizing Invisible Assets. Harvard University Press,
Barathi Kamath, G., 2007. The intellectual capital performance of the Indian banking Boston: USA.
sector. J. Intell. Cap. 8 (1), 96–123. Jansen, J.J.P., Simsek, Z., Cao, Q., 2012. Ambidexterity and performance in multiunit
Barney, J., 2001. The resource-based view of the firm: ten years after 1991. J. Manag. 27 contexts: cross-level moderating effects of structural and resource attributes. Strateg.
(6), 625–641. Manag. J. 33 (11), 1286–1303.
Baškarada, S., Watson, J., Cromarty, J., 2016. Leadership and organizational ambi- Jørgensen, K.M., Boje, D.M., 2010. Resituating narrative and story in business ethics. Bus.
dexterity. J. Manag. Dev. 35 (6), 778–788. Eth. Eur. Rev. 19 (3), 253–264.
Becker, G.S., 1962a. Investment in human capital: a theoretical analysis. J. Polit. Econ. 70 Junni, P., Sarala, R.M., Taras, V., Tarba, S.Y., 2013. Organizational ambidexterity and
(5, Part 2), 9–49. performance: a meta-analysis. Acad. Manag. Perspect. 27 (4), 299–312.
Birkinshaw, J., Gupta, K., 2013. Clarifying the distinctive contribution of ambidexterity to Kafouros, M., Love, J.H., Ganotakis, P., Konara, P., 2020. Experience in R&D collabora-
the field of organization studies. Acad. Manag. Perspect. 27 (4), 287–298. tions, innovative performance and the moderating effect of different dimensions of
Bontis, N., 1998. Intellectual capital: an exploratory study that develops measures and absorptive capacity. Technol. Forecast. Soc. Change 150, 119757.
models. Manag. Decis. 36 (2), 63–76. Kamall Khan, Y. (2016). Impact of Structural Capital on Innovation in the Australian SMEs.
Bontis, N., Fitz-Enz, J., 2002. Intellectual capital ROI: a causal map of human capital 302–312.
antecedents and consequents. J. Intell. Cap. 3 (3), 223–247. Kang, S.-.C., Morris, S.S., Snell, S.A., 2007. Relational archetypes, organizational learning,
Bontis, N., Nikitopoulos, D., 2001. Thought leadership on intellectual capital. J. Intell. and value creation: extending the human resource architecture. Acad. Manag. Rev. 32
Cap. 2 (3), 183–191. (1), 236–256.
Burt, R.S., 2017. Structural holes versus network closure as social capital. In: Lin, N., Kang, S.-.C., Snell, S.A., 2009. Intellectual capital architectures and ambidextrous
Cook, K., Burt, R.S. (Eds.), Social Capital Routledge: New York, 1st ed. pp. 31–56. learning: a framework for human resource management. J. Manag. Stud. 46 (1),
Carte, N., 2005. The maximum achievable profit method of patent valuation. Int. J. Innov. 65–92.
Technol. Manag. 02 (02), 135–151. Khalique, M., Bontis, N., Abdul Nassir bin Shaari, J., Hassan Md. Isa, A., 2015. Intellectual
Chang, A.A., Mubarik, M.S., Naghavi, N., 2020. Passing on the legacy: exploring the capital in small and medium enterprises in Pakistan. J. Intell. Cap. 16 (1), 224–238.
dynamics of succession in family businesses in Pakistan. J. Fam. Bus. Manag https:// Kor, Y.Y., Mahoney, J.T., 2004. Edith Penrose's (1959) contributions to the resource-
doi.org/10.1108/JFBM-10-2019-0065. based view of strategic management. J. Manag. Stud. 41 (1), 183–191.
Chu, P.Y., Lin, Y.L., Hsiung, H.H., Liu, T.Y., 2006. Intellectual capital: an empirical study Lavie, D., Stettner, U., Tushman, M.L., 2010. Exploration and exploitation within and
of ITRI. Technol. Forecast. Soc. Change 73 (7), 886–902. across organizations. Acad. Manag. Ann. 4 (1), 109–155.
Crossan, M.M., Lane, H.W., White, R.E., 1999. An organizational learning framework: Lazzarotti, V., Manzini, R., Nosella, A., Pellegrini, L., 2017. Innovation ambidexterity of
from intuition to institution. Acad. Manag. Rev. 24 (3), 522–537. open firms. The role of internal relational social capital. Technol. Anal. Strateg.
Daft, R.L., Weick, K.E., 1984. Toward a model of organizations as interpretation systems. Manag. 29 (1), 105–118.

8
T. Mahmood and M.S. Mubarik Technological Forecasting & Social Change 160 (2020) 120248

Lev, B., Zambon, S., 2003. Intangibles and intellectual capital: an introduction to a special competitive advantage. J. Knowl. Manag. 7 (1), 20–31. https://doi.org/10.1108/
issue. Eur. Account. Rev. 12 (4), 597–603. 13673270310463590.
Liu, H., Chen, C., 2009. Corporate organizational capital, strategic proactiveness and firm Simsek, Z., Heavey, C., Veiga, J.F., Souder, D., 2009. A typology for aligning organiza-
performance: an empirical research on Chinese firms. Front. Bus. Res. China 3 (1), tional ambidexterity's conceptualizations, antecedents, and outcomes. J. Manag.
1–26. Stud. 46 (5), 864–894.
Lopes-Costa, J.A., Munoz-Canavate, A., 2015. Relational capital and organizational per- Stettner, U., Lavie, D., 2014. Ambidexterity under scrutiny: exploration and exploitation
formance in the Portuguese hotel sector (NUTS II Lisbon). Procedia Econ. Financ. 26, via internal organization, alliances, and acquisitions: exploration and exploitation.
64–71. Strateg. Manag. J. 35 (13), 1903–1929.
Lund Vinding, A., 2006. Absorptive capacity and innovative performance: a human ca- Stewart, T.A., 1997. Intellectual Capital. The New Wealth of Organizations. Nicholas
pital approach. Econ. Innov. New Technol. 15 (4–5), 507–517. Brealey, London.
March, J.G., 1991. Exploration and exploitation in organizational learning. Organization Sullivan, P.H., Sullivan, P.H., 2000. Valuing intangibles companies – an intellectual ca-
science 2 (1), 71–87. pital approach. J. Intell. Cap. 1 (4), 328–340.
Mom, T.J.M., van Neerijnen, P., Reinmoeller, P., Verwaal, E., 2015. Relational capital and Teece, D.J., 2012. Dynamic capabilities: routines versus entrepreneurial action: routines
individual exploration: unravelling the influence of goal alignment and knowledge versus entrepreneurial action. J. Manag. Stud. 49 (8), 1395–1401.
acquisition. Organ. Stud. 36 (6), 809–829. Tiwana, A., 2008. Do bridging ties complement strong ties? An empirical examination of
Mouritsen, J., 1998. Driving growth: economic value added versus intellectual capital. alliance ambidexterity. Strateg. Manag. J. 29 (3), 251–272.
Manag. Account. Res. 9 (4), 461–482. Vlačić, E., Dabić, M., Daim, T., Vlajčić, D., 2019. Exploring the impact of the level of
Mubarik, M.S., Chandran, V.G.R., Devadason, E.S., 2018. Measuring human capital in absorptive capacity in technology development firms. Technol. Forecast. Soc. Change
small and medium manufacturing enterprises: what matters? Soc. Indic. Res. 137 (2), 138, 166–177.
605–623. Volberda, H.W., Lewin, A.Y., 2003. Co-evolutionary dynamics within and between firms:
Mubarik, M.S., Govindaraju, C., Devadason, E.S., 2016a. Human capital development for from evolution to co-evolution. J. Manag. Stud. 40 (8), 2111–2136.
SMEs in Pakistan: is the “one-size-fits-all” policy adequate? Int. J. Soc. Econ. 43 (8), Wernerfelt, B., 1984. A resource-based view of the firm. Strateg. Manag. J. 5 (2),
804–822. 171–180.
Mubarik, M.S., Naghavi, N., Mahmood, R.T., 2019a. Intellectual capital, competitive Wong, K.Y., Aspinwall, E., 2005. Knowledge management: case studies in SMEs and
advantage and the ambidexterity liaison. Hum. Syst. Manag. 38 (3), 267–277. evaluation of an integrated approach. J. Inf. Knowl. Manag. 04 (02), 95–111.
https://doi.org/10.3233/HSM-180409. Youndt, M.A., Subramaniam, M., Snell, S.A., 2004. Intellectual capital profiles: an ex-
Mubarik, M.S., Naghavi, N., Mubarak, M.F., 2019b. Impact of supplier relational capital amination of investments and returns: intellectual capital profiles. J. Manag. Stud. 41
on supply chain performance in Pakistani textile industry. Asian Econ. Financ. Rev. 9 (2), 335–361.
(3), 318–328. Zahra, S.A., George, G., 2002. Absorptive capacity: a review, reconceptualization, and
Mubarik, S., Chandran, V., Devadason, E.S., 2016b. Relational capital quality and client extension. Acad. Manag. Rev. 27 (2), 185–203.
loyalty: firm-level evidence from pharmaceuticals, Pakistan. Learn. Organ. 23 (1), Zameer, H., Wang, Y., Yasmeen, H., Mubarak, S., 2020. Green innovation as a mediator in
43–60. the impact of business analytics and environmental orientation on green competitive
Naghavi, N., Mubarak, M.S., 2019. Negotiating with Managers from South Asia: India, Sri advantage. Manag. Decis. https://doi.org/10.1108/MD-01-2020-0065. Ahed of print.
Lanka, and Bangladesh. In: The Palgrave Handbook of Cross-Cultural Business Becker, G. S. (1964). Investment in human beings. The Journal of Political Economy, 70,
Negotiation. Palgrave Macmillan, Cham, pp. 487–514. Part-2(5), 9–49.
Nonaka, I., Konno, N., 1998. The concept of “Ba”: building a foundation for knowledge Mubarik, M. S. (2015). Human capital and performance of small & medium manufacturing
creation. Calif. Manag. Rev. 40 (3), 40–54. enterprises: a study of Pakistan (Doctoral dissertation, University of Malaya). Accesed
Penrose, E., 1959. The theory of the growth of the firm. Oxford University Press, New from: http://studentsrepo.um.edu.my/6573/1/Shujaat_Mubarik_Thesis_Final_
York, NY. Submission_.pdf (Date: 03 March 2020).
Prahalad, C.K., Hamel, G., 2007. Strategy as a field of study: why search for a new
paradigm? Strateg. Manag. J. 15 (S2) 5–16. 02. Tarique Mahmood: Tarique Mahmood is senior Lecturer and PhD research fellow at
Raisch, S., Birkinshaw, J., Probst, G., Tushman, M.L., 2009. Organizational ambidexterity: Mohammad Ali Jinnah University, jarachi. His-areas of interest are intellectual capital
balancing exploitation and exploration for sustained performance. Organ. Sci. 20 (4), management and organizational ambidexterity.
685–695.
Reza, S., Mubarik, M.S., Naghavi, N., Nawaz, R.R., 2020. Relationship marketing and
third-party logistics: evidence from hotel industry. J. Hosp. Tour. Insights 3 (3), Prof. Dr. Muhammad Shujaat Mubarik: Muhammad Shujaat Mubarak is Professor at
371–391. Mohammad Ali Jinnah University Karachi, Pakistan. He is also Dean of Faculty of
Russo, M.V., Fouts, P.A., 1997. A Resource-Based perspective on corporate environmental Business Administration & Social Sciences, Karachi. His-areas of interest are intellectual
capital, human capital, supply chain economics, and sustainability. On the methodology
performance and profitability. Acad. Manag. J. 40 (3), 534–559.
Secundo, G., Ndou, V., Del Vecchio, P., De Pascale, G., 2020. Sustainable development, side, he is also expert in applying Multi-attributes Data Management like AHP, ANP,
intellectual capital and technology policies: a structured literature review and future DEMATEL, etc. He has published in journals of high repute including Management
research agenda. Technol. Forecast. Soc. Change 153, 119917. Decision, International Journal of Emerging Markets, Social Indicator Research and
Seleim, A.A.S., Khalil, O.E.M., 2011. Understanding the knowledge management-in- Emerging Markets Trade and Finance. Dr. Mubarik also hosts the TV talk show “Corporate
Lounge” at a private satellite TV channel in Pakistan. He also provides consultancy on
tellectual capital relationship: a two-way analysis. J. Intell. Cap. 12 (4), 586–614.
Sharkie, R., 2003. Knowledge creation and its place in the development of sustainable supply chain and service excellence to national and multinational companies.

9
Oper Manag Res (2016) 9:1–10
DOI 10.1007/s12063-016-0106-z

The direct digital manufacturing (r)evolution: definition

of a research agenda
Jan Holmström 1 & Matthias Holweg 2 & Siavash H. Khajavi 1 & Jouni Partanen 3

Received: 22 September 2015 / Revised: 21 December 2015 / Accepted: 10 January 2016 / Published online: 18 January 2016
# Springer Science+Business Media New York 2016

Abstract Direct digital manufacturing, or ‘3D printing’ as it operations management researchers to question established
is more commonly known, offers a wealth of opportunities for practices such as scheduling, batch sizing and inventory man-
product and process innovation, and is often touted to ‘revo- agement in low-volume, high-variety contexts. Furthermore,
lutionize’ today’s manufacturing operations and its associated an increasing adoption of direct digital manufacturing will
supply chains structures. Despite a growing number of suc- drive structural shifts in the supply chain that are not yet well
cessful 3D printing applications, however, evidence of any understood. We summarize these challenges by defining the
displacement of traditional manufacturing is limited. In this research agenda at factory, supply chain, and operations strat-
paper we seek to separate hype surrounding DDM from eco- egy level.
nomic reality in order to ground the future research agenda for
the Operations Management field. By opposing direct digital Keywords 3D printing . Additive manufacturing . Direct
manufacturing with traditional tool-based manufacturing, we digital manufacturing . Process innovation . Supply chain
show that direct digital manufacturing so far lags behind by structure
several orders of magnitude compared to traditional
manufacturing methods. Yet we also find that direct digital
manufacturing clearly is on an improvement trajectory that 1 Introduction
eventually will see it being able to compete with traditional
manufacturing on a unit cost basis. As such we conclude that ‘Additive manufacturing’, ‘rapid prototyping’ or ‘3D printing’
direct digital manufacturing will increasingly challenge all describe a set of rapidly evolving manufacturing techniques
that are beginning to complement, and even replace, tradition-
al manufacturing techniques in low-volume and high-variety
* Jan Holmström contexts. These techniques enable the direct manufacture of
[email protected]
parts from an original digital design (or physical scan) without
Matthias Holweg
tooling and set-up, much in the same way as laser printers
[email protected] enables printing from an original or copied file without type
setting (Hopkinson et al. 2006; Gibson et al. 2010). In the
Siavash H. Khajavi
[email protected] following we will thus refer to this class of manufacturing
techniques as direct digital manufacturing (DDM). As DDM
Jouni Partanen
[email protected]
technologies become more widely used, the cost per unit has
been reducing accordingly, making their application in
low-volume and service supply chains increasingly common
1
Department of Industrial Engineering and Management, Aalto (Economist 2013, Wohlers 2015). High-profile applications
University, Helsinki, Finland include NASA’s rocket engine (NASA 2013), Nike and
2
Saïd Business School, University of Oxford, Oxford, UK Adidas making their high-end trainers to custom-fit
3
Department of Engineering Design and Production, Aalto University, (Fitzgerald 2013), and GE’s fuel nozzle for the LEAP engine.
Helsinki, Finland In addition to the common predictions of growth for the
2 Holmström J. et al.

sector, there are also critical voices outlining the associated innovative engineering and chemistry, parts are projected into
challenges in terms of printing times and cost (Berger 2013, liquid from the bottom through an oxygen permeable surface.
Financial Times 2014). In this context it is instructive to note The light triggers photopolymerization while the oxygen in-
that DDM in Gartner’s annual technology review (Gartner hibits it, allowing a part to be continuously grown while lifted
2013) of commercial promise has been classified as likely up from the liquid. The method may eventually allow
reaching its productivity threshold within the next 2–5 years. projecting a 3D model of a part (hologram) into a liquid that
Much of the operations management (OM) research on solidifies according to the projection 25 to 100 times faster
DDM centers on opportunities for innovation and customiza- than in current stereo-lithography machines. From an OM
tion (de Jong and de Bruijn 2013), while research on the perspective the method is interesting as it promises a reduction
implications on operations management in the wider in the time to produce a part by an order of magnitude, at a
manufacturing sector is still nascent (Chen et al. 2015). quality similar as for injection-molded parts, with equipment
Early studies have investigated how DDM changes operations and material costs that are no more expensive than for current
management in the factory (Mellor et al. 2014) and the supply stereo-lithography methods. As the CLIP example demon-
chain (Hopkinson et al. 2006; Holmström et al. 2010), as well strates, one should expect DDM technologies to make consid-
as strategic and operational implications at firm and supply erable improvements in quality, cost and speed over the com-
chain level (Weller et al. 2015). An integrated approach or ing years.
framework how to consider the strategic aspects of DDM
however is still amiss. In this paper we seek to close this gap
by developing a research agenda for operations management
of DDM at the factory, supply chain and operations strategy 3 Research agenda at factory level
level. We proceed with a general introduction to the concept of
direct digital manufacturing. DDM provides a prompt to rethink many traditional OM prac-
tices such as inventory management, job shop scheduling, and
batch sizing. Replacing process-based job shop operations
2 A primer on direct digital manufacturing with product-model driven operations will not only change
material flows within the manufacturing facility dramatically,
The origins of direct digital manufacturing (DDM) are ‘rapid but also fundamentally change the nature of the operational
prototyping’ applications used predominantly in product de- planning task. Fig. 1 illustrates how in job shops today batches
sign as of the late 1980s (Hopkinson et al. 2006). Today there of parts are produced, stored in inventory, possibly kitted, and
is a widespread use not just for prototyping but also eventually assembled as final products on lines or in cells. In a
tool-making and low-volume manufacturing across industry DDM operation, the elimination of tools and related set-ups
sectors. To illustrate the variety in manufacturing methods, as makes batching entirely redundant. Instead, kits of different
well as the breadth of materials that can be used in DDM, parts for a particular assembly can be configured and printed
Table 1 provides an overview of additive manufacturing in a common build envelope before being assembled as spec-
methods within DDM, their typical applications, and equip- ified by the customer, with little need for production schedul-
ment cost. The list of reported applications is long: the aero- ing or inventory management. In DDM, real-time tracking and
space industry is a pioneer in DDM of parts for final products, decentralized control of individual parts and kits can be inte-
with Boeing for example 3D printing parts for the F-18 Super grated into the product model offering a novel mechanism to
Hornets and 787 commercial jetliners, while BAE systems manage not only the manufacturing, but the overall operation
prints components for the Tornado fighter jets (Hopkinson et (Främling et al. 2007).
al. 2006, Freedman 2011, BAE Systems 2014). Another early A range of research questions emerges for OM research at
adopter is the medical industry producing customized pros- the operations or factory level, as outlined in Table 2. In gen-
thetics and orthopedic implants, hearing aid shells, dental eral terms, DDM and decentralized control of operations share
crown and bridges, and custom orthodontic braces (Wohlers the premise that each unit is a unique instance with it an
2015). DDM is also used for more conventional manufactured identity, history and purpose, i.e. a product individual. When
products: LUXeXcel for example has developed DDM of designs are produced directly from a digital design model, the
lenses and offers car headlight lenses for spare parts control information required to assemble the final product
(Economist 2014). from its component parts can be embedded in the design mod-
In addition to the established DDM methods already in use el (Främling et al. 2007), reducing the need for scheduling and
there are potentially disruptive new methods under develop- planning. Thus, DDM is likely to increase the relevance of
ment. For example Continuous Liquid Interface Production research dealing with decentralized control of production and
(CLIP), a novel method for stereo-lithography shows consid- handling of products (van Brussel et al. 1998; Kärkkäinen
erable potential (Castelvecchi 2015): using a combination of et al. 2003; McFarlane et al. 2003).
Table 1 Main types of material processes used in additive manufacturing. Sources: Wohlers (2015), Roland Berger (2013), Gibson et al. (2010), ASTM (2013)

Method (alternative Description of process Materials used Typical applications Typical equipment cost
names in brackets) in US$
(range in brackets)

Material extrusion Solid material is heated up and selectively Plastics (ABS, polycarbonate, PC/ABS blend, R&D prototypes, investment ~ 30,000
(Fused deposition extruded in liquid form through a nozzle PLA, Polyamide) and also biocompatible casting pattern production (10,000–380,000)
The direct digital manufacturing (r)evolution

modelling, FDM) to produce thin layers of object upon cooling plastics (PEEK and PEKK)
down and solidification
Material jetting The droplets of raw material (e.g.: photopolymer Plastics (Acrylic, acrylate, epoxies), metals Marketing prototypes, investment ~ 100,000
or wax) are selectively deposited to form the (stainless steel which requires post casting pattern production (20,000–700,000)
object, layer by layer production furnace cycle)
Binder jetting A liquid binder is selectively deposited to Metals (tool steel or stainless steel with Marketing prototypes, moulds for casting ~ 200,000
join the raw powder bronze infiltrate), glass, plaster-based powder (16,000–1,800,000)
Sheet lamination The sheets of raw material being cut (e.g.: by Metals (aluminium alloys, copper alloys, Prototyping for simple parts ~ 40,000
LASER or blade) to shape the layers and titanium alloys and stainless steel), paper (36,000–73,000)
then bonded together (the process can be
done in reverse)
Vat photo-polymerization The vat of liquid photopolymer is cured Plastics (acrylic, acrylate, epoxies), photopolymers Prototyping and demonstration ~ 300,000
(Stereo-lithography, SLA) selectively and layer by layer through for metal casting, ceramics (alumina) products, tool-making and (4000–800,000)
the exposure to light pattern production
Powder-bed fusion A source of thermal energy (usually LASER) Plastics (nylon, polystyrene, polypropylene, Industrial final parts production, ~ 500,000
(Selective LASER sintering, is used to selectively fuse the raw material and glass, carbon and aluminium-filled tool-making, orthopedic and (85,000–2000,000)
SLS or Selective Laser powder to form the layers of final part polyamide powders, DuraForm Flex, Windform), dental implants manufacturing
Melting, SLM) metals (tool steels, stainless steel, commercially
pure titanium, titanium alloys, aluminium alloys,
nickel-based alloys, cobalt-based alloys,
copper-based alloys, gold, silver), ceramics
(alumina), polystyrene for metal casting
Direct energy deposition A flow of raw material powder is being Metals (tool steels, stainless steel, commercially Full density near shape final parts ~ 500,000
melted before selective deposition through pure titanium, titanium alloys, Al alloys, (300,000–5,000,000)
the use of a focused thermal energy nickel-based alloys, cobalt-based alloys,
(e.g.: LASER, electron beam or plasma arc) copper-based alloys, gold, silver, combination
of powdered material)
3
4 Holmström J. et al.

Fig. 1 A schematic of tool-based manufacturing versus DDM

The continuous supply of batches of components to the customer, and managing WIP inventories. Shifting attention
assembly line has many drawbacks, from storage to handling from batches to individual parts and customized kits calls into
and quality cost. These challenges are amplified with question the need for inventory management based on the
increasing product variety, as manufacturers have to accounting of stock-keeping-units (SKUs), and movement of
procure and provide the required components for each SKUs between inventory locations. Tracking product individ-
product variant. DDM in turn could provide customized uals and parts can replace more conventional inventory
kits of components for the respective product variant at management without the loss of functionality and infor-
the point of assembly, just-in-time. mation (Rönkkö et al. 2007). On the contrary, replacing
In addition, DDM effectively reduces the need to schedule inventory management with item and product tracking
in job shop operations to merely sequencing print jobs for further increases the ability to control material flows,
each printer: rather than having to coordinate multiple con- leading to greater responsiveness and efficiency
nected production steps, most parts will be produced in ‘n = 1’ (Holmström et al. 2011; Arnäs et al. 2013).
production steps. The default order penetration point effec- The introduction of DDM within a factory also introduces a
tively becomes the single-stage manufacturing process new planning problem: How to combine DDM with other
(Olhager 2003). Thus the entire uncertainty about throughput types of manufacturing? Currently DDM is used in distinct
times vanishes, as does the challenging tasks of creating viable phases of a product lifecycle for prototyping, tooling and ini-
production schedules, issuing reliable delivery dates to the tial (low) volume production (Khajavi et al. 2015), and does

Table 2 Research agenda, at factory level

Tool-based manufacturing DDM Key research questions

Production technology Tooling and process based Digital model based What are the realms of applicability
for direct manufacturing technologies?
In what contexts is DDM relevant,
or even superior?
Material flow Production batches flow Kits on demand to on-site assembly What are the potential efficiency gains in
through production stages material flow that DDM offers?
Planning structures Stock keeping units (SKUs) Product level, What are the most appropriate planning
Bill of materials (BOM) assembly kits structures for DDM?
Production planning Lot size, load profiles per process, Total load of printers How to combine tool-based and DDM
production schedule inside the factory?
The direct digital manufacturing (r)evolution 5

not directly affect how volume production is managed and converse, when to switch to specialized suppliers in global
planned. However, as adoption increases also in volume pro- supply chains. The switch from tool-based to DDM is not
duction DDM needs to be integrated in the plans of tool-based primarily a supply chain issue, but a production technology
manufacturing without loss of benefits. Production planning issue. When technologically feasible the switch is desirable
approaches do not become redundant until all manufacturing for the front end of product introductions and low volume
for a product uses direct digital methods. production as it offers attractive opportunities in risk reduction
and simplified engineering change management (Khajavi et
al. 2015).
4 Research agenda, at supply chain level Warranty, liability and IP protection may inhibit the adop-
tion of localized DDM, but such considerations are being
The advent of DDM already is, and will continue to lead, to overtaken by the actual growth in many supply chains. The
structural shifts in value-added in the supply chain that require emerging challenge for supply chain management research is
SCM researchers to address a number of emerging research how to manage the distribution of design models in the supply
questions, as outlined in Table 3. A widespread adoption of chain, and whether to use open source or controlling strate-
DDM is likely to cause a collapse of supply chain tiers where gies. In mass-manufacturing, designs and tools do not need to
parts can be made locally, as is readily seen in early examples be widely distributed and available. However, if manufactur-
of direct manufacturing using numerically controlled ma- ing is to be provided locally, this requires a global solution for
chines (cf. Nielsen and Holmström 1995). Responding to distributing and managing the design models. Opening up the
widespread adoption requires research on when and how to design model by making it available to everyone simplifies the
switch to supply via DDM. For example, DDM of spare parts distribution and management of designs, but restricts the op-
(cf. Khajavi et al. 2014) reduces the need for having any tiers portunity for designers to generate revenues from their IP. The
in the supply chain, but may require capacity pooling or spe- ‘maker’ economy so far is largely based on open source de-
cialized service providers for local access to DDM equipment signs, creating value through different combinations of cus-
(Holmström et al. 2010, Holmström and Partanen 2014). The tomer and supplier co-creation (Anderson 2012). The inability
low barriers to entry will also provide new opportunities for to protect IP may prove to be a hindrance for the adoption of
firms to provide customized and novel products at low vol- DDM for established products.
ume. New markets and service providers are already emerg- A further change is related to technological improvements
ing, as for example shown in the case of the start-up of products, i.e. when should firms develop and distribute new
Shapeways that offer online contract printing services. products, as opposed to systematically improving existing
Furthermore, how to manage a mix of tool-based and DDM products? DDM would effectively enable firms to improve
is an emerging challenge for SCM research. To date DDM has or update the product after every single unit produced (or
primarily been used for prototypes and tooling, and to a lim- printed), without any constraints related to tooling, existing
ited degree for customized products. However, as DDM is stocks, or coordination with suppliers. In combination with
being adopted for low volume production, the issue becomes local manufacturing this implies the possibility of continuous
more challenging. When should a switch from DDM to updating and refurbishing of products in use.
tool-based manufacturing take place to leverage the econo- When products are continuously updated and refurbished
mies of scale offered by specialized suppliers in global supply using DDM a novel challenge for supply chain management
chains? Here, the issue is not when to switch to DDM, but the research emerges in how to manage digital design models for

Table 3 Research agenda, at supply chain level

Tool-based manufacturing DDM Key research questions

Supply chain structure Specialized suppliers in Pools of local (generalized) What determines the trade-off to switch to
global supply chains service providers supply via DDM?
How manage a mix of tool-based and DDM
based supply chains?
Intellectual property Controlled designs and tools Global distribution of design models How to manage the distribution of design models
rights (IPR) required for localized DDM in the supply chain, and whether to use open
source or controlling strategies?
Product improvement Replacement: develop Upgrade/refurbish: ability to improve When should firms develop and distribute new
new models products-in-use products, as opposed to systematically improving
existing products?
How to manage digital design models in revisions of
heterogeneous products in use?
6 Holmström J. et al.

revisions of heterogeneous products in use. When products in prototyping and tooling, be used to improve tool-based
use are continuously refurbished and updated, over time fur- manufacturing. The most widely adopted DDM based opera-
ther revisions run the risk of becoming increasingly complex. tional practice is prototyping, but also on demand manufactur-
In software engineering the analogous problem has been dealt ing of customized parts and on spare parts can be found
with through object oriented programming and solution archi- (Wohlers 2015). Supporting tool-based manufacturing by pro-
tectures that can be modified according to design patterns ducing tools using DDM is potential practice in this same
(Gamma et al. 1994). A design pattern is a meta-design devel- category. However, currently DDM is only used more exten-
oped for a specific purpose and which can be modified and sively for so called Bsoft-tooling^, i.e. making tools for short
reused in a wide range of settings (Alexander, 1999). As in production runs. Another research question in support of pro-
software engineering a focus on objects and design patterns cess improvement is how DDM can be used to support lean
for handling objects may provide the basis for architectures and six sigma implementations used in traditional repetitive
that can effectively deal with diversity of designs that change manufacturing contexts (Bicheno and Holweg, 2016).
over time (Främling et al. 2007). As yet, the operations strategy to replace tool-based
DDM and the shift of economic value from manufacturing manufacturing with DDM (where cost competitive) cannot
to design will reduce the dependence on contract manufacturers be observed in practice. To date on-demand manufacturing
and parts suppliers. In the global supply chains that are common of parts has not replaced tool-based manufacturing to stock,
today manufacturing is strongly driven by economies of scale: but has been used to bridge demand in special situations
sharing the cost of design and tooling across larger production (Khajavi et al. 2015). The research question for OM is to
volumes reduces unit cost, as does reduced change-over and assess the gap in performance between tool-based
ramp-up costs from increasing lot sizes. Repetitive scale-driven manufacturing and DDM. To provide a current assessment
manufacturing processes are the core reason why in most situ- of the potential for DDM to begin replacing tool-based
ations it is advantageous to buy parts and components from manufacturing in a practical setting, we have investigated
contractors and parts suppliers rather than making them the cost and time to make a common type of product using
in-house. However, in DDM it is the re-use of the design model DDM. For this assessment we approached a vehicle manufac-
that is a potential source of economies of scale. Here, a turer who provided us with the complete bill of materials,
pattern-based design model that is easy to modify and can be including weight, for one of its small executive sedans.
reused many times and for different purposes may become a Based on this data we estimated the DDM cost and time for
novel source of competitiveness for manufacturing firms. an entire vehicle, for each material type found in the car. Table
5 below shows the breakdown of parts by material category, as
well as our estimates of the respective time and cost of man-
5 Research agenda, at operations strategy level ufacture. It is important to consider this extrapolation of
3D-printing a ‘real’ steel car body in the context of recent
Many lofty claims that DDM will replace all traditional claims by ‘Urbee 2’, for example, that a passenger car can
manufacturing in a dramatic fashion have been made (see be manufactured from plastic in a general purpose 3D printing
for example: D’Aveni 2013, 2015). Such claims are based facility at a cost estimated to be around $50,000 (Wired 2013).
on the conceptual potential of DDM, but lack in the assess- While the estimate is based on a series of necessary as-
ment of the economic case for introducing DDM in a given sumptions about the number and geometry of parts to be
setting (Holweg, 2015). From an operations strategy perspec- printed, it indicates where DDM currently stands in relation
tive the relevant question is how to harness the potential of this to a tool-based manufacturing. As can be seen, there still is
technology for strategic purposes. We see three mutually sup- significant gap of several orders of magnitude for digital
portive operations strategies: (1) using DDM to improve manufacturing techniques becoming cost competitive in the
tool-based manufacturing, (2) replacing tool-based large-volume manufacturing of car parts. However, for special
manufacturing with DDM when and where it becomes a cost applications such as spare parts for machines no more sup-
competitive alternative, and (3) developing innovative solu- ported by the manufacturer or customized products the elim-
tions leveraging the unique characteristics of DDM. Table 4 ination of tooling in DDM based operations significantly re-
summarizes a research agenda for these complementing oper- duces both cost and response time and enables replacement of
ations strategies. tool-based manufacturing. In the typographic industry direct
To date the trade-off between higher manufacturing cost printing based on the combination of desktop publishing and
and more direct value adding process has meant that DDM laser printing has in two decades replaced tool-based printing
has been introduced in special situations and for specific tasks for professional quality products for volumes of up to several
to support tool-based manufacturing. The primary research hundred as desktop publishing and printers have improved.
question for supporting an improvement strategy in operations A further research question related to the replacement of
is how can DDM based supportive practices, such as tool-based manufacturing with DDM is to identify the
The direct digital manufacturing (r)evolution 7

Table 4 Research agenda, at operations strategy level

Operations strategy Current state of the art Potential practices Key research questions

Improve tool-based DDM used for prototyping Support tool-based manufacturing How can DDM be introduced to improve
manufacturing and on-demand manufacturing through faster and cheaper tooling tool-based manufacturing?
of parts How can DDM-based practices, such as
prototyping and tooling, be used to support
lean and six sigma implementations?
Replace tool-based Strategy not observed Replace tool-based manufacturing What is the gap in performance between
manufacturing as equipment improves, designs tool-based manufacturing and DDM
can be optimized for DDM in different settings?
What are the mechanisms through which
DDM can close the performance gap?
What is the time frame for DDM to close
the performance gap in different settings?
Innovate based Strategy not observed Use DDM for innovative solutions What are potential DDM based innovative solutions?
on DDM that improve products in use What are the development and adoption
or develop highly context thresholds for different types of DDM
adapted processes based innovative solutions?

mechanism of improvement in DDM. In comparison to the manufacturing is context-specific, we can examine its evolu-
situation in typographic printing the replacement of tool-based tion to date. To illustrate the development Fig. 2 approximates
manufacturing with DDM is driven not only by improvements the number of days required to print the steel of the example
in equipment but by the additional mechanism of product car over a ten year period. At each point in time the best
redesign. Replacement is enabled by 3D modeling software available laser sintering equipment is being used. In 2004,
to optimize the design of DDM produced parts for both re- 780 days would have been required; as of 2014, the time to
duced weight and reduced assembly by integrating parts print the steel components has decreased to 144 days. Over the
(Eyers and Dotchev 2010; Chen et al. 2015). Such design ten year time period the build throughput in terms of cubic
change speeds up producing the part using DDM while at centimeters of steel deposited per hour increased more than
the same time it reduces cost. Thus, our expectation is that five-fold (from 7.2 to 39 cm3/h). An important enabler of this
in the same way as direct printing is incrementally replacing improvement is the doubling of the number of lasers used
offset printing in the typographic industry DDM will start from one to two, which when continued can be expected to
replacing tool-based manufacturing through the combination yield similar improvements also in the future.
of faster and cheaper 3D printers and 3D modeling software Perhaps the most interesting operational strategy from a
that enable lighter but equally durable designs that are cheaper research perspective is the use of DDM for the development
and faster to manufacture. of innovative solutions. The first question for the research
A third question related to the replacement strategy is the agenda is here the potential DDM based innovative solutions.
time frame where DDM becomes a competitive alternative in Proponents of DDM as a revolution point to innovative solu-
different settings. While predicting when DDM will reach tions for improvement of products in use or for highly cus-
levels that are competitive to conventional tool-based tomized and contextualized products (Chen et al. 2015;

Table 5 Estimated cost and days to print the metallic and plastics material parts of a small sedan (using two of the fastest available 3D printers in 2014:
ProX 500 Plus, and SLM 500 HL equipped with four 400 watts lasers)

Material Build volume (cm3) AM machine Print Estimated cost Estimated printing Estimated Cost of printing including
rate (cm3/h) of printing time (days) the machine and raw material (USD)
(USD/cm3)

Steel/iron 135,000 39 ~2.13 144.1 ~287,020

Aluminium alloys 48,000 105 ~0.72 19 ~34,680
Thermoplastics polymers 1,565,000 2,000 ~0.018 32.8 ~27,600
Elastomers 435,000 2,000 ~0.018 9.1 ~7,700
Total - - - 205 ~357,000

Note: For the plastic parts the 3D printing cost calculations are based on the volume of the final parts while for the metal parts the numbers are derived
from the raw volume of the metal content. For all materials the AM machine lifespan assumed to be 5 years
8 Holmström J. et al.

Fig. 2 Estimated days to print the 1000

steel parts of a passenger car on a
single selective laser melting 3D 900
printer. Note: Data is based on 780
manufacturer performance data 800

Time to produce (days)

for the best available 3D printers
700
in 2004, 2010, 2013 and 2014,
respectively, for an assumed 600
volume of steel parts to be printed
of 135,000 cm3 500
401
400

300
225
200 144

100

Year

D’Aveni 2015; Waller et al. 2015). However, from an opera- a rethink of how we run low-volume manufacturing opera-
tions management perspective closer examination will reveal tions, and how one customizes and upgrades manufactured
that many revolutionary solutions require a high level of goods. Some of the most established OM practices, like job
DDM adoption in terms of all parts of a product having been shop scheduling, batch sizing and inventory management will
modeled for DDM, or that all products in a supply chain have change, or may even become redundant. A specific topic mov-
product models that are used for logistics and supply chain ing towards the main stream of OM research is the embedded,
control. This indicates that a reason for the lack of operational or product-centric, ways of controlling operations
innovation strategy in practice is its dependence on evolved (Kärkkäinen et al. 2003). Furthermore, information in the
DDM and operational practice and leads us to the second form of digital product models is set to increasingly replace
question for the research agenda. How evolved needs DDM inventory as the focus of operations and supply chain man-
practice be for different innovative solutions to be possible, i. agement. This shift makes the use of design patterns
e. what are the development and adoption thresholds? With (Främling et al. 2007) and others means to manage the rela-
this type of knowledge innovative operations strategy can be tionships between design models and products in use impor-
developed to focus on contexts where DDM based practice is tant topic on the research agenda. We argue that DDM will
sufficiently evolved to enable specific revolutionary practices. reinvigorate the research on the decentralized control of pro-
duction and handling of products (van Brussel et al. 1998;
McFarlane et al. 2003), and initiate new research streams on
6 Outlook: evolution or revolution? digitalization of supply chain management.
From a supply chain point of view, several new research
The general perception of DDM is one of a disruptive innova- questions follow on from the shift in value-added when sourc-
tion that will revolutionize (read: displace) traditional ing components via DDM, rather than through established
manufacturing and its supply chains by leveraging its unique multi-tier supply chains. First and foremost, it is yet to be
characteristics of enabling local high-variety manufacturing. determined what variables define the trade-offs between
This prediction is based on the conceptual potential of DDM DDM to tool-based manufacturing. Unit cost will surely play
technologies, and bears little resemblance to economic reality. a major role, but so will more strategic considerations related
Instead, the adoption of DDM is likely to be a gradual one to volume flexibility and risk management. From a supply
where manufacturing firms will introduce the new technology chain point of view, what are the implications of collapsing
to improve its current operations, increasingly replacing multi-tier supply chains down to one tier? A widespread adop-
tool-based manufacturing with DDM as economics permit, tion of DDM would drive fundamental structural changes in
therein gradually introducing innovative (and quite possibly, the way in which manufacturing firms organize their supply
revolutionary) practices enabled by DDM’s distinct capabilities. chains; here neither the economic trade-offs that will drive this
While DDM is unlikely to replace batch- or line-based change nor the ramifications for existing manufacturing sup-
manufacturing in the short or medium term, it does requires ply chains are so far understood. Who, for example, will
The direct digital manufacturing (r)evolution 9

capture the value-added in low-volume and aftermarket sup- Bicheno J, Holweg M (2016) The lean toolbox, 5th edn. PICSIE Books,
Buckingham
ply chains if parts can simply be printed close to the point of
Castelvecchi D (2015) Chemical trick speeds up 3D printing. Nature. doi:
use? For example, there will be opportunities for logistics 10.1038/nature.2015.17122
service providers to innovate services combining inventory Chen D, Heyer S, Ibbotson S, Salonitis K, Steingrímsson JG, Thiede S
management and transportation with DDM services (2015) Direct digital manufacturing: definition, evolution, and sus-
(Holmström and Partanen 2014), but what specific capabilities tainability implications. J Clean Prod 107:615–625
D’aveni RA (2013) 3-D printing will change the world. Harvard Bus Rev
will firms need to compete for this business? Much explorato-
91(3):34–35
ry work with early adopters remains to be done here. D’Aveni RA (2015) The 3-D revolution. Harvard Bus Rev 93(5):40–48
The topic of DDM is also of relevance for the theoretical De Jong JP, De Bruijn E (2013) Innovation lessons from 3-D printing.
understanding of changing governance (Williamsson 1996) MIT Sloan Manage Rev 54(2):43–52
and resources of competitive firms (Teece 2007) in a digita- Economist (2013) 3D printing scales up. Economist, Technology
lizing economy. DDM reduces the specific assets needed to Quarterly. http://www.economist.com/news/technology-quarterly/
21584447-digital-manufacturing-there-lot-hype-around-3d-
engage in manufacturing while increasing the business oppor- printing-it-fast. Accessed 7 September 2013
tunity of investing in design. Manufacturing may become a Economist (2014) Smooth operator: a Dutch company has come up with
resource moving outside of the firm (into the Bcloud^) in a a way to print lenses which one day could be used to make prescrip-
way similar to computing. On the other hand resources needed tion glasses on demand. Economist, Technology Quarterly. http://
www.economist.com/news/technology-quarterly/21603237-dutch-
for the control and development of digital product models company-has-come-up-way-print-lenses-which-one-day-could-be-
used over the life-cycle of a product becomes pivotal for cap- used. Accessed 7 June 2014
turing the economic value added. Thus, DDM provides a set- Eyers D, Dotchev K (2010) Technology review for mass customisation
ting to further develop theory on outsourcing (Williamsson using rapid manufacturing. Assembly Autom 30(1):39–46
2008), servitization (Ostrom et al. 2010), and dynamic capa- Financial Times (2014) 3D printing: a powerful technology, but no pan-
acea. Financial Times. http://www.ft.com/cms/s/0/006d60f6-f7a4-
bilities (Teece 2007) in digitalizing supply chains. 11e3-b2cf-00144feabdc0.html#axzz36FWRGpRk. Accessed 24
In summary, DDM is an emerging class of manufacturing June 2014
methods that already offers a wealth of processes and mate- Fitzgerald M (2013) With 3-D printing, the shoe really fits. MIT Sloan
rials that cover many applications across industry sectors. It is Manage Rev blog. http://sloanreview.mit.edu/article/with-3-d-
printing-the-shoe-really-fits/. Accessed 28 June 2014
on a clear trajectory to further reducing printing times and
Främling K, Ala-Risku T, Kärkkäinen M, Holmström J (2007) Design
cost, which will make it increasingly viable in low-volume – patterns for managing product life cycle information. Commun
and eventually also repetitive – manufacturing contexts. The ACM 50:75–79
implications of this shift for operations and supply chain man- Freedman DH (2011) Layer by layer. MIT technology review. http://
agement research are far from understood. We have highlight- www.technologyreview.com/featuredstory/426391/layer-by-layer/.
Accessed 1 July 2014
ed potential changes in practice, as well as emerging research
Gamma E, Helm R, Johnson R, Vlissides J (1994) Design patterns: ele-
questions, yet fully acknowledge that this research agenda is ments of reusable object-oriented software. Pearson Education,
far from exhaustive. The objective of this paper is to provide a Upper Saddle River
nucleus for starting research on the implications of DDM for Gartner (2013) Emerging technologies hype cycle for 2013. Gartner
operations management. As the rate of improvement in DDM Research. http://www.gartner.com/newsroom/id/2575515.
Accessed 19 August 2013
begins to affect practice, it clearly suggests that operations
Gibson I, Rosen DW, Stucker B (2010) Direct digital manufacturing. In:
management research cannot afford not to take notice. Gibson I, Rosen DW, Stucker B (eds) Additive manufacturing tech-
nologies. Springer, New York, pp. 363–384
Holmström J, Partanen J, Tuomi J, Walter M (2010) Rapid manufacturing
in the spare parts supply chain. J Manuf Technol Manage 21(6):
687–697
References Holmström J, Partanen J (2014) Digital manufacturing-driven transfor-
mations of service supply chains for complex products. Supply
Alexander C (1999) The origins of pattern theory: the future of the theory, Chain Manag 19:5–5
and the generation of a living world. Software, IEEE 16(5):71–82 Holmström J, Tenhiälä A, Kärkkäinen M (2011) Item dwell time in pro-
Anderson C (2012) Makers: the new industrial revolution. Crown ject inventories: a field experiment. Comput Ind 62:99–106
Publishing Group, New York Holweg, M (2015) The limits of 3-D printing, Harvard bus rev, online
Arnäs PO, Holmström J, Kalantari J (2013) In-transit services and hybrid forum, June 23
shipment control: the use of smart goods in transportation networks. Hopkinson N, Hague R, Dickens P (2006) Rapid manufacturing: an in-
Transportation Research Part C: Emerging Technologies 36:231–244 dustrial revolution for the digital age. John Wiley & Sons,
ASTM International (2013) Standard Terminology for Additive Chichester, p. 197
Manufacturing Technologies. Designation: F2792-12a Kärkkäinen M, Ala-Risku T, Främling K (2003) The product centric
BAE Systems (2014) 3D printed metal part flown for first time on uk approach: a solution to supply network information management
fighter jet Retrieved January 6, 2014 from http://www.baesystems. problems? Comput Ind 52:147–159
com/article/BAES_164442/3d-printed-metal-part-flown-for-first- Khajavi SH, Partanen J, Holmström J (2014) Additive manufacturing in
time-on-uk-military-jet the spare parts supply chain. Comput Ind 65:50–63
10 Holmström J. et al.

Khajavi SH, Partanen J, Holmström J, Tuomi J (2015) Risk reduction in Munich http://www.rolandberger.com/media/pdf/Roland_Berger_
new product launch: a hybrid approach combining direct digital and Additive_Manufacturing_20131129.pdf. Accessed 14 November
tool-based manufacturing. Comput Ind 74:29–42 2013
Mellor S, Hao L, Zhang D (2014) Additive manufacturing: a framework Rönkkö M, Kärkkäinen M, Holmström J (2007) Benefits of an item-
for implementation. Int J Prod Econ 149:194–201 centric enterprise-data model in logistics services: a case study.
McFarlane D, Sarma S, Chirn JL, Wong C, Ashton K (2003) Auto ID Comput Ind 58:814–822
systems and intelligent manufacturing control. Eng Appl Artif Intell Teece DJ (2007) Explicating dynamic capabilities: the nature and
16:365–376 microfoundations of (sustainable) enterprise performance. Strateg
NASA (2013) NASA tests limits of 3-d printing with powerful rocket Manage J 28(13):1319–1350
engine check http://www.nasa.gov/press/2013/august/nasa-tests- Van Brussel H, Wyns J, Valckenaers P, Bongaerts L, Peeters P (1998)
limits-of-3-d-printing-with-powerful-rocket-engine-check/#. Reference architecture for holonic manufacturing systems: PROSA.
U6QdaPmSx94. Accessed 18 June 2014 Comput Ind 37:255–274
Nielsen NPH, Holmström J (1995) Design for speed: a supply chain Weller C, Kleer R, Piller FT (2015) Economic implications of 3D print-
perspective on design for manufacturability. Comput Integr Manuf ing: market structure models in light of additive manufacturing
8(3):223–228 revisited. Int J Prod Econ 164:43–56
Olhager J (2003) Strategic positioning of the order penetration point. Int J
Wired (2013) 3-D printed car is as strong as steel, half the weight, and
Prod Econ 85(3):319–329
nearing production. Wired. http://www.wired.com/2013/02/3d-
Ostrom AL, Bitner MJ, Brown SW, Burkhard KA, Goul M, Smith-
printed-car/. Accessed 27 February 2013
Daniels V, Demirkan H, Rabinovich E (2010) Moving forward
and making a difference: research priorities for the science of ser- Wohlers TT (2015) Additive manufacturing and 3D printing state of the
vice. J Serv Res-us 13(1):4–36 industry: annual worldwide progress report. Fort Collins, Wohlers
Berger R (2013) Additive manufacturing: a game changer for the Associates
manufacturing industry? Roland Berger Strategy Consultants,
Information Systems Frontiers
https://doi.org/10.1007/s10796-020-10089-2

Shaping Digital Innovation Via Digital-related Capabilities

Mina Nasiri 1 & Minna Saunila 1 & Juhani Ukko 1 & Tero Rantala 1 & Hannu Rantanen 1

Accepted: 3 November 2020

# The Author(s) 2020

Abstract
The aim of this research was to explore the conditions under which digital innovation opportunities emerge in small and medium-
sized enterprises (SMEs). The research answered the question of what capabilities are required to shape the exploitation of digital
innovation, namely market offerings and the digital business process. To address the research question with a quantitative
research method, data were collected through survey questionnaires distributed among 280 SMEs operating in the service and
manufacturing industries in Finland. The results revealed that among four digital-related capabilities—namely human, collabo-
ration, technical, and innovation capabilities—human, technical, and innovation capabilities contribute to market offerings, while
human, collaboration, and technical capabilities contribute to the business process.

Keywords Digital innovation . Digital-related capabilities . Digital transformation . Market offerings . Digital business process

1 Introduction driven insight, as well as develop capabilities to implement

these recognized possibilities. In other words, companies need
As digital transformation is shaping the business environ- to find ways to exploit digital innovation. Therefore, the dom-
ments of contemporary companies, companies need the adjust inant role of digital innovation is being increasingly observed
their operations and find ways to respond to the changes. To in the business environment due to new opportunities that
operate in increasingly digitalized business environments, and offer firms the potential to broaden into new and special ex-
as a part of the digital ecosystems (Delgosha et al. 2020; periences (Jahanmir and Cavadas 2018; Kolloch and
Pappas et al. 2018), companies need to not only find innova- Dellermann 2018; Nylén and Holmström 2015), leading to
tions and innovative ways to change their businesses but also high profit potential and customer satisfaction (Bednar and
develop capabilities to exploit innovation in changing sur- Welch 2019; Parida et al. 2015). The origins of digital inno-
roundings. Further, Mikalef et al. (2020), argued that to derive vation can be traced to digitalization (Chan et al. 2019), which
value from the growing opportunities of digitalization, such as acts as an enabler of and creates essential conditions for
big data, companies need to develop the organizational capac- exploiting digital innovation (Yoo et al. 2010). Such exploi-
ity to recognize how their businesses can benefit from data- tation requires capabilities, relevant to digitalization, that re-
spond appropriately to market opportunities and digital tran-
sitions. Further, in responding to digital transitions, align-
* Mina Nasiri ments between organizational capabilities and market de-
[email protected] mands must be created, as misalignments in this dynamic
can lead to business failure (Chan et al. 2019).
Minna Saunila
[email protected] Pappas et al. (2018) argued that to reach digital transfor-
mation and the creation of sustainable societies, none of the
Juhani Ukko
operators in society should be seen in isolation; instead, there
[email protected]
is a need to improve the understanding of how their interac-
Tero Rantala tions lead to knowledge, innovation, and value creation.
[email protected]
Digitalization provides many ideal opportunities, particularly
Hannu Rantanen for small- and medium-sized enterprises (SMEs), and the in-
[email protected] terplay between companies’ different types of capabilities and
1 innovation in digitalizing business is of growing interest to
School of Engineering Science, Department of Industrial
Engineering and Management, LUT University, Mukkulankatu 19, both academics and practitioners (Bednar and Welch 2019;
15210 Lahti, Finland Mikalef et al. 2018). For example, Mikalef et al. (2019)
 Inf Syst Front

examined the indirect relationship between a Big Data analyt- lacking. This is referred to as the process perspective on digital
ics capability and two types of innovation capabilities— innovation, and it includes the actions and outcomes of digital
incremental and radical. As a part of the companies’ innova- innovation (Kohli and Melville 2019). Traditionally, the main
tion activities in digitalizing business environments, such as focus of innovation management research has been on either
those related to Big Data analytics or business ecosystem ac- innovation development actions or innovation outcomes (i.e.,
tivities, it is important to develop theory and conduct practical Ahmad et al. 2013; Sivasubramaniam et al. 2012). Digital
research that will incorporate the phenomenon called digital innovation involves both actions and outcomes, as suggested
innovation (Nambisan et al. 2017). In the big picture, digital previously (i.e., Lee and Berente 2012; Nambisan et al. 2017).
transformation requires companies of all sizes to rethink Few studies have concentrated on the firm-level capabilities
and innovate their businesses, yet SMEs may have little needed to identify, assimilate, and apply valuable knowledge
time and few resources for experimenting with their busi- from both inside and outside the firm with regard to opportu-
nesses or for implementing new strategies (Bouwman nities for digital innovation, known as initiate activity (Kohli
et al. 2019). Recent research has investigated how SMEs and Melville 2019). Therefore, the scope of this study was
with inadequate capabilities and limited resources drive formed on the following basis: innovation is considered to
digital transformation (Li et al. 2018) and has further cover both the digital-related capabilities that comprise the
demonstrated the lack of understanding of the capabilities basis of the initiate activity and digital innovation as an
of SMEs with respect to different aspects of digital trans- outcome.
formation. As such, the topic has gained interest among To address the abovementioned research gap, this study
academics. For example, Cenamor et al. (2019) examined utilizes a quantitative research method. The aim of the study
the effect of digital platform capability and network is to explore the conditions under which digital innovation
capability on the financial performance of opportunities emerge in SMEs. The following research ques-
entrepreneurial SMEs, while Neirotti et al. (2018) ex- tion was addressed: What capabilities are required to shape the
plored how SMEs develop ICT-based capabilities in re- exploitation of digital innovation? To answer this research
sponse to their environment. Overall, few studies have question, a structured survey questionnaire was administered
concentrated on the firm-level capabilities needed for dig- to 280 SMEs operating in the service and manufacturing in-
ital innovation (Kohli and Melville 2019). This study dustries in Finland. First, this research contributes to the pro-
sheds light on exploiting digital innovation by identifying cess perspective of digital innovation, considering digital in-
the required capabilities in digital transformation and un- novation as an outcome and the capabilities needed to create
derstanding the way of utilizing those capabilities. In this it. Further, by incorporating the effects of digital-related capa-
study, digital-related capabilities were defined as organi- bilities based on the digital innovation type, the research ex-
zational capabilities in terms of multidimensional con- tends the digital innovation literature. The research focused on
structs, such as human (Chan et al. 2019; El Sawy et al. SMEs because, despite the perception that SMEs are frequent-
2016; Kane et al. 2015; Legner et al. 2017), collaboration ly hampered by a lack of resources and capabilities, the prob-
(Amit and Han 2017; Chuang and Lin 2015; El Sawy ability of successfully exploiting digital innovation was higher
et al. 2016; Pagani and Pardo 2017; Sjödin et al. 2016), for SMEs than it was for large and well-established companies
technical (El Sawy et al. 2016; Parida et al. 2015; Sjödin (Street et al. 2017). This may be because SMEs have the
et al. 2016; Yoo et al. 2010), and innovation (Parida et al. capability to move quickly and easily, whereas large compa-
2015; Sjödin et al. 2016; Sousa and Rocha 2019; Xue nies are frequently unwilling to adopt digital innovations be-
2014) capabilities. These capabilities enable companies cause of the risk of losing their current competitive advan-
to respond quickly to digital transformation and to thereby tages. Moreover, organizations that are mature and well-
exploit digital innovation (Fichman et al. 2014; Kohli and developed are marked by institutionalized processes, capabil-
Mekville 2019; Nambisan et al. 2017; Nylén and ities, and cultures that, in this case, may hinder their ability to
Holmström 2015; Urueña et al. 2016). Successful digital respond to digital transitions (Chan et al. 2019).
innovation greatly depends on how capabilities are under- The rest of the paper is structured as follows. Next, the
stood and, subsequently, how such capabilities are theoretical framework for the study, including its theoretical
adapted with regard to innovation outcomes, processes, underpinnings and key concepts, is presented. After this, the
and related markets (Nambisan et al. 2017; Nylén and research model used in the study is described, including a
Holmström 2015). Moreover, digital innovation is incor- discussion of the research hypotheses. Afterward, the research
porated into the continuous matching of digital-related methodology is presented, followed by a discussion of the
capabilities with traditional market offerings (Nambisan research results. Lastly, the conclusions of the study are
et al. 2017). summarized.
However, research on the types of capabilities required to
shape the evolution of the digital innovation phenomenon is
Inf Syst Front

2 Theoretical Framework capabilities based on dynamic and fluid perspectives

(Lyytinen et al. 2016; Yoo et al. 2010). This means that capa-
2.1 Dynamic Capabilities and Digital Innovation bilities once considered practical may now be impractical or
even damaging for the new, emerging condition (Chan et al.
Innovation is the creation and adoption of a practice or object 2019). As a result, there are demands for developed and com-
considered novel (Rogers 1983). Creating innovation is bined capabilities that will permit companies to offer modern
achieved through distinct resources. Resources can be either service offerings. These combined capabilities include service
tangible, such as humans and technology, or intangible, such development capabilities (i.e., developing new offerings and
as knowledge (Davila et al. 2012). These resources ensure the solutions), network management capabilities (i.e., knowledge
efficient execution of innovative routines within firms. As one sharing with the right partners), and digitalization capabilities
of the driving forces of innovation creation, digital transfor- in terms of technical abilities (i.e., integrated and smart sys-
mation not only provides novel opportunities for companies tems), which play a key role in the provision of advanced
but also necessitates an understanding of what capabilities are market offerings (El-Haddadeh 2020; Sjödin et al. 2016).
required and how those capabilities should be adapted with Thus, in this study, digital-related capabilities were defined
regard to innovation outcomes, processes, and related markets as organizational capabilities in terms of multidimensional
(Kohli and Melvill 2019; Nambisan et al. 2017). Current re- constructs, such as human, collaboration, technical, and inno-
search has stated that firms use dynamic capabilities to adapt, vation capabilities, for moving toward digitalization. These
integrate, or reconfigure existing resources and skills to read- digital-related capabilities help companies to respond quickly
just to environmental change (Teece et al. 1997). The dynamic to digital transformation and to exploit digital innovations
capabilities theory deals with the unique, difficult-to-replicate (Kohli and Mekville 2019; Sjödin et al. 2016; Urueña et al.
capabilities that allow rapid adaptability to changes in the 2016).
external environment (Teece 2014; Teece et al. 1997). Thus, Referring to human capabilities, as the source of digital
this theory is well-suited to examine changes caused by digital innovation, digitalization requires employee support, readi-
transformation. Dynamic capabilities reside in “interrelated ness, and digital know-how (Chan et al. 2019; El Sawy et al.
routines within firms for performing specific tasks” (Ngo 2016; Kane et al. 2015; Legner et al. 2017). Companies that
and O’Cass 2013, p. 1135). To classify exisiting capabilities, lack digitally skilled employees and/or staff skilled at problem
research on the routines and operations used by firms for de- solving suffer from deficient capabilities because of the com-
veloping digital innovation is needed. Thus, the following plexity typifying the current digital era (Kache and Seuring
sections review the characteristics of digital innovation as well 2017; Lerch and Gotsch 2015). Thus, in this study, digital-
as the related capabilities presented in prior research. related human capabilities can be defined as the set of capa-
Digital innovation has radically altered the structure of new bilities, knowledge, and skills that employees need in order to
products, services, and business processes, leading to novel move toward digitalization.
value creation and competitive advantages for companies Because of the complex nature of digitalization, achieving
(Fichman et al. 2014; Nambisan et al. 2017). Greater exploi- competitive advantages through single actors is not possible
tation of digital innovation requires bundles of new and com- (Canhoto et al. 2016; Kohli and Mekville 2019; Pagani and
plementary capabilities (Fichman et al. 2014; Lusch and Pardo 2017). Therefore, it is necessary to acquire collabora-
Nambisan 2015; Nambisan et al. 2017; Nylén and tion capabilities in order to sustain a viable alliance and build
Holmström 2015). Digital innovation can be explained as value networks with the right partners (Amit and Han 2017; El
the creation of market offerings and business processes as an Sawy et al. 2016; Pagani and Pardo 2017; Sjödin et al. 2016).
outcome of using digital technologies (Nambisan et al. 2017). Since digitalization has changed the structure of social rela-
Consequently, the definition of digital innovation includes tionships, in both internal and external company spaces
two noteworthy and concurrent phenomena—namely, (Pagani and Pardo 2017), SMEs must seek opportunities for
products/services and business processes (i.e., the way of do- collaborating with partners, thereby complementing their cur-
ing things in an organizational setting)—both of which are rent capabilities with partners’ capabilities (Chan et al. 2019).
enabled by digitalization (Fichman et al. 2014; Nambisan Collaboration capabilities enable learning via the exchange
et al. 2017; Nylén and Holmström 2015). and sharing of knowledge and experience through digital
channels, which are facilitated by digitalization (Chuang and
2.2 Digital-related Capabilities Lin 2015; Legner et al. 2017; Maravilhas and Martins 2019).
Therefore, in this study, digital-related collaboration capabil-
In the current digital era, digitalization is blurring all industry ities are among the capabilities necessary to move toward
boundaries, and yet fixed and bounded strategies continue to digitalization.
be applied, consequently limiting the full exploitation of dig- In the current digital era, the utilization of both internal and
ital innovations. Thus, it is necessary to develop strategies and external organizational cooperation requires technical
 Inf Syst Front

capabilities as a complementary channel for ensuring the pro- hypotheses were developed and will be discussed in the fol-
vision of advance market offerings (Legner et al. 2017; Sjödin lowing section. In addition, to understand the mechanisms
et al. 2016). Thus, these capabilities play a critical role in the behind the eight direct hypotheses, this paper examined
use of digitalization for integrating products and services (El whether digital-related capabilities indirectly affect digital in-
Sawy et al. 2016) and for accessing updated and borderless novation. This analysis provides a basis for further research
services and activities (Parida et al. 2015; Sjödin et al. 2016; related to hypothesizing and testing mediation and moderation
Yoo et al. 2010). Hence, in this study, digital-related technical effects.
capabilities refer to the bundle of technical capabilities that
facilitate the implementation of digitalization in companies. 3.2 Hypotheses
Disruptive digital business, which emerges from digitaliza-
tion, requires innovation capabilities to be managed success- 3.2.1 Effects of Digital-related Capabilities on Market
fully (Sousa and Rocha 2019). The capacities to innovate, Offerings
identify, and exploit business opportunities, as well as to di-
versify the business area, are among the innovation skills Digital transformation in companies and contemporary digital
needed for disruptive digital business (Sousa and Rocha solutions can reduce the development time needed for gener-
2019). Moreover, the ability to develop new ideas, solutions, ating innovations and launching them on the market (Marion
and novel offerings are among the most important capabilities et al. 2015). As such, most recent digital solutions and digital
for companies to possess in the current digital era (Parida et al. technologies have been integrated into companies’ product
2015; Sjödin et al. 2016; Xue 2014). Thus, digital-related and service development, thereby affecting their market offer-
innovation capabilities are required for companies to advance ings (Nylen and Holmström 2015; Yoo et al. 2012). Although
toward digitalization. digital transformation provides possibilities for automatizing
In sum, in responding to the exploitation of digital innova- products and making services smarter, human senses and
tion, organizations frequently require development capabili- capabilities still form an important part of successful digital
ties, which correspond closely to new situations and dynamic innovation. Nylen and Holmström (2015) argued that digital
changes. In addition, organizations simultaneously require the services and solutions must not only be efficient to learn and
mitigation of inflexibility in order to better equip themselves easy to use but must also consider user experience. To com-
in the face of transformations (Chan et al. 2019; Kohli and prehend the usability and user experience of digital innova-
Melville 2019; Parida et al. 2015; Sjödin et al. 2016). Thus, tion, companies must understand their customers’ needs as
with regard to digital-related capabilities, human, collabora- well as the possibilities entailed by the utilization of developed
tion, technical, and innovation capabilities are all recommend- products and services. The gathering of feedback and the un-
ed for companies to increase their odds of survival in the derstanding of customers’ needs require human senses and
current competitive environment. digital-related human capabilities. Hoe (2017) claimed that
the main argument for these thinking skills in digital business
environments is putting end users’ and customers’ wishes and
3 Research Model and Hypothesis needs first and developing digital innovations to fulfill them.
Development In other words, employees need to understand the interplay
between the possibilities created by digital transformation and
3.1 Research Model the wishes of digital innovation users.
According to Porter and Heppelmann (2015), making
This study aimed to explore the conditions under which digital products and services smarter widens opportunities for com-
innovation opportunities emerge in SMEs. Further, this re- panies’ capabilities to, for example, monitor products and ser-
search sought to contribute to the process perspective of dig- vices and personalize product functioning. Human capabilities
ital innovation, considering digital innovation as an outcome are needed to integrate organizations’ core competitive
and the capabilities needed to create it. Figure 1 depicts the advantages and knowledge with these widened opportunities
proposed research model, which postulates a number of direct to generate successful digital innovations that can leverage
linkages between digital-related capabilities (namely human, market offerings. As Fuchs and Sevignani (2013) suggested,
collaboration, technical, and innovation capabilities) and dig- the creation of new products in the digital work context also
ital innovation (namely market offerings and the business requires human brains and human experiences. According to
process). Nylen and Holmströn (2015), the realization of digital inno-
vation requires new skills, and companies must develop their
To better understand the effects of digital-related capabili- mechanisms for supporting the advancement of digital-related
ties on the exploitation of digital innovation, specifically in human capabilities and continuous learning in digitalizing op-
market offerings and business processes, eight direct erating environments. As such, companies must ensure that
Inf Syst Front

Fig. 1 Research model and

Digital-related capabilities Digital innovation
hypotheses
H1 & H5
Human capabilities
Market offerings
H2 & H6
Collaboration capabilities

H3 & H7
Technical capabilities
Business process
H4 & H8
Innovation capabilities

their employees are trained in the use and acceptance of digital Pardo 2017; Peppard and Rylander 2006). According to
solutions. Based on these arguments, the following hypothesis Maravilhas and Martins (2019), the exchange and sharing
was proposed: of knowledge, information, and experiences among users
H1: Human capabilities relate positively to digital via digital channels stimulate innovation, and this is
market offerings. called collaborative innovation. Digital channels provide
an opportunity for aggregating product demands, resulting
The idea of creating innovative products and services via in an increase in the diversity of the developed market as
collaboration has gained significant scholarly attention and well as more diverse product and service offerings
has become the main focus of many companies (Luo et al. (Brynjolfsson et al. 2010). Moreover, shaping the nature
2010; Santoro et al. 2018). For instance, a joint venture be- of digital innovation as collective actions requires differ-
tween IBM, Sony, and Toshiba facilitated the development of ent collaboration capabilities among partners, which are
advanced chips for consumer electronics, culminating in a cell enabled by digitalization (Lyytinen et al. 2016; Nambisan
processor that supplied power for Sony’s PlayStation 3 video et al. 2017; Yoo et al. 2012). Therefore, building on the
game console. Masashi Muromachi, chief executive officer literature discussed above, the following hypothesis was
(CEO) of Toshiba, called this “a winning combination” (Luo proposed:
et al. 2010, p. 245). Furthermore, many different researchers H2: Collaboration capabilities relate positively to dig-
have considered collaborative activities as a way to stimulate ital market offerings.
digital innovation (Fichman et al. 2014; Maravilhas and
Martins 2019). Fichman and colleagues (2014) used the term Technical capabilities are presented as main enablers of
“network effects” as a label for capturing the tendencies of organizational capabilities toward digital innovation (Banker
digital innovation and for addressing the relative ability to et al. 2006; Mithas et al. 2011; Setia et al. 2013; Tanriverdi
communicate or share digital assets among users. 2005) because digital technology creates a bond between the
Furthermore, Lyytinen and colleagues (2016) referred to net- physical and digital characteristics of products and services
works as a catalyst for the faster expansion of digital product (El Sawy et al. 2016). At present, digital technology is more
innovation via borderless access to digital tools. In their re- affordable and ubiquitous than ever before, which has in turn
search, they mentioned that, in the current digital era, there is a facilitated more engagement with digital innovation, thereby
need for networks to collect knowledge flows around new allowing new configurations of actors to develop, generate,
products and services alongside digital tools for the further and invest in new digital products and services (Nylén and
development of these products and services in order to suc- Holmström 2015; Yoo et al. 2010). Digital technology has
cessfully pursue radical innovations (Lyytinen et al. 2016). made it technically possible to upgrade the functionality of
Hence, product innovation can be achieved via both internal products and services due to the reprogrammable nature of
development activities and external collaboration (Hull and these capabilities. Thus, the technical capabilities concerning
Covin 2010; Luo et al. 2010; Lyytinen et al. 2016). digital technologies are related to value propositions of digital
Because of the complex nature of market offerings in products and services, addressing how value is generated and
the digital era, providing successful market offerings with captured in each digital product and service (Nylén and
a single actor does not work (Canhoto et al. 2016; Kohli Holmström 2015).
and Mekville 2019; Lyytinen et al. 2016; Pagani and The technical capabilities of digitalization has permit-
Pardo 2017). Therefore, adopting collaboration perspec- ted activities to cross the boundaries of time, place, and
tive has been mentioned as a practical option for compa- function (Parida et al. 2015; Yoo 2010; Yoo et al. 2012;
nies involved with digitalized supply chains (Pagani and Nylén and Holmström 2015) demonstrated that digital
 Inf Syst Front

technology provides the potential for a reinvention of 3.2.2 Effects of Digital-related Capabilities On the Business
sales and distribution channels. Furthermore, they noted Process
that firms are now able to carefully place and integrate
their products and services with an abundance of mobile The increased digitalization of companies’ operating environ-
operating systems, social media sites, and app stores. In ments provides possibilities for digital innovation to occur. As
sum, this paper suggests that by keeping up to date with related to business processes, real-time activities and automat-
technical capabilities (e.g., analyzing the progress of ed production processes (Fichman et al. 2014), for example,
digital technology and associated usage patterns; Nylén continuously affect the approach to development and innova-
and Holmström 2015), firms can develop their digital tion. New types of digital solutions provide broadened oppor-
market offerings in terms of novel digital products and tunities for data gathering, and different types of artificial in-
services. In line with the considerations above, the follow- telligence solutions can assist companies and their employees
ing hypothesis was proposed: and managers in decision making (Jarrahi 2018). This devel-
H3: Technical capabilities relate positively to digital opment of digital solutions creates circumstances in and by
market offerings. which different groups of people have increased options for
generating digital innovations and affecting the business pro-
Advancing innovation potential among companies is cesses of companies. However, although developed digital
crucial for digital innovation (Sia et al. 2016; Sousa and solutions can provide enhanced data and support for develop-
Rocha 2019) studied digital innovation in terms of disrup- ment, analysis, and decision making, humans remain solely
tive technological phenomena, such as mobile technolo- responsible for making decisions and generating digital inno-
gies, artificial intelligence, big data, and robotics, and vations. As such, the capacity to effectively apply digital so-
they found that the exploitation of such phenomena lutions and take advantage of the opportunities these solutions
requires various innovation skills. These skills include permit for developing business processes depends on compa-
the capability for creativity, the recognition of novel nies possessing the requisite digital-related human capabili-
business openings, and the arrangement of indispensable ties. For this reason, companies must develop a culture and
resources corresponding to these openings. Similarly, atmosphere that supports and promotes the development and
Fichman et al. (2014) stated that digital innovation re- utilization of their employees’ digital skills.
quires an understanding of what has become possible Richter et al. (2018) further highlighted that, in the context
due to advances in technology, as well as the exploitation of digital work design, companies must cultivate the under-
of this understanding to create something valuable for the standing of which work practices allow work to be more au-
company or for society. The development of digital inno- tonomous while at the same time allowing employees to learn
vation also depends on the relative capability to balance and collaborate with each other. Digital-related human capa-
between present and necessary capabilities (Svahn et al. bilities ensure that the increased digitalization within compa-
2017). Nylen and Holmström (2015) believed that this nies can be easily accepted by employees. If the adopted dig-
was due to the unique features of the processes related ital solutions cannot be properly used and, for example, the
to digital innovation. They also suggested that companies collected data are not properly understood, then digital inno-
should question their existing product and service portfo- vations cannot be generated. While real-time activities, artifi-
lios, digital surroundings, and traits used to promote dig- cial intelligence solutions, and automated production process-
ital innovation. Exploiting novel opportunities for innova- es assist and support companies, they cannot effectively re-
tion is key to generating a compiled value by creating spond to common-sense situations (Guszcza et al. 2017;
digital products and services (Nylen and Holmström Jarrahi et al. 2018) or to the entire structure of the digital
2015), thereby casting innovation capabilities as crucial business process. Different types of digital solutions can assist
proficiencies for developing digital market offerings. in different aspects of the business process, but the combina-
Concerning digital market offerings, an advantage may tion of information derived from these aspects requires a com-
be a company’s capability to generate and select produc- mon understanding of the digital business process, especially
tive ideas but also to have the requisite processes in place under uncertain conditions. As such, human capabilities and
to advance novel products and services, coupled with a senses are still needed to correctly interpret the work generat-
willingness to develop new solutions. Such a digital- ed by digital solutions (Brynjolfsson and McAfee 2012) and
related innovation capability may be difficult for compet- to convert such interpretations into digital innovations, and
itors to imitate, potentially providing an initiation to dig- companies must therefore ensure that their employees are well
ital market offerings. Thus, the following hypothesis was trained in using digital solutions. Based on these arguments,
formulated: the following hypothesis was developed:
H4: Innovation capabilities relate positively to digital H5: Human capabilities relate positively to the digital
market offerings. business process.
Inf Syst Front

Digital-related capabilities, in terms of collaborations like (Kallinikos et al. 2013; Langlois 2003; Merrifield et al.
knowledge sharing, social media, digital makerspaces, and 2008; Yoo et al. 2010). All of this is possible via digital tech-
relationships in virtual worlds, have enabled digital innova- nologies that enable physical products to be programmable,
tions in business processes (Nambisan et al. 2017). In addi- addressable, sensible, communicable, memorable, traceable,
tion, due to the changes that digitalization has catalyzed in the and associable (Yoo 2010). These advances in digital technol-
implementation of business processes, it is necessary to recon- ogies, in turn, permit developments in digital infrastructures
sider cooperation among companies (Pagani and Pardo 2017). toward the support of innovations in terms of figurative flex-
For instance, using three-dimensional constructs as digital ibility, semantic coherence, temporal and spatial traceability,
tools in the construction process has created many different, knowledge brokering, and linguistic calibration. (Lyytinen
unexpected collaborations and interactions between different et al. 2016). These types of technical capabilities allow the
partners, designers, and trades, which has in turn expanded successful generation of novel information technology (IT)-
innovation (Boland et al. 2007) while underscoring the need enabled products, services, and processes, whereby the pro-
for collaboration capabilities. Therefore, advanced digital cess innovation outcomes refer to process redesign and sim-
channels are required to encourage collaborative activities in plification (productivity; Kohli and Melville 2019).
the business process. Furthermore, different researchers have Embedding digital technologies in the operations and produc-
reconsidered digital tools as a facilitator of collaboration tions of companies also allows online activities, both within
among business partners, which can be a challenging pursuit, the company and with the company’s customers (Chen et al.
one that often requires adaptations and new business models 2015; Yoo et al. 2012).
(Foltean et al. 2019; Nath et al. 2010). Digital technologies are increasingly appearing in the in-
Despite the potential that social media, as digital channels, dustrial manufacturing context (Nylén and Holmström 2015).
have for transforming business processes (Foltean et al. 2019), Instead of scheduled servicing, embedded digital capabilities
most marketers have not fully distinguished them as either allow the utilization of service forecasting and real-time mon-
opportunities or threats with respect to the creation of new itoring (Nylén and Holmström 2015; Westergren and
business (Cortez and Johnston 2017). For instance, Holmström 2012). For example, with direct digital
Kietzmann and colleagues (2011) identified a lack of knowl- manufacturing (an interconnection between modern ICT and
edge and/or skills with regard to the adoption of social media additive manufacturing equipment), it is possible to reconcile
and the integration of related strategies in the business process supply capacities and consumer demands in real time (Chen
(Kietzmann et al. 2011). As mentioned by Luo and colleagues et al. 2015). Furthermore, the different forms of direct digital
(2010), despite the appeal of sharing knowledge, capabilities, manufacturing include the potential for modifying material
and resources among partners, including external partners in efficiency in product business models and process chains,
the product development process often leads to greater coor- and even in the product–user relationship (Chen et al. 2015).
dination costs and integration difficulties. Thus, a successful In sum, the current research suggests that the technical capa-
digital business strategy requires effective cooperation among bilities of digital technologies affect digital business processes
companies in terms of the product, process, and service do- in terms of, for instance, online activities and advanced
mains, thereby rendering the enhancement of the exploitation manufacturing efficiency (Chen et al. 2015; Nylén and
of digital innovation more complex (Iansiti and Lakhani Holmström 2015; Yoo et al. 2012). In line with these consid-
2014). In addition to these issues, it is worth mentioning that erations, the following hypothesis was proposed:
the results achieved from cooperation in innovation strategies H7: Technical capabilities relate positively to the dig-
are not always positive due to involuntary knowledge spill- ital business process.
overs, different learning speeds, divergent attitudes
concerning the ultimate goal, and lack of flexibility and adapt- Responding to both the threats and opportunities posed by
ability (Faems et al. 2005). Therefore, based on the digitalization by initiating digital business processes also re-
abovementioned literature, the following hypothesis was quires new capabilities in the area of innovation (Henfridsson
formulated: and Yoo 2014; Sia et al. 2016). Since rapid changes in the
H6: Collaboration capabilities relate positively to the business environment make innovation with digital technolo-
digital business process. gies more demanding, various areas of business must be more
organized (Lokuge et al. 2019) called this engaging in an
It has been suggested that digital ecosystems include arti- innovation-savvy culture that contributes to digital innova-
facts and operations that increasingly derive utility from the tion. Thus, this type of innovation capability, which assists
functional relations they maintain (Kallinikos et al. 2013), as companies in sharing information (Lokuge et al. 2019) that
demonstrated by the growing prospects of combining soft- corresponds to the opportunities and threats caused by digita-
ware and software constituents and mixing content across lization (Sia et al. 2016) and in applying digital technologies
platforms, infrastructures, and production systems most relevant for their business (Nylen and Holmström 2015),
 Inf Syst Front

becomes crucial. Further, digital innovation requires a cul- Table 1 Demographic information of the respondents (N = 280)
ture that permits improvisational efforts throughout the com- Characteristics Number of enterprises Percentage (%)
pany while simultaneously recognizing novel opportunities
for innovation (Nylen and Holmström 2015). Age (years since established) 138 49.29
With regard to digital business processes, innovation ca- fewer than 29 142 50.71
more than 30
pabilities are also a key mechanism by which companies can Number of employees 197 70.35
exploit digital innovation. Digital business processes usually Small (fewer than 49) 78 27.85
require companies to modify their current production or gen- Medium (49–250) 5 0.8
eral business processes to enhance their functional efficien- No response
Sector 160 57.14
cy. Therefore, it was expected that a company’s innovation Service 118 42.15
capability, i.e., its capability to generate and select ideas, Manufacturing 2 0.71
advance novel products and services, and develop new so- No response
lutions, is a crucial means by which it can exploit digital
innovation. Thus, the following hypothesis was formulated:
H8: Innovation capabilities positively relate to the dig- point Likert type scale using multiple items, where a response
ital business process. of 1 indicated “strongly disagree” and 7 indicated “strongly
agree.” Due to the lack of empirical research on digital-related
capabilities and digital innovation, specific scales had to be
created based on previous literature, and new items were de-
4 Methodology veloped to measure both digital-related capabilities and digital
innovation. To assess digital-related capabilities, four different
4.1 Sample and Data Collection digital-related capabilities, including human (El Sawy et al.
2016; Lerch and Gotsch 2015), collaboration (Amit and Han
This study used managers’ perceptions as reported in data 2017; Chuang and Lin 2015; El Sawy et al. 2016), technical
collected via a survey questionnaire of 280 SMEs operating (El Sawy et al. 2016; Parida et al. 2015; Xue 2014), and
in the service and manufacturing industries in Finland. As a innovation (Parida et al. 2015; Xue 2014) capabilities were
sampling frame, the list from the random sampling of 6,816 developed as independent variables. Digital-related human
Finnish SMEs was used; in 986 cases, the contact information capabilities consisted of three items addressing questions re-
was invalid, and thus the survey was ultimately conducted lated to companies’ supportive and encouraging attitude to-
with 5,830 SMEs—30% of the 20,0000 SMEs in Finland. ward the development of digital skills, as well as to em-
Then, an invitation letter containing a direct link to the survey ployees’ level of training in digital tool usage and readiness
and a cover letter that described the purpose of the survey was in the digitalization of the operating environment (El Sawy
sent by email to the managers of the Finnish SMEs, asking et al. 2016; Lerch and Gotsch 2015). To measure digital-
them to participate in the survey questionnaire. Four re- related collaboration capabilities, three items were defined to
minders were sent during the one-month data collection pro- resolve questions related to the existence of digital coopera-
cedure, and ultimately a total of 280 valid responses was ob- tion with other companies, the utilization of digital channels to
tained, which is more than the minimum sample size sug- share information with other companies, and the extent to
gested by Barlett et al. for certain populations (2001). which digitalization has transformed the shape of social rela-
In terms of the respondents’ characteristics, firm experi- tionships in their business (Amit and Han 2017; Chuang and
ence ranged from 2 to 123 years, with an average of 35.85 Lin 2015; El Sawy et al. 2016). For measuring digital-related
years, demonstrating that most of the companies were relative- technical capabilities, four items were included; in each case,
ly mature in their fields. Approximately 70% of the companies digital technology was considered as an enabler, enhancing
in the sample were small enterprises with fewer than 49 em- the value of both products and services; integrating products
ployees, while the rest were medium-size enterprises. As men- and services; working across boundaries of time, places, or
tioned above, all the sampled companies operate in the service activities; and providing up-to-date and borderless services
(about 57%) or manufacturing (about 42%) industries. Table 1 (El Sawy et al. 2016; Parida et al. 2015; Xue 2014). To assess
outlines the demographic information of the respondents. digital-related innovation capabilities, three items were devel-
oped, each addressing how digitalization enables innovations
and new ideas, the development of new solutions, and the
4.2 Construct Operationalization production of new products and services (Parida et al. 2015;
Sia et al. 2016; Xue 2014). For measuring digital innovation,
The constructs were conceptualized as reflective measures. two noteworthy and concurrent phenomena, market offerings
Those involved in the research model were measured on a 7- and business processes, were developed as dependent
Inf Syst Front

variables. Market offerings included two items addressing not at risk of common-method bias, late-respondent bias, or
questions about whether digitalization can be defined as dig- lack of validity and reliability.
ital services and digital products (Nambisan et al. 2017; Nylén Prior to the hypothesis tests, the reliability and conver-
and Holmström 2015). To measure the business process, two gent and discriminant validity of the constructs were test-
items were developed, each posing questions about whether ed with confirmatory factor analysis (CFA) and different
digitalization should be defined as automating production pro- indices, including loadings, Cronbach’ alpha, average var-
cesses and real-time activities (Fichman et al. 2014). A com- iance extracted (AVE), and composite reliability (CR), as
plete list of items is presented in Appendix. shown in Table 2. First, Cronbach’s alpha was used to
Three control variables were used to reduce the likeli- assess the items’ reliability. Although a value of 0.7 or
hood of confounded results occurring due to differences 0.6 is recommended as the acceptable cutoff for
in company age and size, as well as in the type of industry Cronbach’s alpha (Taber 2018) in both exploratory re-
(Bstieler 2005). In this research, firm size was measured search (Boyer and Pagell 2000) and in the development
by the number of employees, firm age was determined of new constructs (Flynn et al. 1990; Nunnally 1978)
according to when the company was established, and type mentioned that a smaller value is permissible if the scales
of industry was measured by asking respondents if their are new and contain a small number of items. As shown
companies operated in either service or manufacturing. in Table 2, all the constructs except market offerings had
values greater than 0.6, illustrating the reliability of the
4.3 Data Analysis, Validity, and Reliability constructs. Principal component analysis (PCA) is loaded
in one factor with an eigenvalue greater than 1; this sug-
To confirm the reliability and validity of the data, and to re- gested for items representing a single unidimensional con-
duce both common method bias and non-respondent bias, struct, and if the constructs have the factor loadings great-
different approaches and statistical tests were used at each step er than 0.4, they are shown to be valid (Carmines and
of data collection and analysis, as the necessity of these ap- Zeller 1979). Convergent validity was verified by the val-
proaches and tests has been suggested by Flynn et al. (1990). ue of the factor loadings, AVE and CR. As suggested by
Regarding pre-tests, in an initial step, iterative sessions with Fornell and Larcker (1981), the convergent validity of the
expert researchers were held while designing the survey ques- construct is still adequate for an AVE less than 0.5 if the
tionnaire, and the measured items were determined based on CR is higher than 0.6. As shown in Table 2, each loading
the theoretical foundation provided by previous research. had a value of more than 0.4, and the value of the AVE
Moreover, the items were constructed in such a way that the for all the constructs except business process (AVE =
risk of recognition of a cause-and-effect relationship between 0.482) and market offerings (AVE = 0.428) was more than
the dependent and independent variables by the respondents 0.5, which, in this case, because the CR values were more
was extremely low. In addition, the survey respondents were than 0.6, the convergent validity of the constructs was
ensured anonymity, which encouraged them to respond more confirmed. Discriminant validity was supported by collat-
honestly and to resist the pressure to reply in a socially desir- ing the AVE and maximum shared variance (MSV)
able way. Thus, in the initial phase, the probability of common values, which showed that all the values of MSV were
method bias was minimized via these justifications. smaller than those of the AVE.
Furthermore, as suggested by MacKenzie and Podsakoff A correlation matrix was used to assess the validity of the
(2012), selecting respondents with the necessary experience constructs; the results are presented in Table 3.
in the relevant topic, avoiding the use vague concepts, and Multicollinearity was assessed by calculating the variance in-
using clear and concise language reduce common method bias flation factors (VIFs) and tolerance values among the indepen-
in survey research. Thus, to avoid common method bias, this dent variables. As recommended by Kleinbaum et al. (1988),
study recruited respondents who were managers of the SMEs multicollinearity is not an issue if the VIF value is lower than a
and had the necessary experience in digital innovation and threshold of 5–10 and has a tolerance greater than 0.2; in this
disseminated a survey questionnaire in Finnish, the native study, all the VIF values were in the range of 1.061–2.881
language of all respondents. Regarding non-respondent bias, with a tolerance greater than 0.3. As suggested by Fornell and
an analysis of the variance test (ANOVA) was conducted to Larcker (1981), the discriminant validity of the constructs can
determine whether any significant differences existed between be confirmed if the value of the square root of AVE is greater
the respondents’ answers on the questionnaire after the first than the value of the correlation between constructs.
reminder, called early respondents, and their answers after the Comparing the diagonal with non-diagonal values in Table 3
third reminder, named late respondents (Armstrong and confirmed the convergent validity of the constructs.
Overton 1977). The results revealed that there was no signif- Therefore, multicollinearity and convergent validity were not
icant difference between these two groups. Therefore, due to a significant concern in this study, as the requirements of all
the application of all these approaches, the study results were the recommended statistical tests were met.
 Inf Syst Front

Table 2 Results of validity and reliability testing

Latent variable Observed variable Loadings Cronbach’ α AVE CCR

Human capabilities Digital skills development is supported and promoted in our company. 0.778 0.772 0.514 0.760
Our employees are well trained in using digital tools. 0.718
Digitalization of the operating environment is easily accepted by our employees. 0.649
Collaboration capabilities Digital cooperation with other companies occurs. 0.882 0.799 0.597 0.812
Digital channels are used to share information with other companies. 0.824
Digitality transforms the social relationships in our business. 0.578
Technical capabilities Digitality increases the value of our products or services. 0.856 0.862 0.623 0.867
Digitality enables the integration of products and services into our company. 0.874
Digitality enables up-to-date, location-independent services for our customers.
Digitality allows us to work across boundaries of time, place, or activities. 0.795
Innovation capabilities Digitality enables innovation and new ideas in our company. 0.878 0.859 0.679 0.864
Digitality forces us to develop new solutions. 0.745
Digitality helps produce new products and services. 0.844
Business process Digitalization refers to real-time activities. 0.630 0.638 0.482 0.666
Digitalization refers to the automation of the production process. 0.754
Market offerings Digitalization refers to digital services. 0.666 0.591 0.428 0.616
Digitalization refers to digital products. 0.643

5 Results The first model was able to explain 47% of the variance in
digital market offerings and 29% of the variance in digital
Partial least squares (PLS)-based structural equation modeling business process. The path from technical capabilities (B =
(SEM) was used to test the hypotheses (SmartPLS v. 3.3.1). 0.617, p ≤ 0.10) to digital market offerings was significant,
Smart PLS has frequently been recommended due to its accu- whereas the paths from human capabilities (B = -0.215, ns),
racy and overall utility (Chuang and Lin 2015). SEM is suit- innovation capabilities (B = 0.050, ns), and collaboration ca-
able for data with multiple groups of regression, in which the pabilities (B = 0.239, ns) to digital market offerings were non-
dependent variable for one regression analysis is simulta- significant. Thus, technical capabilities had a direct positive
neously an independent variable for another (Hair et al. impact on digital market offerings, whereas human capabili-
1998). Thus, Aside from its capacity to handle multiple ties, innovation capabilities, and collaboration capabilities did
groups, PLS SEM can also be applied flexibly to both forma- not. The path from human capabilities (B = 0.559, p ≤ 0.01) to
tive and reflective constructs and, additionally, is undemand- digital business process was significant, whereas the paths
ing in terms of measurement scales, sample size, and distribu- from technical capabilities (B = -0.447, ns), innovation capa-
tional assumptions” depending on the intended meaning bilities (B = 0.482, ns), and collaboration capabilities (B = -
(Chuang and Lin 2015). Moreover, the PLS path models used 0.083, ns) to digital business process were non-significant.
in top tier journals are based on a significance level of 0.1 Thus, human capabilities had a direct positive impact on dig-
(Hair et al. 2012). Table 4 lists the testing paths from digital- ital market offerings, whereas technical capabilities, innova-
related capabilities to digital innovation. tion capabilities, and collaboration capabilities did not.

Table 3 Correlation matrix

Human Collaboration Technical Innovation Market offerings Business process

capabilities capabilities capabilities capabilities

Human capabilities 0.717a

Collaboration 0.543** 0.773a
capabilities
Technical capabilities 0.591** 0.525** 0.789a
Innovation capabilities 0.533** 0.568** 0.775** 0.824a
Market offerings 0.303** 0.355** 0.464** 0.445** 0.694a
Business process 0.332** 0.232** 0.255** 0.305** 0.073 0.654a
a
Square root of AVE, Sign. *** ≤ 0.001, ** 0.001 < p ≤ 0.01, * 0.01 < p ≤ 0.05
Inf Syst Front

Table 4 Testing the paths from digital-related capabilities to digital innovation

Path Path coefficient t-value

Model 1
Human capabilities -> Market offerings -0.215 1.094
Technical capabilities -> Market offerings 0.617 1.730*
Innovation capabilities -> Market offerings 0.050 0.146
Collaboration capabilities -> Market offerings 0.239 1.263
Human capabilities -> Business process 0.559 3.017**
Technical capabilities -> Business process -0.447 1.524
Innovation capabilities -> Business process 0.482 1.579
Collaboration capabilities -> Business process -0.083 0.433
Model 2
Technical capabilities -> Market offerings 0.675 7.842***
Human capabilities -> Technical capabilities 0.278 3.374***
Innovation capabilities -> Technical capabilities 0.776 9.917***
Collaboration capabilities -> Technical capabilities -0.095 1.079
Model 3
Human capabilities -> Business process 0.499 5,730***
Technical capabilities -> Human capabilities 0.691 3,035**
Innovation capabilities -> Human capabilities -0.200 0,780
Collaboration capabilities -> Human capabilities 0.389 3,795***

Notes: *** Significance ≤ 0.001; ** Significance ≤ 0.01; * Significance ≤ 0.10

The second model was able to explain 43% of the variance 6 Discussion
in digital market offerings. The path estimates of this model
provide further insight into hypotheses 1–4. The path from Digital transformation is shaping the business environments
technical capabilities (B = 0.675, p ≤ 0.001) to digital market of contemporary companies and Big Data analytics and digital
offerings was significant, as in the first model. The paths from business ecosystems are providing increasingly new possibil-
human capabilities (B = 0.278, p ≤ 0.001) and innovation ca- ities for companies to develop their businesses; therefore,
pabilities (B = 0.776, p ≤ 0.001) to technical capabilities were companies need to be able to respond to the changes and
also significant. Thus, human capabilities and innovation ca- understand the possibilities provided by increased digitaliza-
pabilities both had a significant positive impact on digital tion (Delgosha et al. 2020; Mikalef et al. 2020; Pappas et al.
market offerings through technical capabilities. The research 2018). Companies need to find ways to exploit innovation in
model also predicted a mediated path from collaboration ca- digitalizing business environments (Bednar and Welch 2019;
pabilities (B = -0.095, ns) to digital market offerings. Mikalef et al. 2018) and to understand the role of different
However, this path was not supported. types of capabilities as a part of these activities (Mikalef
The third model was able to explain 25% of the variance in et al. 2019).
digital business process. The path estimates of this model The study aimed to investigate what digital-related ca-
provide further insight into hypotheses 5–8. The path from pabilities are required to shape the exploitation of digital
human capabilities (B = 0.499, p ≤ 0.001) to digital business innovation, namely digital market offerings and the digital
process was significant, as in the first model. The paths from business process. The results demonstrated that human,
technical capabilities (B = 0.691, p ≤ 0.01) and collaboration technical, and innovation capabilities contribute to
capabilities (B = 0.389, p ≤ 0.001) to human capabilities were digital market offerings, while human, collaborative, and
also significant. Thus, technical capabilities and collaboration technical capabilities contribute to the digital business
capabilities both had a significant positive impact on digital process. These findings suggest that the studied
business process through human capabilities. The research capabilities play a significant role in the reshaping and
model also predicted a mediated path from innovation capa- digitizing of an organization in a way that enables new
bilities (B = -0.200, ns) to digital business process. However, digital innovations. These results are consistent with
this path was not supported. A summary of the results is pre- research from Gobble (2018) that stated that the processes
sented in Table 5. of digitalization and digital transformation may start as
 Inf Syst Front

Table 5 Summary of the results

Proposition Support Interpretation

H1: Human capabilities relate positively to Supported Human capabilities contribute indirectly to digital innovation in terms of digital market
digital market offerings. offerings. Human capabilities enhance technical capabilities, which in turn contribute to
digital market offerings.
H2: Collaboration capabilities relate positively Not Collaboration capabilities do not facilitate the creation of digital innovation in terms of
to digital market offerings. sup- digital market offerings.
ported
H3: Technical capabilities relate positively to Supported Technical capabilities directly contribute to the extent to which companies are able to create
digital market offerings. digital innovation in terms of digital market offering.
H4: Innovation capabilities relate positively to Supported Innovation capabilities contribute indirectly to digital innovation in terms of digital market
digital market offerings. offering. Innovation capabilities enhance technical capabilities, which in turn contribute
to digital market offering.
H5: Human capabilities relate positively to the Supported Human capabilities directly contribute to the extent to which companies are able to create
digital business process. digital innovation in terms of the digital business process.
H6: Collaboration capabilities relate positively Supported Collaboration capabilities contribute indirectly to digital innovation in terms of the digital
to the digital business process. business process. Collaboration capabilities enhance human capabilities, which in turn
contribute to the digital business process.
H7: Technical capabilities relate positively to Supported Technical capabilities contribute indirectly to digital innovation in terms of the digital
the digital business process. business process. Technical capabilities enhance human capabilities, which in turn
contribute to the digital business process.
H8: Innovation capabilities relate positively to Not Innovation capabilities do not facilitate the creation of digital innovation in terms of the
the digital business process. sup- digital business process.
ported

innovation initiatives, but they ultimately must reach far ital market offerings; in contrast, no direct effects were
beyond the innovation function to reshape the entire or- found for human, collaboration, and innovation capabili-
ganization. One of the main results of this study was that ties. Thus, this finding strongly supports prior statements
human capabilities directly contribute to the extent to that technical capabilities are the main enablers of organi-
which companies can create digital innovation through zational capabilities toward digital innovation (Banker
digital business processes. This result strongly supports et al. 2006; Mithas et al. 2011; Setia et al. 2013;
the findings from Rachinger et al. (2019). This research Tanriverdi 2005). One explanation for this finding is that
studied representatives from two different industries and digital technology creates a bond between the physical and
found that, from a capability perspective, digitalization digital characteristics of products and services, as presented
requires human skills (Rachinger et al. 2019). The authors by El Sawy et al. (2016). In addition, the technical capabil-
also pointed out challenges in the areas of employee re- ities surrounding digital technologies relate to the value
cruitment and qualification, and they highlighted the im- propositions of digital products and services, addressing
portance of companies’ ability to develop the know-how how value is generated and captured in each digital product
required to seize digitalization opportunities (Arnold et al. and service (Nylén and Holmström 2015). For H3, the pres-
2016; Rachinger et al. 2019). Therefore, while digital ent research findings are in line with those of former stud-
transformation provides novel opportunities for compa- ies, indicating that, because of the growing and increasingly
nies, it also requires an understanding of which capabili- ubiquitous presence of digital technologies at a reasonable
ties are required and how those capabilities should be cost, factors hindering market offerings will be removed.
adapted with regard to innovation outcomes, processes, Consequently, the exploitation of products and services
and related markets (Kohli and Melvill 2019; Nambisan will be facilitated with integrated products and services,
et al. 2017). In the following sections, the implications of as well as with updated and borderless activities (Nylén
the results for theory and practice are presented in detail. and Holmström 2015; Sjödin et al. 2016; Yoo et al.
2010). The findings also support previous statements that
6.1 Implications for Theory technical capabilities permit activities to cross the bound-
aries of time, place, and function (Parida et al. 2015; Yoo
Referring to digital innovation in terms of market offerings, 2010; Yoo et al. 2012) and provide the potential to reinvent
technical capabilities have significant direct effects on dig- sales and distribution channels (Nylén and Holmström
Inf Syst Front

2015). This result suggests that technical digital-related ca- Concerning digital innovation in terms of business process,
pabilities play a lead role in the provision and exploitation digital-related capabilities related to the human element had a
of market offerings. direct significant effect on the digital business process, where-
Referring to H1, the results of the study support previous as direct significant effects of collaboration, technical, and
research by indicating the complementary role of digital-related innovation capabilities on the digital business process were
human capabilities and continuous learning in digitalizing oper- not found. However, collaboration and technical capabilities
ating environments (Nylen and Holmströn 2015) and in had a mediating effect between human capabilities and the
launching new products in digital business (Fuchs and digital business process. The reason why human capabilities
Sevignani 2013). Although human capabilities in terms of em- have a predominant role may be the dynamic and cognitive
ployee readiness and digital skills play a role in the shift toward nature of digital business processes, which are more compat-
digitalization (Chan et al. 2019; El Sawy et al. 2016; Kane et al. ible with human capabilities in comparison to others.
2015; Legner et al. 2017), they do not directly reshape and ex- Although a variety of digital solutions provide possibilities
ploit market offerings. Instead, the effect takes place through for data gathering, and different types of artificial intelligence
technical capabilities. This means that skills related to under- solutions can assist companies, employees, and managers in
standing user experience and customer wishes and needs (Hoe decision making, as presented by Jarrahi (2018), humans re-
2017; Nylén and Holmström 2015) are being deployed to use main the sole authority for making decisions and generating
and select the technology needed to realize digital innovation. digital innovations. For H5, this study supports the view of
Regarding H2, the findings show that collaboration capabili- previous research, which suggested that employees’ digital
ties do not facilitate the creation of digital innovation in terms of skills and readiness to exploit the digital business process
digital market offerings. This is contrary to the predominant view (Richter et al. 2018) can be used to interpret the work gener-
in extant literature, which suggests that successful market offer- ated by digital solutions (Brynjolfsson and McAfee 2012).
ings in the blurry and complex arena of digitalization need col- These findings are in line with the notion that while real-
laboration, and that it is not possible to be successful in a time activities, artificial intelligence solutions, and automated
vacuum-like business environment (Canhoto et al. 2016; Kohli production processes assist and support companies, they can-
and Mekville 2019; Lyytinen et al. 2016; Pagani and Pardo not handle common-sense situations (Guszcza et al. 2017;
2017). This may be because digital technology is affordable Jarrahi et al. 2018) nor the entire structure of the digital busi-
and more ubiquitous than ever before, which may in turn facili- ness process. Digital solutions can assist in different aspects of
tate internal development and engagement in digital innovation the business process, but the combination of information de-
(cf. Nylén and Holmström 2015; Yoo et al. 2010). Additionally, rived from these aspects requires a common understanding of
digital technologies provide more sophisticated and easy-to-use the digital business process, which can be advanced by human
tools that together present opportunities for automatizing prod- capabilities. Hence, this finding confirms the important role of
ucts and making services smarter. This can in turn encourage human capabilities in moving toward digitalization in the dig-
companies to be less dependent on their partners and thereby ital business process.
reduce the importance of collaboration. Referring to H6, collaboration capabilities do not have a
In terms of H4, the results support the findings covered in direct effect on the digital business process. This may be be-
previous research that indicate that innovation capabilities are cause of the complex nature of collaboration in digital chan-
key organizational capabilities for creating new products and nels (Boland et al. 2007; Foltean et al. 2019; Iansiti and
services (Sia et al. 2016; Sousa and Rocha 2019). The current Lakhani 2014; Nath et al. 2010), the lack of skills needed
findings are also in line with those of Nylén and Holmström for collaborating with other companies via digital channels
(2015), who argued that exploiting novel opportunities for in activities that are more related to business processes
innovation is key to generating a compiled value via the cre- (Kietzmann et al. 2011), and the coordination costs and inte-
ation of digital products and services. However, like human gration difficulties in the product development process (Luo
capabilities (H1), innovation capabilities do not directly re- et al. 2010). However, collaboration capabilities facilitate the
shape and exploit market offerings. The effect actually takes business process through human capabilities. The role of col-
place through technical capabilities. The explanation for the laboration capabilities can thus be highlighted, since they ad-
mediation effect may be related to Fichman et al.’s (2014) vance human capabilities, which in turn positively affect the
earlier argument, in which digital innovation was claimed to business process.
require an understanding of what has become possible due to In terms of H7, the study results do not directly support the
advances in technology, as well as the exploitation of this hypothesis. In contrast to previous research that mentioned the
understanding to create something valuable for the company role of technical capabilities in digital business processes and
or for society. Overall, this finding illustrates the important operations (Chen et al. 2015; Nylén and Holmström 2015;
role of innovation capabilities in the exploitation of market Westergren and Holmström 2012; Yoo et al. 2012), the pres-
offerings. ent study did not discover a direct significant relationship
 Inf Syst Front

between technical capabilities and digital business processes. should concentrate more on digital-related human capabilities
The reason for this may be the collective nature of technical in order to obtain more benefits and the greater exploitation of
capabilities and the lack of linkages among companies using digital innovation in terms of the business process. Finally,
real-time activities. However, path modelling indicated that SME managers should consider the major role of technical
technical capabilities affect digital business processes through capabilities in digital innovation, namely market offerings,
human capabilities. Technical capabilities can thus be consid- as well as the dominant role of human capabilities in digital
ered as valuable drivers for human capabilities, which enable innovation, namely the business process.
the successful generation of novel information technology-
enabled products, services, and processes, whereby the pro-
cess innovation outcomes refer to the redesign and simplifica- 7 Conclusion
tion of the business process (cf. Kohli and Melville 2019).
Finally, concerning H8, this study did not support previous The aim of this research was to explore the conditions under
research indicating that innovation capabilities facilitate the which digital innovation opportunities emerge in SMEs.
creation of digital innovation in terms of digital business pro- Furthermore, the research identified the required capabilities
cesses. As mentioned in previous research, innovation capa- in digital transformation and understand the way of utilizing
bilities assist companies in sharing information (Lokuge et al. those capabilities to exploit digital innovation. Thus, the study
2019), corresponding to the opportunities and threats caused answered the question of what capabilities are required to
by digitalization (Sia et al. 2016), and appropriating digital shape the exploitation of digital innovation. As a first contri-
technologies relevant for their business activities (Nylen and bution, this research adds to the process perspective of digital
Holmström 2015). However, the results of this study may innovation. Few studies have concentrated on firm-level ca-
indicate that innovation capabilities offer opportunities for pabilities to identify, assimilate, and apply valuable knowl-
other types of digital innovation, such as digital market offer- edge from inside and outside the firm regarding opportunities
ings, rather than business process innovation. for digital innovation, known as initiate activity (Kohli and
Melville 2019). Further, this research extends the digital inno-
6.2 Implications for Practice vation literature by incorporating the effects of digital-related
capabilities based on digital innovation type. The findings
The findings of this research provide valuable insights for demonstrated that digital market offerings require technical-,
managers interested in establishing required digital-related ca- human-, and innovation-related digital capabilities. In con-
pabilities to develop and exploit digital innovations, including trast, the digital business process requires digital-related capa-
both market offerings and business processes in their business. bilities in terms of human, technical, and collaboration. Since
First, the vision of digitalization in this study may help man- the results showed that the capabilities required for the adop-
agers to perceive of their companies and nearby environments tion of digital innovation differ notably across the two digital
as a universal network, in which human, innovation, and es- innovation types, future research in this area should consider
pecially technical capabilities can create more opportunities the digital innovation types separately.
and potentials for the effective exploitation of market offer- This study had some limitations; however, these limitations
ings. Therefore, in the current digital era, managers may need can serve as the foundation for future research opportunities.
to focus more on these capabilities for enhancing their market First, the cross-sectional nature of the research imposed a limit
offerings in terms of digital products and services, as they are on understanding over time; simultaneously, however, it pro-
more likely to lead to the achievement of competitive advan- vided the possibility for longitudinal research in the future.
tages. Second, managers in charge of digital innovation Second, the provision of data from a single country as well
should consider and shape types of digital innovation in a as the focus on managers’ perceptions restricted the study’s
way that is compatible with the nature of digital-related capa- generalizability and increased the potential for bias; at the
bilities. For example, digital innovation in terms of market same time, however, these limitations present opportunities
offerings involves technology-related actions; thus, shaping to replicate this study in other countries and with other re-
market offerings requires technology understanding, such as sources. As mentioned above, many different methodological
digital-related capabilities emphasizing human, innovation, adjustments were made to correct or mitigate bias-related is-
and technical skills. In contrast, business processes in compa- sues. Third, the low reliability of the outcome measures may
nies are more dynamic and cognitive actions. Therefore, these have influenced the research results. A smaller value of alpha
capabilities make human digital-related capabilities more ap- is permissible if the scales are new and contain a small number
plicable in this regard. Third, managers in charge of digital of items (Nunnally 1978); however, there is a need for a sys-
transformation should better familiarize employees to and at- tematic approach to tackle this issue. Finally, since the study
tune working environments for digitalization; in addition, specialized in digital business processes inside companies,
managers should work to enhance digital culture, and they there is certainly more room for future research that would
Inf Syst Front

seek to link business processes outside the organization by

utilizing real-time activities and digital twinning, as well as
via collaboration among companies.

Funding Open access funding provided by LUT University.

Appendix

Table 6 Items and their measurement

Latent Observed variable References* Type of Scale

variable measurement

Human Digital skills development is supported and El Sawy et al. 2016; Lerch and Reflective 7-point scale ranging from “strongly
capabilities promoted in our company. Gotsch 2015 disagree” (1) to “strongly agree” (7)
Our employees are well trained in using
digital tools.
Digitalization of the operating environment
is easily accepted by our employees.
Collaborative Digital cooperation with other companies Amit and Han 2017; Chuang and Reflective 7-point scale ranging from “strongly
capabilities occurs. Lin 2015; El Sawy et al. 2016 disagree” (1) to “strongly agree” (7)
Digital channels are used to share
information with other companies.
Digitality transforms the social
relationships in our business.
Technical Digitality increases the value of our El Sawy et al. 2016; Parida et al. Reflective 7-point scale ranging from “strongly
capabilities products or services. 2015; Xue 2014 disagree” (1) to “strongly agree” (7)
Digitality enables the integration of
products and services into our company.
Digitality enables up-to-date,
location-independent services for our
customers.
Digitality allows us to work across
boundaries of time, place, or activities.
Innovation Digitality enables innovation and new Parida et al. 2015; Sia et al. 2016; Reflective 7-point scale ranging from “strongly
capabilities ideas in our company. Xue 2014 disagree” (1) to “strongly agree” (7)
Digitality forces us to develop new
solutions.
Digitality helps produce new products and
services.
Business In our company, digitalization refers to Fichman et al. 2014 Reflective 7-point scale ranging from “strongly
processes real-time activities. disagree” (1) to “strongly agree” (7)
In our company, digitalization refers to the
automation of the production process.
Market In our company, digitalization refers to Nambisan et al. 2017; Nylén and Reflective 7-point scale ranging from “strongly
offerings digital services. Holmström 2015 disagree” (1) to “strongly agree” (7)
In our company, digitalization refers to
digital products.

Note: *All the observed variables were developed based on the mentioned references. Thus, existing observed variables or implications of the mentioned
studies were used as a reference to formulate new laten variables suitable for the current study
 Inf Syst Front

Open Access This article is licensed under a Creative Commons Carmines, E. G., & Zeller, R. A. (1979). Reliability and validity
Attribution 4.0 International License, which permits use, sharing, assessment (Vol. 17). Thousand Oaks: Sage publications.
adaptation, distribution and reproduction in any medium or format, as Cenamor, J., Parida, V., & Wincent, J. (2019). How entrepreneurial
long as you give appropriate credit to the original author(s) and the SMEs compete through digital platforms: The roles of digital plat-
source, provide a link to the Creative Commons licence, and indicate if form capability, network capability and ambidexterity. Journal of
changes were made. The images or other third party material in this article Business Research, 100, 196–206.
are included in the article's Creative Commons licence, unless indicated Chan, C. M., Teoh, S. Y., Yeow, A., & Pan, G. (2019). Agility in
otherwise in a credit line to the material. If material is not included in the responding to disruptive digital innovation: Case study of an SME.
article's Creative Commons licence and your intended use is not Information Systems Journal, 29(2), 436–455.
permitted by statutory regulation or exceeds the permitted use, you will Chen, D., Heyer, S., Ibbotson, S., Salonitis, K., Steingrímsson, J. G., &
need to obtain permission directly from the copyright holder. To view a Thiede, S. (2015). Direct digital manufacturing: definition, evolu-
copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. tion, and sustainability implications. Journal of Cleaner Production,
107, 615–625.
Chuang, S. H., & Lin, H. N. (2015). Co-creating e-service innovations:
Theory, practice, and impact on firm performance. International
Journal of Information Management, 35(3), 277–291.
References Cortez, R. M., & Johnston, W. J. (2017). The future of B2B marketing
theory: A historical and prospective analysis. Industrial Marketing
Ahmad, S., Mallick, D. N., & Schroeder, R. G. (2013). New product Management, 66, 90–102.
development: impact of project characteristics and development Davila, T., Epstein, M., & Shelton, R. (2012). Making innovation work:
practices on performance. Journal of Product Innovation How to manage it, measure it, and profit from it. FT press.
Management, 30(2), 331–348. Delgosha, M. S., Saheb, T., & Hajiheydari, N. (2020). Modelling the
Amit, R., & Han, X. (2017). Value creation through novel resource con- asymmetrical relationships between digitalisation and sustainable
figurations in a digitally enabled world. Strategic Entrepreneurship competitiveness: a cross-country configurational analysis.
Journal, 11(3), 228–242. Information Systems Frontiers. https://doi.org/10.1007/s10796-
Armstrong, J. S., & Overton, T. S. (1977). Estimating nonresponse bias in 020-10029-0.
mail surveys. Journal of Marketing Research, 14, 396–402. El Sawy, O. A., Kræmmergaard, P., Amsinck, H., & Vinther, A. L.
Arnold, C., Kiel, D., & Voigt, K. I. (2016). How the industrial internet of (2016). How LEGO built the foundations and enterprise capabilities
things changes business models in different manufacturing indus- for digital leadership. MIS Quarterly Executive, 15(2), 141–166.
tries. International Journal of Innovation Management, 20(8), El-Haddadeh, R. (2020). Digital innovation dynamics influence on
1640015-1-1640015-25. organisational adoption: the case of cloud computing services.
Banker, R. D., Bardhan, I. R., Chang, H., & Lin, S. (2006). Plant infor- Information Systems Frontiers, 22(4), 985–999.
mation systems, manufacturing capabilities, and plant performance. Faems, D., Van Looy, B., & Debackere, K. (2005). Interorganizational
MIS Quarterly, 30(2), 315–337. collaboration and innovation: Toward a portfolio approach. Journal
of Product Innovation Management, 22(3), 238–250.
Barlett, J. E., Kotrlik, J. W., & Higgins, C. C. (2001). Organizational
Fichman, R. G., Santos, D., B.L., and Zheng, Z. (2014). Digital innova-
research: Determining appropriate sample size in survey research.
tion as a fundamental and powerful concept in the information sys-
Information Technology, Learning, and Performance Journal,
tems curriculum. MIS Quarterly, 38(2), 329–353.
19(1), 43–50.
Flynn, B. B., Sakakibara, S., Schroeder, R. G., Bates, K. A., & Flynn, E.
Bednar, P. M., & Welch, C. (2019). Socio-technical perspectives on smart
J. (1990). Empirical research methods in operations management.
working: Creating meaningful and sustainable systems. Information
Journal of Operations Management, 9(2), 250–284.
Systems Frontiers, 22, 1–18.
Foltean, F. S., Trif, S. M., & Tuleu, D. L. (2019). Customer relationship
Boland, R. J., Jr., Lyytinen, K., & Yoo, Y. (2007). Wakes of innovation in
management capabilities and social media technology use:
project networks: The case of digital 3-D representations in archi- Consequences on firm performance. Journal of Business
tecture, engineering, and construction. Organization Science, 18(4), Research, 104, 563–575.
631–647.
Fornell, C., & Larcker, D. F. (1981). Evaluating structural equation
Bouwman, H., Nikou, S., & de Reuver, M. (2019). Digitalization, busi- models with unobservable variables and measurement error.
ness models, and SMEs: How do business model innovation prac- Journal of Marketing Research, 18(1), 39–50.
tices improve performance of digitalizing SMEs? Fuchs, C., & Sevignani, S. (2013). What is digital labour? What is digital
Telecommunications Policy, 43(9), 101828. work? What’s their difference? and why do these questions matter
Boyer, K. K., & Pagell, M. (2000). Measurement issues in empirical for understanding social media? TripleC, 11(2), 237–293.
research: improving measures of operations strategy and advanced Gobble, M. M. (2018). Digitalization, digitization, and innovation.
manufacturing technology. Journal of Operations Management, Research-Technology Management, 61(4), 56–59.
18(3), 361–374. Guszcza, J., Lewis, H., & Evans-Greenwood, P. (2017). Cognitive col-
Brynjolfsson, E., Hu, Y., & Smith, M. D. (2010). Research laboration: Why humans and computers think better together. New
commentary—long tails vs. superstars: The effect of information York,: Deloitte University Press 1 (20): 7–30.
technology on product variety and sales concentration patterns. Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L.
Information Systems Research, 21(4), 736–747. (1998). Multivariate data analysis (Vol. 5, No. 3, pp. 207-219).
Brynjolfsson, E., & McAfee, A. (2012). Winning the race with ever- Upper Saddle River, NJ: Prentice hall.
smarter machines. MIT Sloan Management Review, 53(2), 53–60. Hair, J. F., Sarstedt, M., Ringle, C. M., & Mena, J. A. (2012). An assess-
Bstieler, L. (2005). The moderating effect of environmental uncertainty ment of the use of partial least squares structural equation modeling
on new product development and time efficiency. Journal of in marketing research. Journal of the Academy of Marketing
Product Innovation Management, 22(3), 267–284. Science, 40(3), 414–433.
Canhoto, A. I., Quinton, S., Jackson, P., & Dibb, S. (2016). The co- Henfridsson, O., & Yoo, Y. (2014). The liminality of trajectory shifts in
production of value in digital, university-industry R & D collabora- institutional entrepreneurship Organization Science. Organization
tive projects. Industrial Marketing Management, 56, 86–96. Science, 25(3), 932–950.
Inf Syst Front

Hoe, S. L. (2017). Thinking about how to think: cognitive skills to stay MacKenzie, S. B., & Podsakoff, P. M. (2012). Common method bias in
relevant in a digital economy. Human Resource Management marketing: Causes, mechanisms, and procedural remedies. Journal
International Digest, 25(5), 1–3. of Retailing, 88(4), 542–555.
Hull, C. E., & Covin, J. G. (2010). Learning capability, technological Maravilhas, S., & Martins, J. (2019). Strategic knowledge management a
parity, and innovation mode use. Journal of Product Innovation digital environment: Tacit and explicit knowledge in Fab Labs.
Management, 27(1), 97–114. Journal of Business Research, 94, 353–359.
Iansiti, M., & Lakhani, K. R. (2014). Digital ubiquity: how connections, Marion, T. J., Meyer, M. H., & Barczak, G. (2015). The influence of
sensors, and data are revolutionizing business. Harvard Business digital design and IT on modular product architecture. Journal of
Review, 92(11), 91–99. Product Innovation Management, 32(1), 98–110.
Jahanmir, S. F., & Cavadas, J. (2018). Factors affecting late adoption of Merrifield, R., Calhoun, J., & Stevens, D. (2008). The next revolution in
digital innovations. Journal of Business Research, 88, 337–343. productivity. Harvard Business Review, 86(6), 72–80.
Jarrahi, M. H. (2018). Artificial intelligence and the future of work: Mikalef, P., Boura, M., Lekakos, G., & Krogstie, J. (2019). Big data
Human-AI symbiosis in organizational decision making. Business analytics capabilities and innovation: the mediating role of dynamic
Horizons, 61, 577–586. capabilities and moderating effect of the environment. British
Kache, F., & Seuring, S. (2017). Challenges and opportunities of digital Journal of Management, 30(2), 272–298.
information at the intersection of Big Data Analytics and supply Mikalef, P., Pappas, I. O., Krogstie, J., & Giannakos, M. (2018). Big data
chain management. International Journal of Operations & analytics capabilities: a systematic literature review and research
Production Management, 37(1), 10–36 agenda. Information Systems and e-Business Management, 16(3),
Kallinikos, J., Aaltonen, A., & Marton, A. (2013). The ambivalent ontol- 547–578.
ogy of digital artifacts. MIS Quarterly, 37(2), 357–370. Mikalef, P., Pappas, I. O., Krogstie, J., & Pavlou, P. (2020). Big data and
Kane, G. C., Palmer, D., Phillips, A. N., & Kiron, D. (2015). Is your business analytics: A research agenda for realizing business value.
business ready for a digital future? Sloan Management Review, Information & Management, 57(1), 103237.
56(4), 37–44. Mithas, S., Ramasubbu, N., & Sambamurthy, V. (2011). How informa-
Kietzmann, J. H., Hermkens, K., McCarthy, I. P., & Silvestre, B. S. tion management capability influences firm performance. MIS
(2011). Social media? Get serious! Understanding the functional Quarterly, 35(1), 237–256.
building blocks of social media. Business Horizons, 54(3), 241–251. Nambisan, S., Lyytinen, K., Majchrzak, A., & Song, M. (2017). Digital
innovation management: reinventing innovation management re-
Kleinbaum, D. G., Kupper, L. L., & Muller, K. E. (1988). Applied re-
search in a digital world. MIS Quarterly, 41(1), 223–238.
gression analysis and other multivariate methods: student’s partial
solutions manual. Boston: PWS-Kent Publishing Company. Nath, P., Nachiappan, S., & Ramanathan, R. (2010). The impact of mar-
keting capability, operations capability and diversification strategy
Kohli, R., & Melville, N. P. (2019). Digital innovation: A review and
on performance: A resource-based view. Industrial Marketing
synthesis. Information Systems Journal, 29(1), 200–223.
Management, 39(2), 317–329.
Kolloch, M., & Dellermann, D. (2018). Digital innovation in the energy
Neirotti, P., Raguseo, E., & Paolucci, E. (2018). How SMEs develop
industry: The impact of controversies on the evolution of innovation
ICT-based capabilities in response to their environment. Journal of
ecosystems. Technological Forecasting and Social Change, 136,
Enterprise Information Management. https://doi.org/10.1108/
254–264.
JEIM-09-2016-0158.
Langlois, R. N. (2003). The vanishing hand: the changing dynamics of Ngo, L. V., & O’Cass, A. (2013). Innovation and business success: The
industrial capitalism. Industrial and Corporate Change, 12(2), 351– mediating role of customer participation. Journal of Business
385. Research, 66(8), 1134–1142.
Lee, J., & Berente, N. (2012). Digital innovation and the division of Nunnally, J. C. (1978). Psychometric theory (2nd ed.). New York:
innovative labor: Digital controls in the automotive industry. McGraw-Hill.
Organization Science, 23(5), 1428–1447.
Nylén, D., & Holmström, J. (2015). Digital innovation strategy: A frame-
Legner, C., Eymann, T., Hess, T., Matt, C., Böhmann, T., Drews, P., & work for diagnosing and improving digital product and service in-
Ahlemann, F. (2017). Digitalization: opportunity and challenge for novation. Business Horizons, 58(1), 57–67.
the business and information systems engineering community. Pagani, M., & Pardo, C. (2017). The impact of digital technology on
Business & Information Systems Engineering, 59(4), 301–308. relationships in a business network. Industrial Marketing
Lerch, C., & Gotsch, M. (2015). Digitalized product-service systems in Management, 67, 185–192.
manufacturing firms: A case study analysis. Research-Technology Pappas, I. O., Mikalef, P., Giannakos, M. N., Krogstie, J., & Lekakos, G.
Management, 58(5), 45–52. (2018). Big data and business analytics ecosystems: Paving the way
Li, L., Su, F., Zhang, W., & Mao, J. Y. (2018). Digital transformation by towards digital transformation and sustainable societies. Information
SME entrepreneurs: A capability perspective. Information Systems Systems and e-Business Management, 16(3), 479–491.
Journal, 28(6), 1129–1157. Parida, V., Sjödin, D. R., Lenka, S., & Wincent, J. (2015). Developing
Lokuge, S., Sedera, D., Grover, V., & Dongming, X. (2019). global service innovation capabilities: How global manufacturers
Organizational readiness for digital innovation: Development and address the challenges of market heterogeneity. Research-
empirical calibration of a construct. Information & Management, Technology Management, 58(5), 35–44.
56(3), 445–461. Peppard, J., & Rylander, A. (2006). From value chain to value network:
Luo, C., Mallick, D. N., & Schroeder, R. G. (2010). Collaborative prod- Insights for mobile operators. European Management Journal,
uct development: Exploring the role of internal coordination capa- 24(2), 128–141.
bility in supplier involvement. European Journal of Innovation Porter, M. E., & Heppelmann, J. E. (2015). How smart, connected prod-
Management, 13(2), 244–266. ucts are transforming companies. Harvard Business Review, 93(10),
Lusch, R. F., & Nambisan, S. (2015). Service innovation: A service- 96–114.
dominant logic perspective. MIS Quarterly, 39(1), 155–175. Rachinger, M., Rauter, R., Müller, C., Vorraber, W., & Schirgi, E. (2019).
Lyytinen, K., Yoo, Y., & Boland, R. J. Jr. (2016). Digital product inno- Digitalization and its influence on business model innovation.
vation within four classes of innovation networks. Information Journal of Manufacturing Technology Management, 30(8), 1143–
Systems Journal, 26(1), 47–75. 1160.
 Inf Syst Front

Richter, A., Heinrich, P., Stocker, A., & Schwabe, G. (2018). Digital Taber, K. S. (2018). The use of Cronbach’s alpha when developing and
work design – the interplay between human and computer in future reporting research instruments in science education. Research in
work practices as an interdisciplinary (grand) challenge. Journal of Science Education, 48(6), 1273–1296.
Business & Information Systems Engineering, 60(3), 259–264. Tanriverdi, H. (2005). Information technology relatedness, knowledge
Rogers, E. M. (1983). Diffusion of innovations. New York: A Division of management capability, and performance of multi business firms.
Macmillan Publishing Co. MIS Quarterly, 23, 311–334.
Santoro, G., Vrontis, D., Thrassou, A., & Dezi, L. (2018). The internet of Teece, D. J. (2014). The foundations of enterprise performance: Dynamic
things: building a knowledge management system for open innova- and ordinary capabilities in an (economic) theory of firms. Academy
tion and knowledge management capacity. Technological of Management Perspectives, 28(4), 328–352.
Forecasting and Social Change, 136, 347–354. Teece, D. J., Pisano, G., & Shuen, A. (1997). Dynamic capabilities and
Setia, P., Venkatesh, V., & Joglekar, S. (2013). Leveraging digital tech- strategic management. Strategic Management Journal, 18(7), 509–
nologies: How information quality leads to localized capabilities and 533.
customer service performance. MIS Quarterly, 37, 565–590. Urueña, A., Hidalgo, A., & Arenas, ÁE. (2016). Identifying capabilities
Sia, S. K., Soh, C., & Weill, P. (2016). How DBS Bank pursued a digital in innovation projects: Evidences from eHealth. Journal of Business
business strategy. MIS Quarterly Executive, 15(2), 105–121. Research, 69(11), 4843–4848.
Westergren, U. H., & Holmström, J. (2012). Exploring preconditions for
Sivasubramaniam, N., Liebowitz, S. J., & Lackman, C. L. (2012).
open innovation: Value networks in industrial firms. Information
Determinants of new product development team performance: A
and Organization, 22(4), 209–226.
meta-analytic review. Journal of Product Innovation Management,
Xue, L. (2014). Governance–knowledge fit and strategic risk taking in
29(5), 803–820.
supply chain digitization. Decision Support Systems, 62, 54–65.
Sjödin, D. R., Parida, V., & Kohtamäki, M. (2016). Capability configu- Yoo, Y. (2010). Computing in everyday life: A call for research on ex-
rations for advanced service offerings in manufacturing firms: Using periential computing. MIS Quarterly, 34, 213–231.
fuzzy set qualitative comparative analysis. Journal of Business Yoo, Y., Boland Jr, R. J., Lyytinen, K., & Majchrzak, A. (2012).
Research, 69(11), 5330–5335. Organizing for innovation in the digitized world. Organization
Sousa, M. J., & Rocha, Á (2019). Skills for disruptive digital business. Science, 23(5), 1398–1408.
Journal of Business Research, 94, 257–263. Yoo, Y., Henfridsson, O., & Lyytinen, K. (2010). Research
Street, C. T., Gallupe, B., & Baker, J. (2017). Strategic alignment in commentary—the new organizing logic of digital innovation: an
SMEs: Strengthening theoretical foundations. Communications of agenda for information systems research. Information Systems
the Association for Information Systems, 40, 20. Research, 21(4), 724–735.
Svahn, F., Mathiassen, L., & Lindgren, R. (2017). Embracing digital
innovation in incumbent firms: how Volvo cars managed competing Publisher’s Note Springer Nature remains neutral with regard to juris-
concerns. MIS Quarterly, 41(1), 239–254. dictional claims in published maps and institutional affiliations.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
1

Radio Resource Management for QoS

Guarantees in Cyber-Physical Systems
Shao-Yu Lien, Shin-Ming Cheng, Member, IEEE, Sung-Yin Shih, and Kwang-Cheng Chen, Fellow, IEEE

Abstract—The recent deployment of Cyber-Physical Systems (CPS) has emerged as a promising approach to provide extensive
interaction between computational and physical worlds. For a large-scale distributed CPS comprising of numerous machines, sharing
radio resource efficiently with the existing wireless networks while maintaining sufficient quality-of-service (QoS) for machine-to-
machine (M2M) communications becomes an essential and challenging requirement. By clustering CPS machines as a swarm with
the cluster head managing radio resources inside the swarm, spectrum sharing among numerous machines can be achieved in a
distributed and scalable fashion. Specifically, we apply the recent innovation, cognitive radio, and a special mode in cognitive radio,
interweave coexistence, to leverage machines to collect radio resource usage information for autonomous and interference-free radio
resource management in the CPS. To reduce the communication overheads of channel sensing feedbacking from machines, we apply
compressive sensing to construct a spectrum map indicating the radio resource availability on any given locations within the CPS
coverage. Such spectrum map resource management (SMRM) only utilizes a small portion of machines to perform channel sensing
but enables distributed cluster-based spectrum sharing in an efficient way. Through the concept of effective capacity, the SMRM controls
available resources to guarantee the QoS for communications of CPS. By evaluating the performance of the proposed SMRM in the
most promising realization of CPS based on LTE-Advanced machine-type communications coexisting with LTE-Advanced Macrocells
to utilize identical spectrum, the simulation results show effective QoS guarantees of CPS by SMRM in the realistic environments.

Index Terms—Cyber-physical systems (CPS), cognitive radio (CR), spectrum map, compressive sensing, machine-to-machine (M2M)
communication, quality-of-service (QoS), effective capacity, radio resource management

1 I NTRODUCTION essential, especially in the most challenging and vari-

There is an increasing concern for integrating real-world ant environments of wireless links. In other words, the
physical activities and the cyber-world of computations quality-of-service (QoS) guaranteed wireless networking
and communications to design efficient Cyber-Physical among machines is essential for many cases [6], [7].
Systems (CPS) [1]. These systems consist of networks Furthermore, to preserve the precious spectrum re-
of embedded computations and communications ma- sources, communications/connections among machines
chines which monitor and control the physical entities desirably exploit spectrum of the existing radio networks
via sensors and actuators on electrical power grids, (such as LTE-Advanced [3], [8]), and thus the appearance
transportation vehicles and traffic roads, robotic systems, of numerous machines induces catastrophic impacts on
healthcare and medical machines, environmental control the existing radio network. Although, in LTE-Advanced,
and smart buildings. To bridges cyber and physical a base station (BS) is designed to be able to simultane-
worlds, CPS must operate in a reliable, safe, secure, and ously deal with accesses of machines and human [9],
real-time fashion [2]. it is desirable to separate managements of machine-to-
The major characteristic of CPS is that there can be an machine (M2M) communications and human-to-human
extremely large number of machines involved in the sys- (H2H) communications to different BSs to significantly
tem operation. To provide ubiquitous communications reduce the burden of each BS. As a result, a BS dealing
among such a large amount of machines without addi- with M2M communications is referred as a CPS-BS (and
tional cyber-infrastructure deployment costs, connecting forms a CPS-cell). On the other hand, a BS dealing with
all these machines by leveraging cellular communication H2H communications is referred as a Macro-BS (and
systems turns out to be an effective and efficient solution. forms a Macrocell). To control the interference between
As a result, a promising realization of CPS can be CPS-cells and existing Macrocells, a centralized resource
the machine-type communication (MTC) in Long Term management is not feasible due to an unacceptable
Evolution-Advanced (LTE-Advanced) [3]–[5]. Although complexity of the radio resource allocation. As a result,
all machines could be connected via cyber-infrastructure how to distributively share the limited radio resources
of LTE-Advanced, it does not suggest a successful real- according to the QoS demand of machines without in-
ization of CPS. To ensure the effective operation of CPS, ducing harmful interference so that numerous machines
the reliable information delivery of physical events is could underlay with the existing radio network emerges
as a primary challenge.
S.-Y. Lien, S.-M. Cheng, S.-Y. Shih, and K.-C. Chen are with the Graduate To tackle this challenge, the cognitive radio (CR)
Institute of Communication Engineering, National Taiwan University, Taipei
106, Taiwan, e-mail: [email protected], [email protected], technology enabling a station to cognise and adapt to
[email protected], and [email protected] communications environments so as to achieve the op-

Digital Object Indentifier 10.1109/TPDS.2012.151 1045-9219/12/$31.00 © 2012 IEEE

 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
2

timum overall network performance [10], [11] is consid-

erably noted. Under the design constraint of simplicity &ORXG
without imposing additional complexities on current
radio network protocols and operations, the CR technol- &36%6
ogy is well-suited for CPS communications. Under this &36%6
framework, if each machine performs periodical channel &36%6

sensing to identify the radio resources usage of existing

&36FHOO
Macrocells and only utilizing radio resources identified
as unoccupied by existing Macrocells, interference can &36FHOO
thus be mitigated. However, under this framework, all
machines in the CPS-cell shall perform channel sensing
and report sensing results to the CPS-BS. This leads to
an enormous amount of feedback data, which makes the
concept of CR infeasible to be directly applied to CPS-
cells.
&36FHOO
To achieve autonomous interference mitigation by 0DFURFHOO
leveraging the CR technology in the CPS-cell, the key
is to reduce the amount of feedback. One effective so- 0DFURFHOO &36FHOO 0DFURFHOO
0DFUR%6
lution is to significantly reduce the number of machines &36FHOO
required to perform channel sensing. Specifically, only 0DFUR%6 &36%6 0DFUR%6
a small number of machines in the CPS-cell (instead of
all machines) devotes to perform channel sensing and &36%6 0DFUR%6

report sensing results, while interference levels of all

&36%6
machines can be obtained. In addition, when numerous 0DFUR%6 0DFUR%6

machines directly execute monitoring or control sub-

0DFURFHOO
0DFUR%6 &36FHOO 0DFURFHOO
process in CPS simultaneously, a considerable amount
of information is funneled toward the same destination
0DFURFHOO
and utilize common resources. Even with the aid of
'DWD$JJUHJDWRU 0DFKLQH &RPPXQLFDWLRQVLQVLGHWKHVZDUP
CR, the available radio resource might not be sufficient
&RPPXQLFDWLRQVEHWZHHQVZDUPDQGFORXG
and the regional hot spots might be spawn. The re-
sulting large delays and wasteful packet drops violate Fig. 1. The CPS architecture based on LTE-Advanced
QoS requirements [12]. When the density of machines machine-type communications that coexists with LTE-
grow, rudimentary coordination in the form of resource Advanced Macrocells.
management becomes essential.
For the above two goals, (i) it is proposed to group
machines into clusters in the swarm and the moni- specifying the maximum constant arrival rate that the
tor/control information is collected/distributed by an system can support while satisfying the given QoS re-
elected data aggregator, DA (i.e., cluster head), then DA quirement. We consequently propose the spectrum map
transmits aggregated traffic to the cloud via CPS-BS. based resource management (SMRM) for CPS-cells to
In addition, (ii) a powerful technology known as the construct the spectrum map. By analytically deriving the
compressive sensing [13] widely applied to the signal effective capacity of the SMRM, a systematic procedure
compression and restoration is particularly noted, by for radio resource allocation of the CPS-cell is further
which the ordinary sparse signal is sampled with a sam- proposed, thus enabling a successful realization of CPS.
pling rate far lower than the Shannon/Nyquist sampling The remainder of this article is structured as follows.
rate, while the signal can still be recovered with a high After elaborating the system model and related back-
probability. By leveraging the compressive sensing, DAs grounds in Section 2, the SMRM for the CPS-cell is pro-
and a portion of machines (instead of all machines) are posed in Section 3, and the control of sensing period and
required to perform channel sensing and report sensing RBs allocation to achieve QoS guarantees are provided
results to the CPS-BS, and interference levels at any in Section 4. The performance of the proposed SMRM
given swarms/locations within the coverage of the CPS- is evaluated in Section 5 and this paper is concluded in
cells can be constructed, which is particularly referred Section 6.
as the spectrum map. As a result, the amount of channel
information feedback is significantly reduced and the CR 2 S YSTEM M ODEL AND R ELATED BACK -
technology can be applied to CPS-cells for autonomous
GROUNDS
interference mitigation.
To provide QoS guarantees while fully exploiting the We consider the radio resources management problem
radio resource in the swarm, we adopt the concept of with multiple CPS-cells overlaying Macrocells, as shown
effective capacity [14], which is a link-layer channel model in Fig. 1. A CPS-cell is composed of a CPS-BS, V
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
3

6HQVLQJSHULRG 7 5HVRXUFHEORFN5% AR×N xN ×1 , where the dimension R  N . To reconstruct

x from y, there are two necessary conditions to satisfy:
)UHTXHQF\GRPDLQ

1) Sparsity of signal x: The N -dimensional vector

signal x is K-sparse, that is, there are only K  N
elements in x that are non-zero. When x is binary,
x is K-sparse if there are only K  N elements
in x that are non-zero, or there are only K  N
elements in x that are zero.
2) Restricted isometry property (RIP) of measurement
7LPHGRPDLQ 6HQVLQJVXEIUDPH 'DWDVXEIUDPH
matrix A: For any K-sparse signal x, the R × N
Fig. 2. The subframe structure of CPS. In the SMRM, matrix A has the property that
a subframe is referred as a “sensing subframe” if the x2 (1 − δ) ≤ Ax2 ≤ x2 (1 + δ). (1)
subframe is allocated for performing channel sensing;
otherwise, a subframe is referred as a “data subframe”. where 0 ≤ δ ≤ 1.
Then, given A and y, l1 minimization problem with R 
cKlog(N/K) is able to recover x by
machines and S DAs. A Macrocell is composed of a
Macro-BS and multiple user equipments (UEs). In LTE- x∗ = argmin x1 subject to y = Ax (2)
Advanced, all CPS-cells and Macrocells adopt orthog-
onal frequency division multiple access (OFDMA) and
share all available spectrum. Considering that all CPS- 2.3 Preliminary of Statistical QoS Guarantees
cells and Macrocells belong to the same wireless tech- The real-time services typically require bounded delays.
nology (e.g., LTE-Advanced), the subframe structure of Due to the impact of time-varying fading channels, it had
the Macrocell and CPS-cells are identical and boundaries been shown that providing deterministic QoS guarantees
of frames of CPS-cells align to that of Macrocells. Radio (that is, the probability that the transmission delay vi-
resources are allocated in the unit of “resource block” olates the delay requirement is zero) over the Rayleigh
(RB) as shown in Fig. 2. Denote the number of RBs fading channel is impossible [14]. As a result, a practical
in each subframe as M , and these RBs are indexed by solution turns out to provide the statistical QoS guar-
m = 1, . . . , M . antees (i.e., the probability that the transmissions delay
violates the delay requirement is bounded by a required
2.1 Power Control Considerations value). For this purpose, the large deviation theory [14]
provides that, for stationary arrival and service pro-
In the CPS-cell, power control is used to combat the cesses under sufficient conditions, the probability that
channel fading and we choose to adopt the truncated the buffer length B exceeds a certain threshold B  decays
channel inversion [15], by which the signal to interference exponentially fast as the threshold B  increases. That is,
and noise power ratio (SINR) on each RB can be main- 
tained at a required value, so that all RBs in a subframe Pr{B > B  } ≈ e−θB (3)
can carry the same number of bits [15]. As a result, there
is no difference in selecting different RBs. This setting where θ is a positive constant called QoS exponent. When
simplifies the problem of RBs allocation. In this paper, delay is the main QoS metric of interests, an expression
we say that an RB without suffering interference if this similar to (3) is given by
RB is unoccupied by Macrocells. Pr{delay > dmax } ≈ e−θδdmax (4)
where dmax is the delay bound and δ is jointly deter-
2.2 Compressive Sensing mined by the arrival process and the service process.
Based on the revelation that a small collection of a sparse From (4), it can be observed that a small θ implies that
signal contains enough information for reconstruction, the system can only support a loose QoS requirement,
Compressive Sensing (CS) [13], [16], [17] shows high while a large θ means that a strength QoS requirement
promise for distributed scheme. CS is a new paradigm can be supported.
to perform sampling and compression simultaneously To provide statistical delay guarantees, the effective
thus significantly reduce the sampling rate. Traditional bandwidth and the effective capacity serve significant
approaches acquire the entire signal and process it to foundations. The effective bandwidth [14], denoted by
exact the information. CS acquires only a small number EB (θ), specifies the maximum constant service rate needed
of linear measurements that preserve the structure of the by the given arrival process subject to a given θ. On the
signal, and the signal is then reconstructed from these other hand, the effective capacity, denoted by EC (θ), is
measurements using an optimization process. Specifi- the duality of the effective bandwidth, which specifies
cally, CS uses the measuring matrix A to measure the the maximum constant arrival rate that can be supported by
signal x and acquire the measurements y: yR×1 = the system subject to a given θ [14]. If θ∗ can be found
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
4

7KHFRYHUDJHRIWKH&36FHOOLV
as the solution of EB (θ∗ ) = EC (θ∗ ), δ can be obtained GLYLGHGLQWR1 LVRWURSLFJULGV
by [18] as &36FHOO
∗ ∗
δ = EB (θ ) = EC (θ ). (5) &36%6

Consequently, the system can achieve the statistical guar- &36%6

antee
∗
Pr{delay > dmax } ≈ e−θ δdmax &36%6
. (6)
The effective capacity can be formally defined by [14] 0DFKLQHSHUIRUPVFKDQQHOVHQVLQJDQGUHSRUWVVHQVLQJUHVXOW

Λc (−θ) 1   t  0DFKLQHGRHVQRWSHUIRUPFKDQQHOVHQVLQJDQGGRHVQRWUHSRUWVHQVLQJUHVXOW





EC (θ)  − = − lim log E e−θ i=1 R[i] (7)
θ t→∞ θt
t RFFXSLHG5%
where E[·] denotes taking the mean value, i=1 R[i] 5HSRUWLQJZKLFK5%LVRFFXSLHG
LQWKHVHQVLQJVXEIUDPH XQRFFXSLHG
is the partial sum of the discrete-time stationary and 5% 6HQVLQJVXEIUDPH
ergodic service process {R[i], i = 1, 2, . . .} and
1   t  Fig. 3. The compressive sensing is applied to construct
Λc (θ) = lim ( ) log E eθ i=1 R[i] (8) the spectrum map of the CPS-cell by channel sensing
t→∞ t
results of a portion of machines. The coverage of the
is a convex function differentiable for all real θ. To CPS-cell is divided into N isotropic “grids”
achieve the statistical delay guarantees, approaches to
derive the effective bandwidth of real-time streams had
been widely discussed (e.g., the effective bandwidth of overheads. In LTE-Advanced, DAs can be deployed by
the voice traffic can be obtained by the method proposed the system operator (for example, relay stations [20]
in [19]). However, the effective capacity depends on the in LTE-Advanced can be very powerful DAs) and the
system design. We therefore shall derive the effective swarm can be constructed by DAs. In the following, we
capacity of the proposed SMRM. devote to the spectrum map construction in Step 3 of
Algorithm 1 by the compressive sensing.
3 T HE SMRM FOR THE CPS- CELL
3.1 Compressive Sensing for the Spectrum Map
To mitigate interference, the CPS-BS shall avoid allo- Construction
cating occupied RBs (by Macrocells) to DAs (and thus
machines in the swarm). As a result, the key idea of Denote Ψ = [ψ1 ψ2 . . . ψN ]T as the true RBs occupation of
interference mitigation is that all machines shall au- Macrocells, where ψn indicates M RBs occupations of ex-
tonomously estimate the RBs usage of the Macrocell isting Macrocells in a sensing subframe on the nth grid.
and report the sensing result (that is, identifying which Upon receiving feedback from DAs, the compressive
RB is occupied by the Macrocell in a subframe at any sensing is obtained by multiplying a sampling matrix
given locations, i.e., the spectrum map). However, this on the true RBs occupation of Macrocells Ψ,
approach is infeasible due to potentially unacceptable y = AΨ + ε (9)
amount of uplink reporting data. As a result, the number
of machines required to perform channel sensing shall where A is a R × N matrix with each element taking “1”
be significantly reduced. To resolve this issue, we shall with probability q VVn , and taking “0” with probability
adopt the compressive sensing for the spectrum map 1 − q VVn , where Vn is the number of machines within
construction. For this goal, the coverage of the CPS- the nth grid. The spectrum map can be constructed by
cell is divided into N isotropic “grids” indexed by searching the minimum l1 norm of Ψ,
n = 1, . . . , N , as shown in Fig. 3. With the facilitation of
Ψ∗ = arg min Ψ1 s.t. AΨ − y2 ≤  (10)
the spectrum map, interference can be mitigated by only
allocating unoccupied RBs to machines. Such an SMRM by applying the second order corn programming [21] if
is proposed in Algorithm 1 as follows. R = O(K log NKM ), where lp norm of Ψ is Ψp :=
By the proposed SMRM, only V q machines in ex- 
N

pectation shall perform channel sensing and report the ( |ψn |p )1/p , K is the sparsity of A,  = Φε2 and Φ is
n=1
sensing result. Since V q ≤ V , the number of machines a random basis.
required to perform channel sensing can be reduced. Fur- To successfully construct the spectrum map by the
thermore, since the received interference power had been compressive sensing with an acceptable quality, it is
adopted by 3GPP LTE-Advanced as a mandatory sensing suggested that two conditions shall be considered [13]:
quantity and corresponding sensing results report pro- (i) Ψ shall be sparse. (ii) The selection of A shall satisfy
cedures had also been defined by LTE-Advanced [20], the restricted isometry property (RIP). Since, in typical situ-
the proposed SMRM can be applied to CPS-cell without ation in the urban environment, most RBs in a subframe
any hardware modifications and additional signaling could be occupied by Macrocells, Ψ is typically sparse.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
5

Algorithm 1 SMRM manner among subframes. To capture these characteris-

1: The CPS-BS periodically allocates subframes for ma- tics, the CPS-BS is proposed to extract following param-
chines to perform channel sensing to identify which eters from channel sensing in Step 5 of the SMRM.
RB is occupied. The sensing period is T subframes Definition 1: Let MMacro be the number of RBs occu-
and each channel sensing persists for one subframe pied by Macrocells in a subframe, the traffic load of
(Fig. 2). A subframe is referred as a “sensing sub- Macrocells, ρ, is defined as ρ  MMacro /M .
frame” if the subframe is allocated for performing Definition 2: When an RB is occupied by Macrocells
channel sensing; otherwise, the subframe is referred in a subframe, the (aggregated) RB allocation correla-
as a “data subframe”. All machines and DAs can tion probability of Macrocells, η, is the probability that
not perform data transmissions/receptions within a Macrocells will still occupy this RB in the subsequent
sensing subframe. subframe.
2: Within a sensing subframe, each machine performs Definition 3: Let Mη be the number of RBs with non-
channel sensing with probability q to measure the zero η in a subframe, the fraction of correlated RBs
received interference power on each RB. allocation of Macrocells, ϕ, is defined as ϕ  Mη /ρM .
3: if the received interference power on an RB exceeds a To obtain these parameters, machines still only need
certain threshold then to sense the received interference power on each RB,
4: this RB is identified as being occupied; otherwise, by which machines can identify the traffic load of the
the RB is unoccupied. The machine reports the sens- Macrocell, ρ. By further performing the time-domain
ing result to the DA and thus the CPS-BS. correlation of channel sensing results [24], [25], η and
5: The CPS-BS constructs the spectrum map based on ϕ can be obtained. η and ϕ capture the correlation of
the channel sensing report (the method to construct the RBs allocation among subframes in Macrocells. If
the spectrum map will be elaborated later) both η and ϕ are large, it implies that Macrocells allocate
6: In subsequent data subframes, the CPS-BS only allo- RBs in a high correlation manner among subframes. By
cates unoccupied RBs sensed in the sensing subframe obtaining these parameters in Step 5 of Algorithm 1,
to machines. a systematic procedure is proposed in next section to
7: The CPS-BS also extracts following parameters from control T and RBs allocation for QoS guarantees.
the result of reports: (i) the traffic load of existing
Macrocells, (ii) the RBs allocation correlation prob- 3.3 The Impact of the SMRM to Macrocells
ability of existing Macrocells, and (iii) the fraction According to the state-of-the-art operation of LTE-
of correlated RBs allocation of existing Macrocells, Advanced [20],the Macro-BS is able to coordinate each
which are detailed later. UE to estimate the channel quality of RBs and only
allocate RBs with an acceptable channel quality to the
UE. As a result, when an RB is sensed as not being
From [22], [23], although it is suggested that the RIP can occupied by Macrocells and this RB is utilized by the
be achieved if each machine senses the channel the i.i.d. CPS-cell, such an RB is with a poor quality to Macro-
and symmetry Bernoulli random variable, in Section 5, cells and Macrocells avoid utilizing such an RB. Such
we will show that the spectrum map can be constructed an operation is known as the “interweave” coexistence
with an acceptable quality when each machine senses between CPS-cells and Macrocells [26]. However, it does
the channel with probability q of the Bernoulli random not suggest that there is no interfere between Macrocells
variable. and CPS-cells under the interweave coexistence. Due
the channel quality variation, Macrocells may change
3.2 More Considerations on the SMRM the RBs allocation in each subframe. As we will see
later, when Macrocells change the RBs allocation very
By applying the SMRM, if the existing allocations of RBs frequently (i.e., η and ϕ are of small values), interference
change very frequently among subframes, there could between Macrocells and CPS-cells turns to severe. By
be estimation errors of RBs usage. Therefore, a small estimating η and ϕ, in the next section, the control of
sensing period T can decrease the estimation error and T and RBs allocation is optimized for CPS-cells under
thus interference. However, since channel sensing is an any cases of η, ϕ and ρ.
overhead, a small T implies that more radio resource
(more sensing subframes) are used for channel sensing.
As a result, there is a tradeoff between sensing period
4 T HE C ONTROL OF T AND THE RB S A LLO -
T and interference. A good tradeoff shall provide QoS CATION
guarantees for machines while fully utilizing the radio The ultimate goal of the control of T and the RBs
resource. To achieve this goal, the CPS BS should ap- allocation is to provide QoS guarantees while achieving
propriately control T and the RBs allocation. For this an efficient radio resource allocation. From (5) and (6),
purpose, the CPS-BS needs more information about the it is known that the effective capacity facilitate us to
characteristics of RBs allocations of existing users. That achieve this goal. In the following, we derive the effec-
is, RBs are allocated in a high or in a low correlation tive capacity of the SMRM.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
6

0DFURFHOO
4.1 Effective Capacity of the SMRM
0DFURFHOO &36FHOO 0DFURFHOO
By obtaining parameters in Step 5 of Algorithm 1, we 0DFUR%6

start the derivation of the effective capacity of the SMRM &36FHOO

in the following by temporarily assuming that the RB uti- 0DFUR%6 &36%6 0DFUR%6

lized by the CPS-cell suffers no interference. Considering &36%6 0DFUR%6

that all RBs carry the same number of bits (say b bits)
in a subframe, based on (7), the effective capacity of a 0DFUR%6
&36%6
0DFUR%6

machine utilizing one RB without suffering interference

0DFUR%6 &36FHOO
is given by 0DFURFHOO 0DFURFHOO

1   t 
0DFURFHOO
E1C (θ) = − lim log E e−θ i=1 R[i] . (11)
t→∞ θt
Fig. 4. Network topology for simulations. There are seven
For the block fading channel wherein the service process hexagonal-grid Macrocells with wrap-around. Three CPS-
{R[i], i = 1, 2, . . .} is uncorrelated, (11) can be rewritten cells are deployed at the boundary of the central Macro-
as cell with wrap-around. The inter-side distance of Macro-
1  
E1C (θ) = − log e−θR[i] . (12) cells and CPS-cells are 500m.
θ
Since each RB carries b bits, R[i] = b for all i. We therefore (1−ηϕρ)M −l
Clg C(1−ηϕ)ρM −g
obtain frame is (1−ηϕρ)M . Since the maximum value of
C(1−ηϕ)ρM
1   g is min(l, (1 − ηϕ)ρM ), the expected value of g is
E1C (θ) = − log e−bθ = b. (13) min(l,(1−ηϕ)ρM ) Clg C(1−ηϕρ)M −l
θ g (1−ηϕ)ρM −g
. Therefore, among l RBs
g=0 (1−ηϕρ)M
C(1−ηϕ)ρM
To generalize (13), the effective capacity of the SMRM utilized by the CPS-cell, the expected number of RBs
utilizing l RBs with potential interference due to the without suffering interference in each data subframe is
estimation error on the RBs usage of the existing users min(l,(1−ηϕ)ρM ) Clg C(1−ηϕρ)M −l

is given by following theorem. l − g=0 g (1−ηϕ)ρM −g

(1−ηϕρ)M . Further considering
C(1−ηϕ)ρM
Theorem 1: Considering that the SMRM is applied to the overhead of sensing subframes, we can obtain
the CPS-cell, the effective capacity per subframe of a ⎛ ⎞
min(l,(1−ηϕ)ρM )
l (1−ηϕρ)M −l
machine that utilizes l RBs in each data subframe is given T − 1 C g C (1−ηϕ)ρM −g
l
Cl
= ⎝l − g ⎠
(16)
by T (1−ηϕρ)M
g=0 C (1−ηϕ)ρM
ElC (θ) = l
l E1C (l
l θ), (14)
as the average number of RBs without suffering interfer-
where ence in each subframe. By applying the results in [27],
T −1 [28], the effective capacity per subframe that the machine

l = utilizes l RBs in each data subframe without suffering
⎛T min(l,(1−ηϕ)ρM ) ⎞ interference is
(1−ηϕρ)M −l
gClg C(1−ηϕ)ρM −g
· ⎝1 − ⎠ (15)
g=0
(1−ηϕρ)M ElC (θ) = lE1C (lθ). (17)
l · C(1−ηϕ)ρM
l
By substituting l in (17) by l
C in (16), we can obtain
and Cab  b!(a−b)!
a!
. (14) as the effective capacity per subframe of the machine
Proof: Due to the potential estimation error on the that utilizes l RBs in each data subframe.
BRs usage of Macrocells, we need to derive the proba-
bility that there are g RBs suffering interference among
4.2 The Control of T and the RBs Allocation
l RBs utilized by the machine in the data subframe.
For this purpose, we can divide RBs occupied by the Since machines locate at different areas experience dif-
Macrocells into two classes: (i) There are ηϕρM RBs ferent RBs occupations from Macrocells. As a result, ρ,
whose allocations are the same in each subframe. (ii) η, ϕ and thus the effective capacity of a machine may be
There are (1 − ηϕ)ρM RBs whose allocations certainly distinct from that of each other. With the facilitation of
(1−ηϕρ)M
change in each frame. Therefore, there are C(1−ηϕ)ρM Theorem 1, we thus can propose the control of T and the
RBs allocation to achieve the statistical delay guarantees
possibilities of the RBs arrangement of Macrocells in
for each machine, as elaborated in the following.
each subframe. In a data subframe, the CPS-BS only allo-
cates RBs sensed as unoccupied in the sensing subframe
(1−ηϕρ)M −l
to the machine. Therefore, there are Clg C(1−ηϕ)ρM −g 5 P ERFORMANCE E VALUATIONS
possibilities that there are g RBs suffering interference Since the considered CPS is realized based on MTC
among l RBs utilized by the CPS-cell. Therefore, the in LTE-Advanced, we evaluate the performance of the
probability that there are g RBs suffering interference proposed SMRM by adopting system parameters of
among l RBs utilized by the CPS-cell in the data sub- LTE-Advanced [29]. The network deployment for the
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
7

Algorithm 2 The control of T and the RBs Allocation TABLE 1

System parameters and assumptions for simulations
1: The CPS-BS calculates the effective bandwidth EB (θ)
of the real-time traffic. Parameters Values/assumptions
2: To efficiently utilize radio resource, T is initially set Carrier Frequency 2 GHz
to a predetermined value. System bandwidth 20 MHz
(100 RBs per subframe)
3: The CPS-BS first allocates l = 1 RB to the machine
Subframe length 1 ms
in the data subframe. Number of subcarriers in an RB 12
4: The CPS-BS calculates the effective capacity for the Number of OFDM symbols in an RB 7
machine by (14). Diameter of Macrocell coverage 500m
Diameter of CPS-cell coverage 500m
5: Find the solution of θ such that Number of grids per CPS-cell 40
Number of machines per CPS-cell 200
EB (θ) = ElC (θ) = δ. (18) Number of DAs per CPS-cell 10
Cells layout Seven hexagonal-grid
6: Derive the delay violation probability by (as shown in Fig. 4) Macrocells with wrap-
around. Three CPS-cells are
Pr{Delay > dmax } = e−θδdmax (19) deployed at the boundary
of the central Macrocell
7: if e−θδdmax > ε then with wrap-around.
8: l is determined by DAs deployment Uniformly distributed over
each CPS-cell
min {l}, s.t. e−θδdmax ≤ ε for all machines (20) Machines deployment Uniformly distributed over
1≤l≤L each CPS-cell
DL TX power of Macrocells 46 dBm
L is the maximum number of RBs that can be allo- DL TX power of CPS-cells 46 dBm
cated to the machine in a data subframe. Interference threshild -65 dBm
9: if If (20) is not satisfied then Modulation QPSK
Voice/video trace file According to [29], [32]
10: decrease T by one and repeat Steps 4 to 10 to
find the appropriate l and T such that (20) can be
satisfied.

be simultaneously occupied by both the Macrocell

simulation is shown in Fig. 4, where three CPS-cells and the CPS-cell. Under this situation, such an RB
coexist with six Macrocells. Such a deployment is a is identified as with poor channel quality and both
typical coexistence of CPS-cells and Macrocells. CPS- the Macrocell and the CPS-cell select other RBs with
cells are capable of the proposed SMRM and thus the better channel quality. As a result, an appropriate
cognitive radio technology, while the SMRM is not ap- RBs selection scheme is the key for such the state-of-
plied to Macrocells. Details of simulation parameters the-art operation of Macrocells and CPS-cells with-
and assumptions are listed in Table I. This performance out the SMRM. Recently, in [30], [31], an iterative
evaluation is conducted by C++. algorithm known as the Gibbs sampler based radio
In this evaluation, the following two schemes are resources selection is proposed to optimize the se-
selected as performance comparison benchmarks for the lection of RBs. That is, each Macrocell and CPS-cell
SMRM. computes the temperature parameter U = log C(2+u)0
,
2

• Randomize. Without the SMRM, the distributed where u is the age variable denoting the system
interference mitigation scheme widely adopted by time and C0 is an appropriately selected cons tant.
the state-of-the-art OFDMA systems is to randomize Next, each Macrocell computes the energy on every
RBs allocation in each subframe. Such a randomized RB in a subframe as the summation of all inter-
scheme is similar to concept of the interleaved RBs ference from CPS-cells and each CPS-cell computes
allocation to combat the block fading channel. This the energy on every RB in a subframe as the sum-
interference mitigation scheme is widely adopted mation of all interference from Macrocells. Then,
when the radio resources usage of neighboring cells the Macrocell and the CPS-cell select an RB in a
is totally unknown, which is a typical interference subframe according the the Gibbs distribution with
mitigation scheme when the system lacks of the temperature U [30], [31]. To achieve an acceptable
cognitive radio capability. performance, a considerable number of iterations is
• Gibbs Sampler. As mentioned in Section 3.3 that the required. By considering the average performance
state-of-the-art Macrocells as well as CPS-cells have on each iteration, this scheme is adopted to show
the capability of channel quality assessment and the performance of the CPS-cell without the SMRM
avoid to utilize an RB with poor channel quality (i.e., in the interweave coexistence.
severe interference or poor channel gain). However,
without a centralized coordination between Macro- In this simulation, we shall first evaluate the perfor-
cells and CPS-cells and without the SMRM in CPS- mance of the spectrum map construction by compressive
cells, an RB with acceptable channel quality may sensing. Fig. 5 shows the error rate of the spectrum map
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
8

0.17 600
ρ=0.8
0.16 ρ=0.7
ρ=0.6 550
ρ=0.5 SMRM, T=24
η=ϕ=0.8

Effective Capacity (bits/subframe)

0.15
SMRM, T=20
500
Spectrum map construction error rate

SMRM, T=16
0.14
SMRM, T=24
SMRM, T=20
0.13 450
SMRM, T=16
Randomize
0.12
400 Gibbs sampler

0.11
350
0.1 η=ϕ=0.3

0.09
300

0.08 250
0.0001 0.001 0.01 0.1 1
QoS Exponent θ (1/bits)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
q

Fig. 6. Effective capacity of the CPS-cell adopting the

Fig. 5. Error rate of the spectrum map construction under SMRM, the randomized scheme and Gibbs sampler
different q. based RB selection under different T (l = 6, ρ = 0.8),
where the average performance of Gibbs sampler based
RBs selection on each iteration is evaluated.
construction. The error rate is measured by
N M
Υ(m, n)
ξ = n=1 m=1 (21) tion, there could be potential errors on the estimation of
NM
RBs occupation of Macrocells. Therefore, the first thing
where Υ(m, n)=1 if there is an error on the spectrum required to be investigated is whether it is effective to
map construction on the mth RB in a subframe on the perform channel sensing to mitigate interference. It can
nth grid; otherwise, Υ(m, n)=0. We can observe from be observed from Fig. 6 that, when T is appropriately se-
Fig. 5 that, when Macrocells are with a typically high lected, the SMRM outperforms the randomized scheme
traffic load (ρ = 0.8), there is only a 7.5% error (7.5 RBs when RBs are occupied by Macrocell with a high corre-
are occupied by Macrocells but they are estimated as lation among subframes. Even in the extreme case that
unoccupied, or 7.5 RBs are not occupied by Macrocells RBs are occupied by Macrocell with a low correlation,
but they are estimated as occupied, among 100 RBs in the performance of the SMRM is around the same level
a frame) if q ∈ [0.6, 0.8]. Please note that only 60% of of the randomized scheme. These results support the
machines are adopted for channel sensing. This result effectiveness of channel sensing on the SMRM for the
suggests the effectiveness of the SMRM on the spectrum interference mitigation. Since the Gibbs sampler based
map construction. We may particularly note that, in radio resources selection scheme requires a considerable
the compressive sensing, a large q may lead to the number of iterations to achieve an acceptable perfor-
oversampling issue to degrade the performance. mance, we shall evaluate the average performance of the
In Macrocells, there are typically certain correlations Gibbs sampler based radio resources selection scheme on
in RBs allocations among subframes. Since the swarm each iteration. We can also observe that, the SMRM also
suffers aggregated interference from multiple Macro- outperforms the Gibbs sampler scheme when RBs are oc-
cells, there are certain correlations in such aggregated cupied by Macrocell with a high correlation among sub-
interference (RBs occupations) from multiple Macrocells. frames. Even in the extreme case that RBs are occupied
Therefore, we should evaluate the performance of the by Macrocell with a low correlation, the performance
swarm under different correlations of RBs allocations of of the SMRM is around the same level of the Gibbs
Macrocells among subframes. In this performance eval- sampler scheme, which demonstrates the effectiveness
uation, a high correlation is considered by η = ϕ = 0.8 of the proposed SMRM. Furthermore, the performance
and a low correlation is considered by η = ϕ = 0.3. In of the SMRM can be enhanced when T increases, while
the low correlation case, when ρ = 0.8, there are only this performance enhancement becomes marginal when
0.3 × 0.3 × 0.8 × 100 = 8 RBs that will be persistently T keeps increasing. Therefore, by selecting an appro-
occupied by Macrocells in each subframe. This is a very priate T (e.g., 24 subframes), the performance can be
extreme case. Please note that, in Fig. 6 to Fig. 9 shown close to the optimum. Please also note that, due to the
as follows, the error of the spectrum construction, the interweave coexistence between the CPS-cell and the
error on the estimation of the RBs usage of Macrocells Macrocell, when an unoccupied RB is utilized by the
and the error on the estimation of η and ϕ are taken into CPS-cell, Macrocells may not occupy such an RB. As a
the consideration of the performance evaluation. result, the effective capacity of CPS-cell is only affected
In the SMRM, channel sensing is an overhead. In addi- by the sensing and estimation error of RBs usage from
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
9

2000 0.032
SMRM, η=ϕ=0.8

The fraction of l with interference to/from Macrocells

1800 SMRM, η=ϕ=0.3
0.03 T=24, η=ϕ=0.8
1600
Randomize
Effective Capacity (bits/subframe)

T=24, η=ϕ=0.3
1400
0.028 T=20, η=ϕ=0.8
1200
T=20, η=ϕ=0.3
1000 0.026

800
0.024
600

400
0.022
200

0 0.02
0 5 10 15 20
Number of RBs utilized by the CPS−cell
0.018
0 5 10 15 20
Fig. 7. Effective capacity of the machine adopting the Number of RBs utilized by the CPS−cell (l)

SMRM and the randomized scheme under different l.

(T = 24, ρ = 0.8) Fig. 9. The fraction of l with potential interference to/from
Macrocells. (ρ = 0.8)

1100
SMRM, η=ϕ=0.8
1000 SMRM, η=ϕ=0.3 served that when the traffic load of Macrocells increases,
Randomize
900
both the effective capacities of the CPS-cell adopting the
SMRM and adopting the randomized scheme decrease.
Effective Capacity (bits/subframe)

800
The reason is, when the number of RBs occupied by
700 Macrocells increases, the probability that the CPS-cell
600
and the Macrocell utilizing the same RB also increases.
However, since the CPS-cell adopting the SMRM only
500
utilizes the unoccupied RBs, the SMRM outperforms the
400 randomized scheme.
300 Under the proposed SMRM, there could be errors on
200
the spectrum map construction and errors on the estima-
tion of the RBs usage of Macrocells. These errors may
100
0.5 0.6 0.7 0.8 0.9 induce interference to/from Macrocells. Therefore, we
Traffic load of Macrocells (ρ)
shall evaluate whether these estimation errors induces
severe impacts to Macrocells. It can be observed from
Fig. 8. Effective capacity of the machine adopting the Fig. 9 that, these estimation errors only induce a very
SMRM and the randomized scheme under different traffic limited impact to Macrocells. That is, only 2.34% and
loads of the Macrocell ρ. (T = 24, l = 6) 3.13% among l when l RBs are utilized by the CPS-
cell, for η=ϕ=0.8 and η=ϕ=0.3, respectively, as T =24.
When T =20, as expected, the impact can be reduced
the Macrocell and T , which is thus independent of θ. to 1.85% and 2.54% among l for η=ϕ=0.8 and η=ϕ=0.3,
Next, we investigate the performance of the CPS-cell respectively, due to the improvement on the estimation
utilizing different number of RBs in Fig. 7. Please note accuracy of η and ϕ. This result suggests that, by adopt-
that, when l unoccupied RBs are utilized by CPS-cell, the ing the proposed SMRM, the impact induced by the CPS-
worst case of total error rate (jointly considering spec- cell to Macrocells is very limited.
trum map reconstruction error and the estimation error Finally, we focus on the performance of SMRM in
on the RBs usage of Macrocells) can be a constant (as the realistic applications such as healthcare surveillance,
we will see in Fig. 9 later). Therefore, there is a constant traffic monitoring in transportation systems, and envi-
fraction of RBs among l RBs utilized by the CPS-cell ronment monitoring systems, where real-time VoIP and
that may suffer/induce interference from/to Macrocell. video transmissions are necessary. Table 2 shows the sim-
As a result, the effective capacity increases linearly as ulation results of the SMRM on the support of real-time
l increases. Since the CPS-cell adopting the SMRM can VoIP [29] and high quality MPEG4 video transmissions
identify unoccupied RBs to utilize, interference can be [32]. The arrival process of VoIP is the well known ON-
alleviated, especially when the number of RBs required OFF fluid model. The holding times in “ON” and “OFF”
by the machine increases. states are exponentially distributed with means 6.1s and
In Fig. 8, the performance of the CPS-cell under differ- 8.5s, respectively. The data rate of the “ON” state is
ent traffic load of Macrocells is investigated. It can be ob- 32Kbps. The delay bound is 20ms and the delay bound
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
10

TABLE 2
Simulation results of the delay bound violation probabilities of the VoIP (dmax =20ms) and the high quality MPEG4
video (dmax =40ms) transmissions in the swarm (ε=0.02).

Traffic Viol. Prob. Viol. Prob. Viol. Prob. Viol. Prob. Viol. Prob. of Viol. Prob. of Viol. Prob. of Viol. Prob. of
of SMRM of SMRM of SMRM of SMRM Randomize Randomize Randomize Randomize
(simul.) (19) (simul.) (19) (simul.) (19) (simul.) (19)
η=ϕ=0.3 η=ϕ=0.3 η=ϕ=0.8 η=ϕ=0.8 η=ϕ=0.3 η=ϕ=0.3 η=ϕ=0.8 η=ϕ=0.8
Video 0.010 0.0199 0.0066 0.015 0.0464 0.091 0.0331 0.072
VoIP 0.0012 0.003 0.0003 0.0007 0.0013 0.0031 0.0008 0.0017

violation probability is 0.02. The delay bound of MPEG4 [8] K. Doppler, M. Rinne, C. Wijting, C. B. Ribeiro, and K. Hugl,
is 40ms and the delay bound violation probability is 0.02. “Device-to-device communication as an underlay to lte-advanced
networks,” IEEE Commun. Mag., vol. 47, no. 12, pp. 42–49, Dec.
The effective bandwidth calculation of VoIP and video 2009.
traffic can be found by [19], [33]. We can observe from [9] 3GPP TR 23.888 V1.2.0. (2011, Apr.) System improvement
Table 2 that the theoretical results of the delay constraint for machine-type communications. [Online]. Available:
http://www.3gpp.org/ftp/Specs/archive/23 series/23.888/23888-
violation probabilities effectively capture the simulation 120.zip
results, which shows the effectiveness on adopting the [10] S.-Y. Lien, Y.-Y. Lin, and K.-C. Chen, “Cognitive and game-
effective capacity model for providing statistical QoS theoretical radio resource management for autonomous femtocells
with QoS guarantees,” IEEE Trans. Wireless Commun., vol. 10, no. 7,
guarantees. Table 2 also shows the effectiveness of the pp. 2196–2206, July 2011.
SMRM on the support of the statistical delay guarantee, [11] Y.-C. Liang, K.-C. Chen, G. Y. Li, and P. Mahonen, “Cognitive ra-
while the MPEG4 movie can not be supported by the dio networking and communications: An overview,” IEEE Trans.
Veh. Technol., vol. 60, no. 7, pp. 3386–3407, Sept. 2011.
randomized scheme. [12] K. Karenos and V. Kalogeraki, “Traffic management in sensor
networks with a mobile sink,” IEEE Trans. Parallel Distrib. Syst.,
vol. 21, no. 10, pp. 1515–1530, Oct. 2010.
6 C ONCLUSION [13] D. Donoho, “Compressed sensing,” IEEE Trans. Inf. Theory, vol. 52,
no. 4, pp. 1289–1306, Apr. 2006.
In this paper, we resolve the most critical challenge of [14] D. Wu and R. Negi, “Effective capacity: a wireless link model
for support of quality of service,” IEEE Trans. Wireless Commun.,
autonomous radio resource management to provide QoS vol. 12, no. 4, pp. 630–643, July 2003.
guarantees for the swarm in CPS by applying the cogni- [15] A. J. Goldsmith and S. G. Chua, “Variable-rate variable-power
tive radio technology and the distributed radio resource MQAM for fading channels,” IEEE Trans. Commun., vol. 45, no. 10,
pp. 1218–1230, Oct. 1997.
monitoring, which is enabled by the compressive sens- [16] E. Candes, J. Romberg, and T. Tao, “Stable signal recovery from
ing spectrum map. The propose SMRM can be applied incomplete and inaccurate measurements,” Comm. Pure Appl.
into the realistic CPS platform underlying with LTE- Math., vol. 59, no. 8, pp. 1207–1223, Aug. 2006.
[17] J. Haupt, W. U. Bajwa, M. Rabbat, and R. Nowak, “Compressed
Advanced networks and other possible wireless infras- sensing for networked data,” IEEE Signal Process. Mag., vol. 25,
tructure. Performance evaluation results showed that the no. 2, pp. 92–101, Oct. 2008.
proposed SMRM outperforms the randomized scheme to [18] J. Tang and X. Zhang, “Cross-layer modeling for quality of service
guarantees over wireless links,” IEEE Trans. Wireless Commun.,
enable a smooth real-time voice/video transmissions for vol. 6, no. 12, pp. 4505–4512, Dec. 2007.
healthcare surveillance applications or traffic monitoring [19] C.-S. Chang, Performance Guarantees in Communication Networks.
systems over the swarm. Springer, 2000.
[20] 3GPP TS 36.300 V11.0.0, “Evolved universal terrestrial radio
access (E-UTRA) and evolved universal terrestrial radio access
network (E-UTRAN),” Dec. 2011.
R EFERENCES [21] M. A. T. Figueiredo, R. D. Nowak, and S. J. Wright, “Gradient
projection for sparse reconstruction: Application to compressed
[1] E. A. Lee, “Cyber physical systems: Design challenges,” in Proc. sensing and other inverse problems,” IEEE J. Sel. Areas Commun.,
IEEE ISORC 2008, May 2008, pp. 363–369. vol. 1, no. 4, pp. 586–597, Dec. 2007.
[2] R. Rajkumar, I. Lee, L. Sha, and J. Stankovic, “Cyber-physical [22] S. Mendelson, A. Pajor, and N. Tomczak-Jaegermann, “Uniform
systems: the next computing revolution,” in Proc. ACM DAC’10, uncertainty principle for bernoulli and subgaussian ensembles,”
June 2010. Constructive Approximation, vol. 28, no. 3, pp. 277–289, Feb. 2008.
[3] S.-Y. Lien, K.-C. Chen, and Y. Lin, “Toward ubiquitous massive [23] R. Baraniuk, M. Davenport, R. DeVore, and M. Wakin, “A simple
accesses in 3GPP machine-to-machine communications,” IEEE proof of the restricted isometry property for random matrices,”
Commun. Mag., vol. 49, no. 4, pp. 66–74, Apr. 2011. Constructive Approximation, vol. 28, no. 3, pp. 253–263, Feb. 2008.
[4] S.-Y. Lien and K.-C. Chen, “Massive access management for QoS [24] D. S. Bai, “Efficient estimation of transition probabilities in a
guarantees in 3GPP machine-to-machine communications,” IEEE Markov chain,” The Annals of Statistics, vol. 3, no. 6, pp. 1305–
Commun. Lett., vol. 15, no. 3, pp. 311–313, Mar. 2011. 1317, Mar. 1975.
[5] S.-Y. Lien, T.-H. Liau, C.-Y. Kao, and K.-C. Chen, “Cooperative [25] R. M. Gray, Probability, Random Processes, and Ergodic Properties.
access class barring for machine-to-machine communications,” Springer, 2009.
IEEE Trans. Wireless Commun., vol. 11, no. 1, pp. 27–32, Jan. 2012. [26] A. Goldsmith, S. A. Jafar, I. Marić, and S. Srinivasa, “Breaking
[6] F. Xia, L. Ma, J. Dong, and Y. Sun, “Network QoS management spectrum gridlock with cognitive radios: an information theoretic
in cyber-physical systems,” in Proc. ICESS 2008, July 2008, pp. perspective,” Proc. IEEE, vol. 97, no. 5, pp. 894–914, May 2009.
302–307. [27] D. Wu and R. Negi, “Utilizing multiuser diversity for efficient
[7] T. Dillon, V. Potdar, J. Singh, and A. Talevski, “Cyber-physical support of quality of service over a fading channel,” IEEE Trans.
systems: Providing quality of service (QoS) in a heterogeneous Veh. Technol., vol. 54, no. 3, pp. 1198–1206, May 2005.
systems-of-systems environment,” in Proc. IEEE DEST 2011, May [28] J. Tang and X. Zhang, “Cross-layer-model based adaptive re-
2011. source allocation for statistical QoS guarantees in mobile wireless
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
11

networks,” IEEE Trans. Wireless Commun., vol. 7, no. 6, pp. 2318– Kwang-Cheng Chen received his B.S. degree
2328, June 2008. from National Taiwan University in 1983, and
[29] 3GPP TR 36.814 V9.0.0, “Further advancements for E-UTRA M.S. and Ph.D. degrees from the University of
physical layer aspects,” Mar. 2010. Maryland, College Park, in 1987 and 1989, all
[30] S.-M. Cheng, S.-Y. Lien, F.-S. Chu, and K.-C. Chen, “On exploting in electrical engineering. From 1987 to 1998 he
cognitive radio to mitigate interference in macro/femto heteroge- was with SSE, COMSAT, the IBM Thomas J.
neous networks,” IEEE Wireless Commun. Mag., vol. 18, no. 3, pp. Watson Research Center, and National Tsing
40–47, June 2011. Hua University, Hsinchu, Taiwan, working on
[31] F.-S. Chu and K.-C. Chen, “Mitigation of macro-femto co-channel mobile communications and networks. He is a
interference by spatial channel separation,” in Proc. IEEE VTC- Distinguished Professor and the Director at the
Spring, 2011. Graduate Institute of Communication Engineer-
[32] [Online]. Available: http://www.tkn.tu- ing and the Communication Research Center, National Taiwan Univer-
berlin.de/research/trace/trace.html sity. Dr. Chen has received a number of awards and honors, including
[33] B. Maglaris, P. Anastassiou, P. Sen, G. Karlsson, and J. D. Robbins, 2011 IEEE ComSoc WTC Recognition Award and co-authoring 3 IEEE
“Performance models of statistical multiplexing in packet video papers to receive 2001 ISI Classic Citation Award, the IEEE ICC 2010
communications,,” IEEE Trans. Commun., vol. 36, pp. 834–843, Best Paper Award, and IEEE GLOBECOM 2010 GOLD Best Paper.
1988. His research interests include wireless communications and network
science.

Shao-Yu Lien received the B.S. degree in Elec-

trical Engineering from National Taiwan Ocean
University, Taiwan, in 2004, the M.S. degree
in communications engineering from National
Cheng Kung University, Taiwan, in 2006, and
the Ph.D. degree in the Graduate Institute of
Communication Engineering from National Tai-
wan University, Taiwan, in 2011. He received
IEEE ICC 2010 Best Paper Award and his re-
search interests include cognitive/autonomous
technologies and statistical scheduling in wire-
less systems.

Shin-Ming Cheng received B.S. and Ph.D. de-

grees in computer science and information en-
gineering from National Taiwan University, in
2000 and 2007, respectively. He joined the Grad-
uate Institute of Communication Engineering,
National Taiwan University as a postdoctoral
research fellow in 2007. His research inter-
ests include information security, cognitive radio
networks, and network science.

Sung-Yin Shih received B.S. and M.S. de-

grees in department of electrical engineering
and graduate institute of communication engi-
neering from National Taiwan University, Taipei,
in 2008 and 2011, respectively. Her research
interests include compressed sensing, routing,
and cognitive radio networks.
Received: 22 January 2019 Revised: 13 March 2020 Accepted: 17 March 2020
DOI: 10.1111/caim.12376

REGULAR ARTICLE

Building an understanding of how winning products emerge

when open and proprietary products coexist: Evidence from
the RepRap community

Michael A. Stanko

Poole College of Management, North Carolina

State University, Raleigh, NC, USA Online innovation communities have altered the nature of collaborative innovation.
Within these communities, coexistence of open and closed source offerings is
Correspondence
Michael A. Stanko, Poole College of becoming commonplace, though potential diffusion and product advantages from
Management, North Carolina State University,
each form are not well understood. Patterns of derivative innovation within these
Raleigh, NC 27695-7229, USA.
Email: [email protected] communities affect designers' focus; thus, this work is grounded in the attention-
based view. Beyond open vs. closed source development, we find that the presence
of sibling designs (designs based on the same source material) and self-remix (itera-
tion on material by the same designer) have notable diffusion and product effects.
Diffusion effects are investigated using 354 co-existing open and closed source 3D
printers from the RepRap community, while a subset of these printers is used for an
analysis of key product attributes: value and ease of use. While previous researchers
have argued for an early stage open source diffusion advantage, this is not observed
here. However, customers perceive open source products to have value advantages,
while closed source offerings are easier to use. Sibling designs have a diffusion
advantage, particularly early on. Self-remixes have both diffusion and product advan-
tages. By better understanding these contextual elements of derivative innovation,
designers' attention can be shaped to achieve desired outcomes.

KEYWORDS

3D printing, attention-based view, closed vs. open source, innovation communities, open
innovation

1 | I N T RO DU CT I O N Community members innovate in ways not available to an isolated

party, promoting cumulative, derivative innovation. Online innovation
Within online innovation communities, ideas are changed, expanded communities represent a unique context for derivative innovation
and combined (i.e., remixed) by community members who share an given the level of transparency, designer diversity and the ease of
interest, but differ in expertise, age, geography and even language electronic communication. Firms are keen to participate in and even
(Faraj, von Krogh, Monteiro, & Lakhani, 2016; Flath, Friesike, Wirth, & host these platforms as a form of inbound open innovation.
Thiesse, 2017). A member of an online innovation community—which Recently, within online innovation communities, coexistence
may be an interested hobbyist or a firm's employee—is able to remix a between open source and proprietary products has become more
source innovation developed by another designer without direct con- common. In communities such as Android and RepRap, there have
tact. Remixing is an increasingly popular digital concept; a form of been an increasing number of instances of closed source appropria-
knowledge reuse in which existing designs are adapted into something tion of community efforts. Many firms have their genesis in open
new (Friesike, Flath, Wirth, & Thiesse, 2019; Stanko, 2016). source communities (e.g., Red Hat, Sun Microsystems, MakerBot);

Creat Innov Manag. 2020;1–15. wileyonlinelibrary.com/journal/caim © 2020 John Wiley & Sons Ltd 1
2 STANKO

some “fork off” a closed source version of a community's open source community is likely to shape derivative innovations and their market
contributions (Ciesielska & Westenholz, 2016). For instance, when success. This research will guide firms participating in these communi-
Amazon developed the Kindle Fire operating system, it began with an ties to best ensure product diffusion and desirable product character-
open source version of Android and developed a closed source deriva- istics based on their choice of source material within the community
tive to enable Amazon to control how apps would interface with the (i.e., community context). Here, three contextual elements are concep-
operating system and users' exposure to Amazon services tualized: closed vs. open source, sibling designs and self-remix.
(Appleyard & Chesbrough, 2017; Pon, Seppälä, & Kenney, 2014). Simi- A designer's choice to develop a closed vs. open source design
larly, MakerBot's earliest products were open source 3D printers likely has implications for the product developed and its diffusion. For
based on designs from the RepRap community. Over time the com- instance, past scholars have argued that open source products have a
pany shifted to proprietary development to differentiate from other cost advantage (Dahlander & Gann, 2010), given their ability to har-
RepRaps and protect their intellectual property (West & Kuk, 2016). ness community input. But, the open vs. closed source development
Coexistence of open and closed source development within innova- decision is only part of the context of derivative innovation; the pat-
tion communities is a relatively new phenomenon. Scholars have terns of related innovations within the community that are formative
recently noted hybrid business strategies and switching tactics to the products developed. Some remixed innovations have many sib-
(i.e., open to closed or vice versa) as firms attempt to reap the learning ling designs (designs derived from the same source, i.e., parent, mate-
benefits of a community while exerting control (Appleyard & rial), allowing designers to view changes by others working from the
Chesbrough, 2017). Ciesielska and Westenholz (2016) outline a con- same source material, but also potentially creating differentiation chal-
tinuum of open source involvement, ranging from imitating open lenges. Sibling designs act as visible, parallel experiments leading to
source ideas (low involvement) to combining open and proprietary discoveries that all designers working from the source material may
development (mid) to managing an open source community (high). benefit from (Flath et al., 2017). Sibling designs can enable designers
This is primarily a study of the market diffusion of products to develop solutions beyond their own technical skills, empowering
emerging from an online innovation community where closed and remixed products that could not otherwise be developed. Thus, the
open source innovations coexist. We also conduct an analysis into number of sibling designs is an important contextual element of deriv-
two product attributes: value and ease of use (Figure 1). While we do ative innovations within these communities, acting as an unconven-
not conduct a complete (i.e., mediated) test of the nomological net- tional conduit for learning. While others have speculated on the
work of constructs contained in Figure 1, we note that an existing lit- innovative potential of sibling designs, we push our understanding for-
erature connects the outcomes of interest from the second set of ward by showing meaningful market and product outcomes. Finally,
hypotheses and empirical tests (value/ease of use) with the outcome community members have the option to remix the work of others or
of interest from the first set (diffusion). The constructs are placed in to use their own prior work as source material (i.e., self-remixing). A
Figure 1 in their logical, causal order consistent with this literature designer refining their own prior work focuses attention very differ-
(e.g., Davis, 1989; Venkatesh & Bala, 2008). The fundamental logic ently when contrasted with a designer searching out a starting point
that this paper advances is that an innovation's context within the across an online community. Thus, self-remixing is the final contextual

F I G U R E 1 Manuscript overview
Constructs are positioned left to right
based on their logical causal order.
For simplicity, only direct effects are
shown. While this study does not
investigate the relationships between
value/ease of use and market
diffusion, we note that an established
literature connects these constructs
(e.g., Davis, 1989; Venkatesh & Bala,
2008)
STANKO 3

element of derivative innovations we examine. The open innovation as well as the debate regarding diversity vs. expertise in
(e.g., Chesbrough, 2003) and crowdsourcing (e.g., Kohler, 2015) litera- crowdsourcing and ideation (Hong & Page, 2004; Tetlock, Mellers,
tures suggest the designer's decision to consult external sources Rohrbaugh, & Chen, 2014).
(or not) will impact innovative outcomes. Next, the contextual elements are more thoroughly discussed,
With these elements introduced, our research question becomes: followed by hypotheses. Diffusion effects are investigated via regres-
in innovation communities where open and closed source products sion, then product attributes are examined using fuzzy set qualitative
coexist, how do the contextual elements of derivative innovation comparative analysis (fsQCA) for a sample subset. The article ends
(open vs. closed source, number of sibling designs, and self-remix) with discussion and limitations.
influence products' attributes and market diffusion? Given our focus
on community members' attention, we ground this research in the
attention-based view, which argues that behavior is the result of deci- 2 | CONT EX TUAL E L EM ENT S OF
sion makers' focus of attention (Ocasio, 1997). Abundant information D E R I V A T I V E I N N O V A T I O N WI T H I N ON L I N E
within online innovation communities makes focusing designers' INNOVATION COMMUNITIES
attention a challenge: scattered attention in an information-rich envi-
ronment affects the ability to prioritize and absorb information Within the context of online innovation communities where open and
(Dahlander & Frederiksen, 2012). Online communities act as an exer- closed source products coexist, we view three elements as affecting
cise in stimulus-driven attentional selection (Ocasio, 2011). The designers' attention and ability to harness the potential benefits of
actions of designers depend on their situated attention, “the particular community collaboration, ultimately impacting the success of the
context or situation they find themselves in” (Ocasio, 1997, p. 188), products developed: open vs. closed source development, sibling
here dictated by the contextual elements, which shift design priorities, designs, and self-remix.
information processing, and problem solving.
Based on a sample of 354 3D printers from the RepRap commu-
nity, we show that these contextual elements meaningfully impact 2.1 | Open vs. closed source development
product diffusion. Given the challenge of assembling a comparable set
of open and closed source products, the diffusion of open vs. closed As mentioned above, there have been an increasing number of
sourced products has mainly been compared previously through simu- instances of closed source appropriation of community efforts. In one
lation and case study (e.g., Almirall & Casadesus-Masanell, 2010; online innovation community this became such a common occurrence
Bonaccorsi & Rossi, 2003). This study represents a rare investigation that the community's founder wrote an essay discussing his reasons
of the market diffusion of a large number of comparable open and for being “rather lax” in pursuing those who appropriate community
closed source products. In attempting to build a more nuanced under- “technology and close it off” (Bowyer, 2011). The coexistence of open
standing, we also draw on data available only for a subset of these and closed source developments can be less than harmonious
printers to explore customer perceptions of two distinct product out- (O'Mahony & Bechky, 2008); it can be an emotional issue, as when a
comes: value (perceived benefits relative to perceived sacrifice) and company co-founder discussed his shame at the company's shift from
ease of use (the extent that using the product is effortless). Open open to closed source development (West & Kuk, 2016). Within these
source research suggests a consumer surplus (Lerner & Tirole, 2002), communities, it is generally at the discretion of the innovator whether
thus value is a logical outcome to examine. Research also suggests a they license their innovation as closed or open source. This said, some
tradeoff in open source development related to ease of use license types preclude remixing an open source innovation into a
(Appleyard & Chesbrough, 2017). While a growing body of research closed source offspring (this is known as a share-alike license), though
(mainly case and simulation based), exploring the outcomes of open enforcement of these licenses is inconsistent, particularly within
vs. closed development gives us a starting point in thinking about online communities (Felin & Zenger, 2014).
these questions, researchers' understanding of the implications of the Company involvement with open source development has been
decision to pursue open vs. closed source development is limited. widespread (Bonaccorsi, Giannangeli, & Rossi, 2006; Henkel, 2008;
There has been little substantiation of these performance effects from Höst & Oručevic-Alagic, 2011), though many firms struggle with the
studies drawing on samples allowing for statistical analysis, marking balance between participating in value creation and attempting to
the first research gap we set out to fill. The effects of the other two appropriate meaningful gains (Schröder & Hölzle, 2010). Intellectual
contextual factors thought to impact designers' attention (sibling property protection within innovation communities is limited (Felin &
designs and self-remix) also represent relevant, timely, unanswered Zenger, 2014) and there are many unresolved questions with respect
theoretical questions we set out to address. While there has been to open source business models (Harison & Koski, 2010; Massa,
speculation regarding the innovative potential of sibling designs, here Tucci, & Afuah, 2017; Vanhaverbeke & Chesbrough, 2014).
we fill another acknowledged research gap by developing an under- Researchers have begun to ask under what conditions firms will
standing of this remix pattern's connection to product success (Flath switch between open and closed source development (Appleyard &
et al., 2017, p. 322). Our findings related to self-remixing inform dis- Chesbrough, 2017; Hepp & Henkel, 2017), though this research has
cussions regarding the tensions created by openness (Lauritzen, 2017) not yet examined switching behavior within innovation communities.
4 STANKO

Reusing modular components is commonplace in open source Montgomery, 1988). While there are certainly transparency-enabled
development (Haefliger, von Krogh, & Spaeth, 2008; MacCormack, benefits to innovating within a community (e.g., learning; Jeppesen &
Baldwin, & Rusnak, 2012), which is thought to lower costs Frederiksen, 2006), transparency also enables the proliferation of sim-
(Fitzgerald, 2006). Beyond this, the implications of developing open ilar offerings (Hill & Monroy-Hernandez, 2012). There are many open
and closed source products within an online innovation community questions regarding the effects of sibling designs, and we view this
are not well understood, though this choice should affect a designer's construct as imperative to better understanding remixing as a contem-
attention during the development process. Open vs. closed source porary form of collaborative innovation.
development differentially affects how a design will draw from and
contribute to the community, as well as openness to having a design
altered by others in the future. Past investigations have largely been 2.3 | Self-remix
limited to open source software; open hardware has received decid-
edly less attention (Höst & Oručevic-Alagic, 2011). There has been Here, we use remix to mean the “use of an existing innovation as
relatively little direct comparison of customers' preference for open source material or inspiration to aid in the development of further
vs. closed source offerings—researchers are limited by scant instances innovations” (Stanko, 2016, p. 773). The iterative nature of product
of directly comparable products in each category (see Paulson, Succi, & development within online innovation communities shapes the prod-
Eberlein, 2004). Accordingly, researchers have looked to qualitative ucts that are ultimately developed, with many community members
approaches and simulation-based studies. For instance, Almirall and (including participating firms) choosing existing designs as a starting
Casadesus-Masanell's (2010) simulation finds that openness leads to point for further development. Typically, social norms within innova-
superior market performance when product complexity is low. tion communities favor crediting source material; in many cases
Bonaccorsi and Rossi's (2003) simulation shows that network exter- designers encourage others to remix their work.
nalities may explain the diffusion of open source products. Self-remix refers to the designer's choice to iterate based on their
Economides and Katsamakas (2006) show that application variety own previous source material. Designers may iterate on their own
may be enhanced for open source products. West and Kuk (2016) previous designs, but participation of other designers drastically shifts
suggest that proprietary development while harnessing user- designer attention. Self-remixing shows designers' commitment to
generated content may have performance potential. The case studies refining their own ideas (Ocasio, 2011 refers to “attentional vigilance”)
in O'Mahony & Bechky (2008) show how boundary organizations as opposed to exploring the community. From an open innovation
enable collaboration between open source communities and firms perspective (Chesbrough, 2003), self-remixing can be considered a
pursuing proprietary development. form of inbound closedness, even within an online innovation commu-
nity. A balance between inbound openness and closedness has been
shown to optimize innovative performance (e.g., Belderbos, Faems,
2.2 | Sibling designs Leten, & van Looy, 2010; Salter, Wal, Criscuolo, & Alexy, 2015). This
may favor a hybrid approach; for instance, a designer monitoring
Whereas much of the content developed in online communities is developments elsewhere within an innovation community but rem-
ignored, a smaller set of material tends to be heavily remixed (Flath ixing their own prior work.
et al., 2017; Hill & Monroy-Hernandez, 2012), as the community's
attention collectively shifts to particular source material. How does a
designer's choice to begin from source material that is heavily remixed 3 | HY P O T H E S E S
affect the product developed and its diffusion? Given the characteris-
tic transparency of online innovation communities, the coexistence of Rather than predicting an overall diffusion advantage for open or
multiple offspring from the same source innovation allows designers closed source products, we argue for timing-based advantages for
to view, critique, and learn from other variants (see Jeppesen & each form of development. Open source innovations will diffuse
Molin, 2003). Sibling designs can bring alternative possibilities to the more quickly early in the life of a community since customers value
attention of designers, potentially enabling a rich information not being locked into a closed (i.e., proprietary) solution early, when
exchange as a series of “experiments running in parallel” unfolds technological shifts are expected (Almirall & Casadesus-Masanell,-
(Appleyard & Chesbrough, 2017, p. 313). 2010; West, 2003). Early on, when the technological trajectory is
Sibling designs can lead to related discoveries from which all not predictable, but the likelihood of newly arriving products is high,
interested designers may benefit (Flath et al., 2017). Conversely, a consumers rationally fear being trapped in a proprietary standard
large number of siblings may scatter designers' attention, hampering (Appleyard & Chesbrough, 2017; West, 2003). Open source devel-
their ability to refine their own designs. Attention implies withdrawal opment is practical early on, distributing efforts across community
from some stimuli to focus on another (Bitgood, 2000; Ocasio, 1997), members and reusing components to lower the time and effort
which is complicated by a sibling rich environment. With respect required for each product's development (Appleyard &
to product diffusion, there may be differentiation challenges for Chesbrough, 2017). Thus, open source developers' tendency to pro-
designs with a large number of similar siblings (Lieberman & duce modular designs (MacCormack et al., 2012) should lead to
STANKO 5

superior open source offerings early, aiding market diffusion. Sim- We theorize that self-remixes have a market diffusion advantage.
plicity is thought to enhance the market performance of open Continuous designer attention devoted to multiple generations of a
source products (Almirall & Casadesus-Masanell, 2010), which is product will lead to a more refined product. Designers are able to
more likely early in the life of a community, before complexity learn from consumers and incorporate this learning into the next gen-
becomes a major concern. eration. Self-remixers' ability to incorporate feedback from users of
Creativity has also been shown to be fostered by open source prior designs should result in improvements in terms of usability, all-
projects (Paulson et al., 2004); which may help open source products owing the designer to address bugs and customer pain points, and
diffuse early in the life of the community. Later in the life of an online generally improving customer response. Self-remixing also enables
innovation community, the market favors refinement over creativity. development tactics such as empathic research (Leonard &
By this point, customer preferences are established and it may be eas- Rayport, 1997), in which the designer observes users (e.g. while
ier to penetrate the market using a closed source product with supporting previous designs) to more completely understand the con-
unshared differentiation advantages. The refinement associated with sumer experience when developing future improvements. While
closed source solutions (where developers needn't concern them- designers may experiment with a first-generation design, they are
selves with the ability to reuse components) may aid diffusion later as more likely to focus attention on refining their design in future
customer expectations shift toward a more seamless experience generations—and do so with a more established set of development
(Appleyard & Chesbrough, 2017). The relative autonomy afforded skills, fostering product diffusion. Self-remix may allow designers to
through closed source development (Ciesielska & Westenholz, 2016) leverage the discovery advantages of the community (see Almirall &
will prove more impactful on market success and the realization of Casadesus-Masanell, 2010; Leminen, Turunen, & Westerlund, 2015),
value (Chesbrough, Lettl, & Ritter, 2018) later in the life of a while executing their own coherent, controlled vision (Appleyard &
community. Chesbrough, 2017).
While we believe that self-remixing will have diffusion advan-
H1. Open source status is more positively related to product diffusion tages, there are also potential counterarguments here; self-remix has
early in the life of an online innovation community (when com- the effect of continuously harnessing one designer's attention, per-
pared to late). haps forsaking beneficial aspects of community development. The
presence of diversity within a crowd has long been thought beneficial
Innovations with many sibling designs are thought to have diffu- (Hong & Page, 2004; Ollila & Yström, 2016). Problem solving and bug
sion advantages for several reasons. First, designs that are based on fixes are aided by drawing on a sizable and diverse community; these
parent designs which have been widely remixed (i.e., those designs are sought after characteristics of crowdsourcing tactics (Boudreau &
having numerous siblings) are likely compatible with customer expec- Lakhani, 2013). Self-remixes may be limited by the designer's fixation
tations (an acknowledged contributor to diffusion; Rogers, 2003). (the propensity to focus on a limited set of their own ideas; Agogué,
Importantly, given the transparency of online communities, numerous Le Masson, Dalmasso, Houdé, & Cassotti, 2015) and there may be
experiments in parallel based on the same source material will benefit flexibility tradeoffs to sustained attentional vigilance (Ocasio &
designers. Learning is an important benefit of participation in innova- Wohlgezogen, 2010). Despite these counterarguments, we expect
tion communities (Lakhani & Wolf, 2005; Stanko, 2016). Designers self-remixing to lead to more refined designs, which will more widely
will benefit from adapting or adopting improvements seen elsewhere diffuse.
based on the same source material, as well as by avoiding mistakes
they observe. H3. Self-remixing is positively related to product diffusion.
While a large number of sibling designs will prove helpful during
early periods of quickly advancing community knowledge, sibling
designs may also complicate efforts to differentiate from an increased
number of offerings later in the life of the community. Online trans- 3.1 | Product outcomes
parency makes the proliferation of unoriginal offerings likely (Hill &
Monroy-Hernandez, 2012). Given the potential for mass proliferation As outlined in Figure 1, we now hypothesize effects on two product
of user-generated content (see Droge, Stanko, & Pollitte, 2010), a attributes: value and ease of use. Open source offerings have advan-
mature community will have developed large numbers of variant tages marshaling resources and building shared community support
products. (Dahlander & Gann, 2010; Henkel, 2006). Typically, open source
development efforts are funneled into activities that can be used
H2A. The number of sibling designs is positively related to product across multiple products (Baldwin & Clark, 2006; Haefliger
diffusion. et al., 2008; MacCormack et al., 2012), which allows for cost reduc-
tion (Appleyard & Chesbrough, 2017; Fitzgerald, 2006). These factors
H2B. The number of sibling designs is more positively related to product should result in users perceiving superior open source value.
diffusion early in the life of an online innovation community (when Members of an innovation community likely have higher toler-
compared to late). ance for ease of use challenges compared with mainstream users
6 STANKO

(West & Kuk, 2016), tolerating (even enjoying) requirements for In cases where a designer iterates on her own work (i.e., self-remix),
lengthy, technical steps on the part of the user. Given that some open this commitment to her own prior work can be considered a form
source offerings are developed for the designer's own use (e.g., von of closedness, even within an innovation community. A balance
Krogh, Spaeth, & Lakhani, 2003), ease of use concerns likely garner between open and closedness may foster beneficial product out-
less attention from these designers. There is also a tendency in open comes (e.g., Laursen & Salter, 2006; Salge, Farchi, Barrett, &
source projects to continually and unnecessarily add complexity Dopson, 2013). In this instance, self-remix enables the designer's
(Paulson et al., 2004), complicating the user experience. Closed source isolated attention, as opposed to being distracted by stimuli across
development is likely to be focused on one specific product with a tar- the community (Bitgood, 2000; Ocasio, 2011). Multiple, distinct
get consumer in mind, rather than a user-innovator, allowing for a designers may unnecessarily complicate products. Loosely connected
more tailored user experience with improved ease of use open source developers have been shown to extraneously lengthen
(Appleyard & Chesbrough, 2017). code over time (Scacchi, 2006). Conversely, self-remixers are more
likely to pursue a unified vision without needless cost and complex-
H4A. Open source status is positively related to value. ity. Constancy of attention by the same designer through multiple
product generations may enable usability and value improvements
H4B. Open source status is negatively related to ease of use. that are unlikely absent this constancy of attention. The designer of
a self-remix is in the uncommon position to have the opportunity to
Sibling designs are also thought to impact product outcomes, though observe users working with the previous generation product. Note
this again represents an open theoretical question. First, sibling devel- that ease of use concerns are not often mentioned by users online
opments expose designers to new approaches and technologies (Flath (Droge et al., 2010), perhaps due to concerns regarding status (see
et al., 2017). Siblings allow for a rich learning experience through Fleming & Waguespack, 2007) and fear of being perceived as tech-
exposing designers to diverse viewpoints based on shared subject nically incompetent. Without observing customers using the previ-
matter. We expect this learning to result in the development of ous generation product, learning benefitting the next generation
remixed products with favorable product outcomes (in this case, value product's ease of use is unlikely.
and ease of use) as the community of designers go through a collec-
tive process of continuous improvement of the same source material H6A. Self-remixing is positively related to value.
(Friesike et al., 2019). Beyond improvements based in designers' learn-
ing, there are other logical reasons to expect superior product ratings H6B. Self-remixing is positively related to ease of use.
for products with numerous sibling designs. Given that consumers
may be familiar with prominent parent designs (those which are
widely remixed), it seems reasonable to expect that designs remixed 4 | METHODS
from these prominent parents may benefit from this familiarity. Logi-
cally, consumers may find that products that are (even somewhat) 4.1 | Context: the RepRap community
familiar tend to be easier for them to use (Rogers, 2003). This is part
of what Tan, Miao, and Tan (2020) refer to as a remixed product's The context chosen to investigate these questions is the RepRap (rep-
“inheritance” from its parent. With respect to value, it also seems rea- licating rapid prototyper) project, an online community devoted to
sonable that the availability of sibling designs could potentially lead to developing 3D printers capable of printing most of their own compo-
price competition since sibling products may vie for the same poten- nents (RepRap.org). Early RepRaps were intended to be customized
tial customers. It logically follows that this price rivalry should aid cus- and improved in the future, allowing community members to tailor
tomers' perception of value. designs to their own tastes and increasing the total population of Rep-
Counterarguments exist here, too; visible attentional Raps (Jones et al., 2011). An evolutionary underpinning is firmly
alternatives within an online innovation community have the poten- established in the RepRap community, with early printers named after
tial to scatter designers' focus (Dahlander & Frederiksen, 2012; prominent biologists (e.g., Darwin, Mendel), preceding the develop-
Ocasio, 2011; Stam, 2009). Processing information from legions of ment of hundreds of different, functioning designs. 3D printers have
sibling offerings is likely to befog designer attention (Bitgood, 2000; been described as one of the first open design efforts to garner wide-
Ocasio, 1997), lessening focus on their own product's value and spread commercial success (West & Kuk, 2016). Within the RepRap
usability. Despite these counterarguments, consistent with our ear- community both firms and individual hobbyists participate, similar to
lier hypothesis of diffusion benefits from sibling products (H2A), we open source software communities (Shah, 2006). More generally, 3D
predict positive product outcomes from having a large volume of printing has been noted as a powerful enabler of entrepreneurship
sibling designs: (Gartner, Maresch, & Fink, 2015). Several prominent 3D printing firms
have their roots in the RepRap project, including Makerbot, Ultimaker,
H5A. The number of sibling designs is positively related to value. and Prusa Research. While the RepRap community largely embraces
open source development, there are numerous examples in which
H5B. The number of sibling designs is positively related to ease of use. RepRap technology has been applied to closed source designs
STANKO 7

(Bowyer, 2011), providing an opportunity to contrast these two forms siblings from all parents are counted. Open source refers to
of development. whether the printer makes use of a license specifying open source
To test the hypotheses put forward, we coupled two distinct hardware and software (1) or not (0). Open source RepRaps typi-
datasets. The first dataset (the RepRap Family Database) includes cally publicly post CAD models and use freely available source
information on printers within the RepRap community and those code, while closed source offerings do not (closed source n = 93).
derived from community technology. These data were compiled by an Community age is the number of days since the online posting of
active member of the RepRap community (Rene Mueller) based on the first RepRap until the focal printer was posted online. De novo
documentation posted by designers as well as his own judgments is used as a control variable indicating printers with no credited
when needed. Once nonfunctional printers and those with incomplete parent printer (1), as opposed to derivative works (0); 278 (79 per-
information are removed, the sample includes 354 3D printers cent) of these printers are classified as derivative works. While
released between 2006 and 2015. This is matched with additional most RepRaps have a clear lineage from their predecessors, de
data from a 3D printing intermediary (3D Hubs; www.3dhubs.com) novo printers do not have a clear connection to prior source work.
which gathers reviews from registered 3D print providers using their Two product outcome ratings are obtained from 3D Hubs; ratings
platform, which we webscraped during May 2017. For the most popu- are only available for printers appearing on 3D Hubs' list of the most
lar (i.e., commonly reviewed) printers, customer ratings of their popular printers. These ratings are based on responses to 3D Hubs'
printers are available. 3D Hubs' list of printers is cumulative; once online survey of their registered platform users who own the printer
printers achieve the necessary number of member reviews (5) they in question. As part of 3D Hubs' survey, registered printer owners are
are not removed. asked to rate their printers on several attributes using single item
scales. Value measures the extent that the printer's owners perceive
the product to be great value for money, or conversely to be over-
4.2 | Variable descriptions priced. Ease of use measures the extent that the printer is straightfor-
ward to operate, or conversely, is very complicated.
Each variable (see Table 1) will be discussed in turn. Diffusion is
operationalized as a binary variable: whether (1) or not (0) the
printer appears on 3D Hubs' list of the most popular printers. Due 5 | RE SU LT S
to the blend of commercial and non-commercial pursuits examined
here in the form of open and closed source products, this type of To test H1–H3, we make use of 3D Hubs' list of the most popular 3D
non-traditional diffusion measure is needed for accurate, compara- printers. Thirty-two of the 354 printers from the RepRap database are
ble tracking (Paulson et al., 2004). Since these printers are a mix of listed among the most popular 3D printers (9 pecent). While only a
self-replicating non-commercial products and more traditional relatively small portion of RepRap printers are listed on 3D Hubs' list
stand-alone commercial offerings, traditional sales figures are of the most popular 3D printers, it is noteworthy that this constitutes
unavailable. Given this, we perform two distinct validity checks nearly half the printers on 3D Hubs' list (32 of 71 listed printers;
(Appendix A) which support the validity of this measure. Limita- 45 percent). This speaks to the substantial commercial impact that
tions of this binary variable are discussed in the final section. Self- products from the RepRap community have had on the 3D printing
remix refers to whether the 3D printer is derived from a parent marketplace. It is interesting to consider whether the coexistence of
printer with the same designer (1) or not (0). Sibling designs is a open and closed source developments has been of benefit to the
count of the number of printers sharing the same parent printer. products collectively developed by the community given the likely
In cases in which multiple parent printers are specified (n = 42), motivational and skillset differences between these developers (see

TABLE 1 Descriptive statistics and Pearson correlations

Mean St. dev. n Min Max 1. 2. 3. 4. 5. 6. 7. 8.

1. Diffusion .090 .287 354 0 1 1
2. Open source .737 .441 354 0 1 −.058 1
3. Sibling designs 14.90 19.18 354 0 100 .0089 .087 1
4. Self-remix .172 .378 354 0 1 .14 ***
.0004 −.16*** 1
5. Community age 2270.4 425.9 354 0 3,146 −.032 −.17 ***
.12** .0022 1
6. De novo .214 .411 354 0 1 .099 *
−.22 ***
−.41 ***
−.24 ***
−.030 1
7. Value 8.19 1.05 32 5.5 9.3 -- .59*** .24 −.021 −.16 −.30 1
8. Ease of use 7.21 1.18 32 5.3 9.7 -- −.46 ***
−.29 .14 .62 ***
.14 −.078 1
*
p < .10
**
p < .05
***
p < .01
8 STANKO

West & Kuk, 2016). While this question is outside of our focus here, it on diffusion earlier in the life of an online innovation community—
will be developed as a future research direction. and conversely have a less positive effect on diffusion later in the
life of the innovation community. Overall, the models explain a
substantial portion of the variance in these printers' ability to dif-
5.1 | Regression analysis fuse (Model 3 Cragg and Uhler's R2 = .15). All variance inflation
factors (VIFs) are well below the standard guideline of 10, indicating
Given our binary dependent variable, we use logistic regression to test that multicollinearity does not pose a serious threat to interpreting
H1–H3 (see Table 2) on the full sample of printers (n = 354). All con- results (highest VIFs: model 1 = 1.41, model 2 = 6.90, model
tinuous variables used in interaction terms are mean centered for ease 3 = 1.47).
of interpretation. In model 1, only main effects are included. Models Next, to provide greater perspective in interpreting these results
2 and 3 sequentially incorporate the interaction terms; model we examine the probability of inclusion in 3D Hubs' list of the most
2 includes the interaction of open source status and community age popular printers across levels of independent variables, using Stata's
to test H1, model 3 includes the interaction of sibling designs and margins command (Williams, 2012), based on model 3. For printers
community age to test H2B. that are not self-remixes, with all other variables at their means, the
In model 1 we observe a significant positive effect of sibling probability of inclusion in 3D Hubs' list is 5.0 percent. For printers that
designs (p < .01), supporting H2A. This effect is also observed in are self-remixes, the probability of inclusion is 27.8 percent; self-
models 2 and 3 (consistently p < .01). The number of sibling designs is remixes are over five times as likely to be included in the 3D Hubs' list
positively related to the printer's market diffusion. Model 1 also indi- (consistent with H3). Next, we examine the effect of the number of
cates a positive (p < .01) effect of self-remixing on diffusion siblings on the probability of inclusion, at one standard deviation
(supporting H3), which is consistently observed in models 2 and below and above the mean age of the community at the time of the
3. Printers that are self-remixed are significantly more likely to achieve printer's release (1,845 and 2,696 days into the life of the community,
market diffusion than those that are not. respectively). This takes into account both the relevant direct and
The hypothesized interaction effect (H1) between open source interaction effects, holding other variables at their mean values.
status and community age is not significant (model 2). The Inspecting Figure 2, siblings have a decidedly more positive effect on
(unhypothesized) main effect of open source status is also non-sig- diffusion early in the life of the community (consistent with H2B).
nificant. Open source status does not significantly affect product In addition to testing these hypotheses, we also examined individ-
diffusion directly, nor through a moderated relationship with com- ual designers to detect switches from open to closed source develop-
munity age. Model 3 shows a negative interaction effect between ment (and vice-versa). Among 37 designers with at least two designs,
the presence of sibling designs and community age (p < .01), four switch from closed to open or from open to closed. While this
supporting H2B. That is, sibling designs have a more positive effect provides some evidence of the switching behavior categorized by

TABLE 2 Binary logistic regression analysis predicting market diffusion

Model 1 2 3
1 = included in list of most popular
D printers 0 = not included Main effects only Open interaction (H1) Sibling interaction (H2B)
Open source −.315 (.397) −.224 (.441) −.293 (.397)
Sibling designs .0261*** (.00856) .0263*** (.00850) .0246*** (.00955)
Self-remix 1.948*** (.525) 1.954*** (.525) 2.001*** (.541)
Community age −.000328 (.000336) .000559 (.00102) −.00114** (.000523)
De novo 1.865*** (.564) 1.824*** (.564) 1.886*** (.577)
Open source × Community age — .00104 (.00107) —
Sibling designs × Community age — — −.0000837*** (.0000307)
2
McFadden's R .089 .093 .119
2
Cragg & Uhler's R .116 .121 .153
χ2 19.18 23.54 23.28
df 5 6 6
n 354 354 354
*
p < .10
**
p < .05
***
p < .01
Unstandardized coefficients shown. Robust standard errors in parentheses.
STANKO 9

F I G U R E 2 Probability of inclusion in list of

most popular 3D printers by number of sibling
designs [Colour figure can be viewed at
wileyonlinelibrary.com]

Appleyard and Chesbrough (2017), it also shows that (in this context) standardizing; Rubera & Kirca, 2017). fsQCA is then conducted
switching behavior is uncommon, with the large majority (89 percent) (Table 3) using Stata's fuzzy command (Longest & Vaisey, 2008). A
of designers (i.e., both firms and individuals) not reconsidering their value of 1 represents full membership in the set, 0 represents full
choice to pursue open or closed source development. non-membership in the set. Consistency refers to the degree to which
cases having the same combination of conditions demonstrate the
outcome. Two consistency thresholds must be exceeded for a config-
5.2 | Fuzzy set qualitative comparative analysis uration to be viewed as a true solution here. First, the standard con-
sistency cutoff of 0.7 is used for all analyses (Ragin, 2008a). Second,
To examine product outcomes (H4–6), we use the subset of 32 3D the configuration's consistency with the outcome in question must be
printers for which product outcome ratings are available through 3D significantly greater (p < .05 via a Wald test) than its consistency with
Hubs. Because of this reduced sample size, we use fuzzy set qualita- the negation of that outcome (Longest & Vaisey, 2008). Coverage
tive comparative analysis (fsQCA), which has emerged as a widely measures the portion of cases of interest explained by the configura-
used tool in innovation research (e.g., Curado, 2018; Gilbert & tion, which is somewhat analogous to regression's R2 (Ragin, 2008b).
Campbell, 2015; Stanko, 2016) and other disciplines over the past For both outcomes, there is one consistent explanatory configura-
decade. As opposed to regression's emphasis on the discrete effect of tion. Note that this is not always the case in fsQCA; it is possible to
individual variables, fsQCA uses Boolean algebra to detect combina- have many explanatory configurations for an outcome (or none at all).
tions of predictors that consistently lead to the specified outcome Printers rated highly in terms of value are consistently (.894) open
(Ragin, 2008a; Ragin & Rubinson, 2011). fsQCA's emphasis on contex- source self-remixes with few sibling designs (consistent with H4A and
tual combinations and low sample size requirements make it an ideal H6A, though contradicting H5A). Printers rated highly in terms of ease
methodology to apply here. of use are very consistently (.969) closed source self-remixes with few
We use fsQCA to identify the contexts that consistently result in sibling designs (consistent with H4B and H6B, though contradicting
3D printers that are rated highly in terms of value and ease of use. H5B). Viewing fsQCA results across these two outcomes together,
Given the need to ensure a meaningful number of observations within configurations including open source are associated with both a bene-
configurations, we limit the specified conditions to the three contex- fit to consumers (value, consistent with H4A) and a drawback (reduced
tual elements: open source, sibling designs, and self-remix. Before per- ease of use, consistent with H4B). Configurations including sibling
forming fsQCA, continuous variables need to be transformed into designs are associated with drawbacks in terms of both product rat-
fuzzy set scores, ranging from 0 to 1 (via rank ordering and ings (lessened perception of value, contradicting H5A and lessened
perception of ease of use, contradicting H5B). Finally, configurations
TABLE 3 fsQCA results: Product outcomes including self-remix are consistently associated with positive customer
Outcome Value Ease of use outcomes (value, supporting H6A and ease of use, supporting H6B).
We now connect these fsQCA findings logically to the earlier
Open source • Θ
regression results. First, it does seem plausible that open source sta-
Sibling designs Θ Θ
tus' mixed relationships with product outcomes (superior value–
Self-remix • •
consistent with H4A, but inferior ease of use–consistent with H4B)
Consistency 0.894 .969
combine for a non-significant open source effect on diffusion, as
Coverage 0.197 .157
observed here in the regression results (H1). The findings related to
Note: • = causal condition (present); Θ = causal condition (absent). self-remix product advantages also map logically to the previous
10 STANKO

finding of a positive effect of self-remix on diffusion. Given that con- comparable, coexisting open and closed source products—fill substan-
figurations including self-remix are associated with both value (consis- tial, acknowledged gaps in our understanding of open vs. closed
tent with H6A) and ease of use (consistent with H6B), it is source products' market performance (Ciesielska & Westenholz, 2016)
unsurprising that these products should more widely diffuse (H3). We and more generally build our understanding of how innovators should
also observe that configurations including a small number of sibling draw on community resources to optimally create value (Chesbrough
designs are associated with desirable product outcomes (contradicting et al., 2018).
H5A and H5B). This appears to lend credence to the possibility that Our findings related to self-remixing and sibling designs also have
sibling designs distract designers' attention from a focus on core attri- implications for managers pursuing innovation through online commu-
butes such as value and usability. Given that the effect of sibling nities as well as for the platforms hosting these communities (Agogué,
designs differs as the community matures (H2B) and that there appear Yström, & Le Masson, 2013). Self-remixing does have a positive effect
to be attention-based product benefits from a lack of sibling designs, on market diffusion (as well as both value and ease of use); the learn-
it is clear that sibling designs play a nuanced role in affecting products ing that happens through consistent attention to multiple generations
and their success (discussed further in the following section), and that of customers pays off in more refined self-remixes. Finally, the num-
further inquiry is warranted here. ber of sibling designs is critical: early in the life of the community,
these are helpful to market diffusion, aiding learning and fostering
benefits from the online community. However, later in the life of the
6 | DISCUSSION community sibling designs do not foster market diffusion, instead act-
ing as barriers to differentiation. Consistent with attention-based
This research contributes to the literature on online innovation com- arguments (Ocasio, 1997), it seems that the scattering of attention
munities by showing that the context that derivative innovations are resulting from a sibling-rich design environment results in diminished
developed within meaningfully impacts ratings of these products and product value and usability. Innovators pursuing community-based
their market success. The growing coexistence of open and closed tactics should consider their priorities and consciously shape
source development renders these findings even more relevant to designers' context within the community, since this materially impacts
both managers and scholars. Sibling designs are a contemporary con- their products and market success. While a large fraction of open
duit for knowledge exchange within online innovation communities innovation research has been at the level of the firm (Chesbrough,
whose presence (particularly early in the life of the community) is Vanhaverbeke, & West, 2006), these insights are made possible
shown to aid products' market diffusion, though we also observe that through a rare examination of a large number of products emerging
configurations with small numbers of sibling designs are associated from the same online innovation community.
with value and ease of use. Further, the choice to self-remix has impli-
cations for designers' attention and their ability to refine products;
self-remix is associated with both market diffusion and favorable 6.1 | Open vs. closed source
product characteristics. Theoretically, consistent with the attention-
based view (Ocasio, 1997), we show that the consistency or scattering A key question since the rise of open source development processes
of designers' attention within these communities meaningfully impacts has been how such efforts compete with (or complement) more tradi-
the products developed and their success. tional proprietary approaches. This work addresses a research gap
The unprecedented investigation of the market diffusion of a related to the diffusion of open vs. closed source products and
large sample of coexisting open and closed source designs also marks responds to Appleyard and Chesbrough's, (2017, p. 319) call to inves-
a contribution to the literatures on innovation communities and open tigate the competitive payoffs to openness. Coexistence of open and
source development, particularly in the context of open hardware closed source products has been investigated through simulation-
(Höst & Oručevic-Alagic, 2011). While many firms struggle with the based study (Almirall & Casadesus-Masanell, 2010; Bonaccorsi &
decision to pursue open vs. closed source development (West & Rossi, 2003), and conceptual and case-based research has considered
Kuk, 2016), this decision is not shown here to have a significant the benefits and costs of each form of development (e.g., Dahlander &
impact on market diffusion—a noteworthy finding given the implica- Gann, 2010; von Hippel & von Krogh, 2003; West, 2003). However,
tions of this and the lack of past examination. Further, the data exam- prior research lacks larger scale comparison of open and closed source
ined do not provide evidence of timing-based advantages of either efforts competing in the same market, likely due to the difficulty in
form of development. However, we do find distinct product out- finding a large number of commensurable efforts in the same category
comes: open source products are associated with value advantages, and time period (see Paulson et al., 2004, pp. 248–249). Here, we
while closed-source products are associated with being easier to use. show that neither open nor closed source development has a signifi-
While we argued (H1) that there may be advantages to open source cant market diffusion advantage. Given the tensions within firms try-
development early, there may also be offsetting disadvantages; for ing to harness the benefits of an open source community while
instance, Ciesielska and Westenholz (2016) argue that open source protecting their intellectual property, our findings are meaningful for
involvement can initially constrain a company as it learns to comply managers: the open vs. closed source development decision does not
with community norms. These findings—drawing on a large sample of significantly impact product diffusion.
STANKO 11

The lack of a diffusion advantage early in the life of an innovation point for further discussion of the attentional and market dynamic
community contradicts Appleyard and Chesbrough's (2017) argument (i.e., pricing, familiarity) effects of sibling designs.
(in this setting) that open source products will have an advantage It seems advisable for platforms to harness this sibling diffusion
attracting customers early on. Within the RepRap community, the effect, for instance by initiating innovation remix contests in which all
switching behaviors discussed by Appleyard and Chesbrough are also entries draw on common source material. Purposely shifting designers'
rare. Importantly, we do observe meaningful differences in terms of attention to learning from the work of others while remixing the same
attributes of the products developed through open and closed source source material – again, early in the maturation of a community – may
development. Configurations involving open source designs are asso- be one way to leverage this diffusion advantage from sibling designs
ciated with superior value, consistent with previous speculation (relatedly, Dahlander & Piezunka, 2014 discuss the notion of proactive
(e.g., Fitzgerald, 2006). Open source printers are also associated with attention). Similarly, seeking out commonly remixed source material is
lower ease of use, which may be an outcome of modular approaches advisable for companies interested in developing mass market prod-
to open source development (Haefliger et al., 2008; MacCormack ucts early in maturity of the community, but not later on.
et al., 2012). It also seems plausible that designers releasing closed
source products are more in tune with their target customers, as
opposed to user innovators, designing products for customers like 6.3 | Self-remixing
themselves, with a high tolerance for ease of use challenges. Consis-
tent with the attention-based view, shifting designer attention to a Self-remixing shows a constancy of designer attention and results in
target customer (as opposed to designing to “scratch their own itch”; a more refined, easier to use product, which the market warmly
von Krogh, Haefliger, Spaeth, & Wallin, 2012, p. A6), affects priorities receives. Our findings related to self-remixing are internally consis-
and the products ultimately developed. Viewed together, our findings tent in that self-remix has positive effects on both product ratings as
regarding the outcomes of open and closed source development are well as market diffusion. Self-remixing can be thought of as an ele-
logically consistent. There appears to be a tradeoff in consumer rat- ment of closed inbound innovation within a (largely open) online
ings of open source products (improved value but lower ease of use). innovation community, as designers draw from their own earlier work
This tradeoff logically explains the non-significant effect of open as opposed to elsewhere within the community. Our findings add to
source status on market diffusion. research from the open search literature that illustrate performance
drawbacks from an over-reliance on external sources (e.g., Belderbos
et al., 2010; Salter et al., 2015). This finding also contributes to the
6.2 | Sibling designs recent discussion regarding the desirability of elite problem solvers
vs. the benefits of involving a diverse group of disconnected problem
Sibling designs promote diffusion, particularly early in the life of an solvers (Hong & Page, 2004; Tetlock et al., 2014) by showing that
online innovation community when related efforts can most mean- attentional vigilance in the form of self-remix leads to the develop-
ingfully provide creative impetus to designers. While prior research ment of winning new products without directly involving other mem-
has hinted at the possibilities from related discoveries through sib- bers of the community.
ling designs (notably Flath et al., 2017), this work fills a gap in our For hosts of online innovation communities, it may seem intuitive
understanding of innovation communities by establishing meaningful to harness the diversity of the crowd by encouraging designers to use
market and product impacts from sibling designs. Sibling designs are the work of others as a starting point, encouraging “cross-pollination”
a conduit for designers to learn from others' experimentation, which (see Ollila & Yström, 2016). However, our results consistently show
are more advantageous early in the life a community, when the that self-remixed products have both product and diffusion advan-
focus is on learning, experiments in parallel, and creativity. Con- tages. Online innovation platforms should encourage self-remixing
versely, later in the life of an online innovation community, when where possible (as should firms active in these communities). Simple
the pace of meaningful innovation has slowed but products prolifer- tactics such as prompting members a few weeks after they post a
ate, sibling designs harm diffusion as designers find it difficult to dif- design, to see if they have considered an alternative approach to their
ferentiate from a crowded field of “me-too” offerings. Numerous earlier innovation, may prod refinement through self-remix.
sibling designs also scatter designers' attention away from refining
value and ease of use. Logically consistent with our findings con-
cerning sibling designs' diffusion effects, deficits in terms of value 6.4 | Limitations and next steps
and usability should become more harmful to diffusion as the com-
munity matures; the earliest adopting consumers are less concerned While the data employed here allow for noteworthy insights regarding
with these attributes given their heightened tolerance for ease of product attributes and diffusion, there are of course limitations. For
use concerns and willingness to pay (Lee, Ryu, & Kim, 2010; instance, all derivative works are considered equivalent here, where
Rogers, 2003). Overall, it is clear that sibling designs play a nuanced clearly there is a range of derivative innovation within online innova-
role in determining product outcomes and diffusion, which shifts tion communities, from slight modifications to complete reimagining
over time. This study plays an important role by providing a starting (Flath et al., 2017; Friesike et al., 2019). A more nuanced examination
12 STANKO

of the depth of derivative innovations would prove fruitful. It is also numbers of observations, often case based. Certainly, future
possible for similar designs to emerge that are not siblings. Under- researchers should consider the rich possibilities from assembling a
standing potential learning differences between sibling designs and data set capable of considering both product attributes and market
unrelated, similar designs may prove interesting. There are also inter- success across a large sample of comparable open and closed source
esting possibilities in bringing different theoretical lenses to the inves- products. This would allow, for instance, estimation of a model
tigation of derivative works and remixing. For instance, actor-network where the contextual factors' direct and indirect diffusion effects
theory (see Sage, Vitry, & Dainty, 2020), with its consideration of (i.e., mediated and/or moderated by product attributes) are
human and non-human actors and emphasis on the affective force of observed, allowing for an even richer understanding of the role of
technologies may prove interesting to apply within the remixing con- community context. The fsQCA findings reported here should only
text where online community members draw on source material be interpreted with regard to the entire configuration, rather than
developed by other designers, without the necessity of direct interac- drawing conclusions with regard to isolated variables.
tion. From an attention-based perspective, it may also be interesting While there remain important questions for researchers in this
to examine how characteristics of the stimulus (i.e., object) and the area to address, it is clear that a design's context within an online
designer's motivation interact (see Ocasio, 2011). It would also prove innovation community affects both the product developed and its dif-
worthwhile to take a more macro perspective to contrast communities fusion. That said, in this study (drawing on a large-scale examination
with and without the coexistence of open and closed source products of comparable open and closed source 3D printers), the decision to
we see here. Are there benefits (or drawbacks) to the entire eco- pursue open vs. closed source development does not significantly
system from this coexistence? While emotional clashes between open affect market diffusion.
and source developers have been noted (West & Kuk, 2016), do
skillset and motivational differences across open and source devel- OR CID
opers have positive effects for the entire community? Perhaps com- Michael A. Stanko https://orcid.org/0000-0001-9177-4121
parative case analysis across innovation communities would allow
insights into these larger scale questions. ENDNOTES
In our context, switching behavior (open to closed source or vice- 1
Given the range of commercial motivations (and lack thereof) within
versa) is seen to be relatively uncommon. Does this hold elsewhere? online innovation communities, we use the term ‘product’ to refer to both
commercial and non-commercial offerings.
There will continue to be interesting questions around switching
2
(Appleyard & Chesbrough, 2017), as well as exploring hybrid forms of Appleyard and Chesbrough (2017) make, but do not statistically test, a
similar argument.
innovation between open and closed source (Bonaccorsi et al., 2006;
3
A “family tree” based on the RepRap Family Database is available at:
Ciesielska & Westenholz, 2016; Leminen et al., 2015). The mixed and
http://reprap.org/wiki/RepRap_Family_Tree
shifting effects of sibling designs should also draw further research
inquiry. Are there other moderators beyond community maturity
(e.g., designer routines) that alter the effect of sibling designs? Here, RE FE RE NCE S

our focus was within online innovation communities where open and Agogué, M., Le Masson, P., Dalmasso, C., Houdé, O., & Cassotti, M. (2015).
Resisting classical solutions: The creative mind of industrial designers
proprietary products coexist, but questions around the effects of sib-
and engineers. Psychology of Aesthetics, Creativity, and the Arts, 9,
ling designs in other contexts (e.g., crowdsourcing contests) may also 313–318.
prove interesting. How can designers capture the learning benefits Agogué, M., Yström, A., & Le Masson, P. (2013). Rethinking the role of
from sibling designs without suffering the scattering of attention and intermediaries as an architect of collective exploration and creation of
knowledge in open innovation. International Journal of Innovation Man-
harmful diffusion effects that are observed here once the innovation
agement, 17(2), 1–24.
community matures? There may also be reasons why designers Almirall, E., & Casadesus-Masanell, R. (2010). Open versus closed innova-
choose to self-remix which cannot be accounted for here. tion: A model of discovery and divergence. Academy of Management
Our investigation of market diffusion effects across a large Review, 35, 27–47.
Appleyard, M. M., & Chesbrough, H. W. (2017). The dynamics of open
(n = 354), comparable sample of open and closed source offerings is
strategy: From adoption to reversion. Long Range Planning, 50,
unprecedented (to our knowledge). However, limitations of our data 310–321.
must also be discussed. The use of a binary variable for market dif- Baldwin, C. Y., & Clark, K. B. (2006). The architecture of participation:
fusion is limiting; there are likely meaningful market penetration dif- Does code architecture mitigate free riding in the open source devel-
opment model? Management Science, 52, 1116–1127.
ferences between printers classified here as diffused. As sources
Belderbos, R., Faems, D., Leten, B., & Van Looy, B. (2010). Technological
become available, more fine-grained measures may help researchers activities and their impact on the financial performance of the firm:
further understand differences in the ability to capture value across Exploitation and exploration within and between firms. Journal of Prod-
open and closed source development. We are also limited to a rela- uct Innovation Management, 27, 869–882.
tively small number of observations of reasonably widely diffused Bitgood, S. (2000). The role of attention in designing effective interpretive
labels. Journal of Interpretation Research, 5(2), 31–45.
products where product ratings are available (n = 32), though this
Bonaccorsi, A., Giannangeli, S., & Rossi, C. (2006). Entry strategies under
must be seen against the backdrop of existing comparisons of open competing standards: Hybrid business models in the open source soft-
and closed source products, typically based on much smaller ware industry. Management Science, 52, 1085–1098.
STANKO 13

Bonaccorsi, A., & Rossi, C. (2003). Why open source software can succeed. Henkel, J. (2006). Selective revealing in open innovation processes: The
Research Policy, 32, 1243–1258. case of embedded Linux. Research Policy, 35, 953–969.
Boudreau, K. J., & Lakhani, K. R. (2013). Using the crowd as an innovation Henkel, J. (2008). Champions of revealing—The role of open source devel-
partner. Harvard Business Review, 91(4), 60–69. 140 opers in commercial firms. Industrial and Corporate Change, 18,
Bowyer, A. (2011). RepRap and Open Source. Retrieved from reprap. 435–471.
org/wiki/RepRap_and_Open_Source Hepp, D., & Henkel, J. (2017). Why young firms change their level of open-
Chesbrough, H. (2003). Open innovation: The new imperative for creating ness. Paper presented at the Academy of Management Meeting.
and profiting from technology. Boston, MA: Harvard Business Press. Hill, B. M., & Monroy-Hernandez, A. (2012). The remixing dilemma: The
Chesbrough, H., Lettl, C., & Ritter, T. (2018). Value creation and value cap- trade-off between generativity and originality. American Behavioral Sci-
ture in open innovation. Journal of Product Innovation Management, 35, entist, 57, 643–663.
930–938. Hong, L., & Page, S. E. (2004). Groups of diverse problem solvers can out-
Chesbrough, H., Vanhaverbeke, W., & West, J. (2006). Open innovation: perform groups of high-ability problem solvers. Proceedings of the
Researching a new paradigm. Oxford: Oxford University Press. National Academy of Sciences, 101, 16385–16389.
Ciesielska, M., & Westenholz, A. (2016). Dilemmas within commercial Höst, M., & Oručevic-Alagic, A. (2011). A systematic review of research on
involvement in open source software. Journal of Organizational Change open source software in commercial software product development.
Management, 29, 344–360. Information and Software Technology, 53, 616–624.
Curado, C. (2018). Human resource management contribution to innova- Jeppesen, L. B., & Frederiksen, L. (2006). Why do users contribute to firm-
tion in small and medium-sized enterprises: A mixed methods hosted user communities? The case of computer-controlled music
approach. Creativity and Innovation Management, 27, 79–90. instruments. Organization Science, 17, 45–63.
Dahlander, L., & Frederiksen, L. (2012). The core and cosmopolitans: A Jeppesen, L. B., & Molin, M. J. (2003). Consumers as co-developers: Learn-
relational view of innovation in user communities. Organization Sci- ing and innovation outside the firm. Technology Analysis & Strategic
ence, 23, 988–1007. Management, 15, 363–383.
Dahlander, L., & Gann, D. M. (2010). How open is innovation? Research Jones, R., Haufe, P., Sells, E., Iravani, P., Olliver, V., Palmer, C., &
Policy, 39, 699–709. Bowyer, A. (2011). RepRap—The replicating rapid prototyper. Robotica,
Dahlander, L., & Piezunka, H. (2014). Open to suggestions: How organiza- 29, 177–191.
tions elicit suggestions through proactive and reactive attention. Koçaş, C., & Bohlmann, J. D. (2008). Segmented switchers and retailer
Research Policy, 43, 812–827. pricing strategies. Journal of Marketing, 72(3), 124–142.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and Kohler, T. (2015). Crowdsourcing-based business models. California Man-
user acceptance of information technology. MIS Quarterly, 13, agement Review, 57(4), 63–84.
319–340. Lakhani, K. R., & Wolf, R. G. (2005). Why hackers do what they do. In
Droge, C., Stanko, M. A., & Pollitte, W. A. (2010). Lead users and early J. Feller, B. Fitzgerald, S. Hissam, & K. R. Lakhani (Eds.), Perspectives on
adopters on the web: The role of new technology product blogs. Jour- free and open source software (Vol. 1) (pp. 3–22). Cambridge, MA: MIT
nal of Product Innovation Management, 27, 66–82. Press.
Economides, N., & Katsamakas, E. (2006). Two-sided competition of pro- Lauritzen, G. D. (2017). The role of innovation intermediaries in firm-
prietary vs. open source technology platforms and the implications for innovation community collaboration: Navigating the membership para-
the software industry. Management Science, 52, 1057–1071. dox. Journal of Product Innovation Management, 34, 289–314.
Faraj, S., von Krogh, G., Monteiro, E., & Lakhani, K. R. (2016). Online com- Laursen, K., & Salter, A. (2006). Open for innovation: The role of openness
munity as space for knowledge flows. Information Systems Research, in explaining innovation performance among UK manufacturing firms.
27, 668–684. Strategic Management Journal, 27, 131–150.
Felin, T., & Zenger, T. R. (2014). Closed or open innovation? Problem solv- Lee, H., Ryu, J., & Kim, D. (2010). Profiling mobile TV adopters in college
ing and the governance choice. Research Policy, 43, 914–925. student populations of Korea. Technological Forecasting and Social
Fitzgerald, B. (2006). The transformation of open source software. MIS Change, 77, 514–523.
Quarterly, 30, 587–598. Lehmann, D. R., Keller, K. L., & Farley, J. U. (2008). The structure of
Flath, C. M., Friesike, S., Wirth, M., & Thiesse, F. (2017). Copy, transform, survey-based brand metrics. Journal of International Marketing, 16(4),
combine: Exploring the remix as a form of innovation. Journal of Infor- 29–56.
mation Technology, 32, 306–325. Leminen, S., Turunen, T., & Westerlund, M. (2015). The grey areas
Fleming, L., & Waguespack, D. M. (2007). Brokerage, boundary spanning, between open and closed in innovation networks. Technology Innova-
and leadership in open innovation communities. Organization Science, tion Management Review, 5(12), 6–18.
18, 165–180. Leonard, D., & Rayport, J. F. (1997). Spark innovation through empathic
Friesike, S., Flath, C. M., Wirth, M., & Thiesse, F. (2019). Creativity and pro- design. Harvard Business Review, 75(6), 102–113.
ductivity in product design for additive manufacturing: Mechanisms Lerner, J., & Tirole, J. (2002). Some simple economics of open source. Jour-
and platform outcomes of remixing. Journal of Operations Management, nal of Industrial Economics, 50, 197–234.
65, 735–752. Lieberman, M. B., & Montgomery, D. B. (1988). First-mover advantages.
Gartner, J., Maresch, D., & Fink, M. (2015). The potential of additive Strategic Management Journal, 9(SI), 41–58.
manufacturing for technology entrepreneurship: An integrative tech- Longest, K. C., & Vaisey, S. (2008). fuzzy: A program for performing quali-
nology assessment. Creativity and Innovation Management, 24, tative comparative analyses in Stata. Stata Journal, 8, 79–104.
585–600. MacCormack, A., Baldwin, C., & Rusnak, J. (2012). Exploring the duality
Gilbert, B. A., & Campbell, J. T. (2015). The geographic origins of radical between product and organizational architectures. Research Policy, 41,
technological paradigms: A configurational study. Research Policy, 44, 1309–1324.
311–327. Massa, L., Tucci, C., & Afuah, A. (2017). A critical assessment of business
Haefliger, S., von Krogh, G., & Spaeth, S. (2008). Code reuse in open model research. Academy of Management Annals, 11, 73–104.
source software. Management Science, 54, 180–193. Ocasio, W. (1997). Towards an attention-based view of the firm. Strategic
Harison, E., & Koski, H. (2010). Applying open innovation in business strat- Management Journal, 18(SI), 187–206.
egies: Evidence from Finnish software firms. Research Policy, 39, Ocasio, W. (2011). Attention to attention. Organization Science, 22,
351–359. 1286–1296.
14 STANKO

Ocasio, W., & Wohlgezogen, F. (2010). Attention and control. In quality of debate. Current Directions in Psychological Science, 23,
S. B. Sitkin, L. B. Cardinal, & K. M. Bijlsma-Frankema (Eds.), Organiza- 290–295.
tional control (pp. 191–221). Cambridge, Cambridge University Press. Vanhaverbeke, W., & Chesbrough, H. (2014). A classification of open inno-
Ollila, S., & Yström, A. (2016). Exploring design principles of organizing for vation and open business models. In H. Chesbrough,
collaborative innovation. Creativity and Innovation Management, 25, W. Vanhaverbeke, & J. West (Eds.), New frontiers in open innovation (–
363–377. 50, 68). Oxford: Oxford University Press.
O'Mahony, S., & Bechky, B. A. (2008). Boundary organizations: Enabling Venkatesh, V., & Bala, H. (2008). Technology acceptance model 3 and a
collaboration among unexpected allies. Administrative Science Quar- research agenda on interventions. Decision Sciences, 39, 273–315.
terly, 53, 422–459. von Hippel, E., & von Krogh, G. (2003). Open source software and the “pri-
Paulson, J. W., Succi, G., & Eberlein, A. (2004). An empirical study of open- vate-collective” innovation model: Issues for organization science.
source and closed-source software products. IEEE Transactions on Soft- Organization Science, 14, 209–223.
ware Engineering, 30, 246–256. von Krogh, G., Haefliger, S., Spaeth, S., & Wallin, M. W. (2012). Carrots
Pon, B., Seppälä, T., & Kenney, M. (2014). Android and the demise of oper- and rainbows: Motivation and social practice in open source software
ating system-based power. Telecommunications Policy, 38, 979–991. development. MIS Quarterly, 36, 649–676.
Ragin, C. C. (2008a). Redesigning social inquiry. Chicago, IL: University of von Krogh, G., Spaeth, S., & Lakhani, K. R. (2003). Community, joining, and
Chicago Press. specialization in open source software innovation: A case study.
Ragin, C. C. (2008b). User's guide to fuzzy-set/qualitative comparative analy- Research Policy, 32, 1217–1241.
sis. Tucson, AZ: University of Arizona. West, J. (2003). How open is open enough? Melding proprietary and open
Ragin, C. C., & Rubinson, C. (2011). Comparative methods. In B. Badie, source platform strategies. Research Policy, 32, 1259–1285.
D. Berg-Schlosser, & L. Morlino (Eds.), International encyclopedia of West, J., & Kuk, G. (2016). The complementarity of openness: How
political science. London: Sage. MakerBot leveraged Thingiverse in 3D printing. Technological Forecast-
Rogers, E. M. (2003). Diffusion of innovations (5th ed.). New York: Free ing and Social Change, 102, 169–181.
Press. Williams, R. (2012). Using the margins command to estimate and interpret
Rubera, G., & Kirca, A. H. (2017). You gotta serve somebody: The effects adjusted predictions and marginal effects. Stata Journal, 12(2),
of firm innovation on customer satisfaction and firm value. Journal of 308–331.
the Academy of Marketing Science, 45, 741–761.
Sage, D., Vitry, C., & Dainty, A. (2020). Exploring the organizational prolif-
eration of new technologies: An affective actor-network theory. Orga- AUTHOR BIOGRAPHY
nization Studies, 41, 345–363.
Salge, T. O., Farchi, T., Barrett, M. I., & Dopson, S. (2013). When does sea-
rch openness really matter? A contingency study of health-care inno- Michael A. Stanko ([email protected]) is Associate Professor
vation projects. Journal of Product Innovation Management, 30,
of Marketing and Innovation at the Poole College of Management
659–676.
Salter, A., Wal, A. L., Criscuolo, P., & Alexy, O. (2015). Open for ideation: at North Carolina State University. His research interests include
Individual-level openness and idea generation in R&D. Journal of Prod- new product development, specifically innovation outside tradi
uct Innovation Management, 32, 488–504. tional firm boundaries. Recent research has examined online inno
Scacchi, W. (2006). Understanding open source software evolution. Soft-
vation communities, 3D printing and crowdfunding. He has
ware Evolution and Feedback: Theory and Practice, 9, 181–205.
Schröder, A., & Hölzle, K. (2010). Virtual communities for innovation: Influ- published previously in the Journal of Product Innovation Manage
ence factors and impact on company innovation. Creativity and Innova- ment, Research Policy, MIT Sloan Management Review, Information
tion Management, 19, 257–268. Systems Research, R&D Management, the Journal of the Academy of
Setia, P., Bayus, B., & Rajogopalan, B. (2020). The takeoff of open source
Marketing Science and other outlets. Mike has also written a small
software: A signaling perspective based on community activities. MIS
Quarterly, Forthcoming. series of case studies and notes for marketing managers, available
Shah, S. K. (2006). Motivation, governance, and the viability of hybrid through Ivey Publishing.
forms in open source software development. Management Science, 52,
1000–1014.
Stam, W. (2009). When does community participation enhance the perfor-
mance of open source software companies? Research Policy, 38,
How to cite this article: Stanko MA. Building an
1288–1299.
Stanko, M. A. (2016). Toward a theory of remixing in online innovation understanding of how winning products emerge when open
communities. Information Systems Research, 27, 773–791. and proprietary products coexist: Evidence from the RepRap
Susarla, A., Oh, J.-H., & Tan, Y. (2012). Social networks and the diffusion of community. Creat Innov Manag. 2020;1–15. https://doi.org/
user-generated content. Information Systems Research, 23, 23–41.
10.1111/caim.12376
Tan, J., Miao, D., & Tan, Q. (2020). Empirical study on influencing factors
of knowledge product remixing in OIC. IEEE Access, 8, 34215–34224.
Tetlock, P. E., Mellers, B. A., Rohrbaugh, N., & Chen, E. (2014). Forecasting
tournaments: Tools for increasing transparency and improving the
STANKO 15

APPENDIX A: | Diffusion validity checks languishes in obscurity (Setia, Bayus, & Rajogopalan, 2020; Susarla,
Oh, & Tan, 2012). For the 32 printers appearing on 3D Hubs' list the
Since we employ a non-traditional measure of market diffusion, we mean presence score is 1.68, compared to 1.06 for those printers not
perform two distinct validity checks to ensure that the diffusion mea- appearing (one tailed p = .0006). The significant relationship between
sure relates to other markers as we would logically expect. the diffusion measure and experts' perception of the products' market
First, we recruited three employees of a campus 3D printing lab, presence supports the validity of the diffusion measure employed.
each with at least 1 year of employment experience who are also Logically, a more widely diffused product should also have a more
active in online communities dedicated to 3D printing. These 3D established online presence. As a second validity check, we performed
printing experts were each presented with the names of the Google searches of the name of each 3D printer (in quotation marks
354 printers in our sample (in random order) and asked to respond to to ensure an exact match), plus “3D printer” and recorded the number
“I often encounter this 3D printer” on a five-point scale ranging from of results available for each search, following Koçaş and
1 (strongly disagree) to 5 (strongly agree; Lehmann, Keller, & Bohlmann's (2008) link popularity measure. For printers on 3D Hubs'
Farley, 2008). The three raters' scores were closely correlated (.62, list, the mean number of search results was 309,591 compared to
.60, .41; all p < .0001) and showed strong reliability: α = .78; κ = .31 72,332 for those printers not on 3D Hubs' list (one tailed p = .009).
(p = .0000); thus the three scores were averaged to make up each Given that the printers on 3D Hubs' list (on average) have more than
printer's presence score. As expected, the experts had not encoun- four times the number of search results as printers not appearing, this
tered many of the 354 printers in our sample; the mean presence again supports the validity of the diffusion measure employed.
score for all printers is 1.11. Much content generated online
 International Journal of Production Research

ISSN: 0020-7543 (Print) 1366-588X (Online) Journal homepage: http://www.tandfonline.com/loi/tprs20

3D printing system: an innovation for small-scale

manufacturing in home settings? – early adopters
of 3D printing systems in China

Qingfeng Wang, Xu Sun, Sue Cobb, Glyn Lawson & Sarah Sharples

To cite this article: Qingfeng Wang, Xu Sun, Sue Cobb, Glyn Lawson & Sarah Sharples (2016) 3D
printing system: an innovation for small-scale manufacturing in home settings? – early adopters of
3D printing systems in China, International Journal of Production Research, 54:20, 6017-6032, DOI:
10.1080/00207543.2016.1154211

To link to this article: https://doi.org/10.1080/00207543.2016.1154211

Published online: 01 Mar 2016.

Submit your article to this journal

Article views: 672

View related articles

View Crossmark data

Citing articles: 4 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tprs20
International Journal of Production Research, 2016
Vol. 54, No. 20, 6017–6032, http://dx.doi.org/10.1080/00207543.2016.1154211

3D printing system: an innovation for small-scale manufacturing in home settings? – early

adopters of 3D printing systems in China
Qingfeng Wanga, Xu Sunb*, Sue Cobbc, Glyn Lawsonc and Sarah Sharplesc
a
Department of Quantitative and Applied Economics, Nottingham University Business School China, Ningbo, China; bDepartment of
Mechanical, Materials and Manufacturing Engineering, University of Nottingham Ningbo China, Ningbo, China; cHuman Factors
Research Group, University of Nottingham, University Park, Nottingham, UK
(Received 19 March 2015; accepted 1 February 2016)

This study investigates Chinese consumers’ adoption of the innovative 3D printing systems for small-scale manufactur-
ing in home settings. Empirical studies were conducted in a survey with 256 participants. The number of significant
determinants that affect an individual’s decision to adopt 3D printing systems has been identified by applying a model
that integrates the Technology Acceptance Model and Innovation Diffusion Theory. A number of moderation effects of
demographic variables (e.g. gender, design background) on the association between motivational variables and partici-
pants’ adoption have also been analysed with factor analysis, structural equation modelling and hierarchical regression.
Our results shed some light on the characteristics of early adopters of home 3D printing systems in China. This study
contributes to the early understanding of Chinese consumers’ adoption of innovative 3D printing systems.
Keywords: 3D printing systems; moderation effect; hierarchical regression; structural equation modelling

1. Introduction
Since the Industrial Revolution, conventional manufacturing techniques such as cutting and forming have been much
refined resulting in sophisticated numerical lathes and milling machines for cutting, and injection moulding and die-
casting machines for forming. The costly and extremely complicated machine tools make it very challenging for users
to operate the same. 3D printing technology provides an innovative alternative to the traditional manufacturing proce-
dure, where a 3D physical model of any shape can be generated directly from a computer-aided design model (Noorani
2005). It creates objects from the bottom-up by adding material to each cross-sectional layer. It allows individual users
to print complicated engineering parts automatically from design files, and this technology has simplified communication
between different actors involved in the development of a product (Lipson and Kurman 2013).
The growth of 3D printing systems in the personal consumption market is gaining speed in China. There are cur-
rently approximately 17,000 home 3D-printers in operation in China being used to make prototypes and quirky objects,
closely following the US, in leading the 3D printing revolution (Davidson and Trentmann 2013). Indeed, China is an
important market for 3D printing systems not only because it has one of the largest consumer bases but also because
China is still the manufacturing centre of the world (Nahm and Steinfeld 2014). As China is transforming from a
low-cost production economy to a middle-to-higher income economy, it will gradually lose its low-cost advantage in
manufacturing to neighbouring countries such as India and Vietnam. To sustain its economic growth, China needs to
transform itself from a low-skilled production economy to one with more value-added production emphasising its cre-
ative and technological industries. Home 3D printing systems enable a user-led open design approach to bring innova-
tive products and services to the public. Public acceptance serves to further incentivize the users to continuously
innovate for both private and social reasons. This virtuous cycle both reduces research and development wastage and
promotes social and economic benefits. Despite the growing popularity and importance of 3D printing technology in the
Chinese market, there is a gap in our understanding of how Chinese consumers understand this innovative technology.
There are very few studies which have been conducted from a Chinese consumers’ perspective to investigate their
perceptions and acceptance of 3D printing technology.
While there seems to be an implicit assumption that 3D printing systems provide a means to manufacture
customised products which have the precise features an individual user requires, there is also some debate over whether

*Corresponding author. Email: [email protected]

© 2016 Informa UK Limited, trading as Taylor & Francis Group

6018 Q. Wang et al.

a 3D printing system can be used successfully by individual consumers for small-scale production in a home setting.
Berman (2012) points out that there are huge potential market opportunities for home 3D printing system providers who
can meet the needs of the evolving consumers. Some authors have argued that price (Ratto and Ree 2012), performance
(Forrest 2014) and software associated with 3D printing systems (Campbell et al. 2011; Ariadi et al. 2012) are not
developing fast enough to make 3D printing systems appealing to consumers for use in a home setting. To develop
successful new products and systems, new product development managers should have a good understanding of their
consumers’ adoption process (Mugge and Dahl 2013). Although 3D printing systems enable users to do things they
were not able to do with traditional manufacturing systems, research has shown that consumers are initially reluctant to
adopt innovative products ‘that provide novel benefits but involve high learning costs’ (e.g. Zhao, Hoeffler, and Dahl
2009). Consumers may have difficulties in understanding the benefits of novel products, therefore, intentions to adopt
such innovative products may remain low (Feiereisen, Wong, and Broderick 2013).
Very few studies have investigated the adoption of 3D printing systems and even fewer studies have empirically
tested how much individuals’ characteristics, and their perceptions towards home 3D printing systems, can explain peo-
ple’s adoption of such systems in China. This study combines Innovation Diffusion Theory (IDT) with the Technology
Acceptance Model (TAM) to explore the factors affecting users’ intentions to use home 3D printing systems and to
investigate the moderation effects of individual characteristics (e.g. demographic variables) on their intention to use the
system. This study contributes to the design of home 3D printing systems for Chinese consumers, the development of
product management strategies in the Chinese market for 3D printing manufacturers and the opportunity of using
consumer-led design approaches to unleash the potential social and economic transformational impact that 3D printing
systems may have on creativity and social innovation.

2. Literature review
IDT (Rogers 1995) and the TAM (Davis 1989) are two leading theories with which to study users’ adoption of technol-
ogy. IDT has been used extensively for information technology and system research (Kwon and Suh 2004). Rogers
(1995) defined diffusion as ‘the process by which an innovation is communicated through certain channels over time
among the members of a social system’, and defined innovation as ‘ideas, customs or objects that are perceived as new
by individuals who use them’. IDT focuses on product characteristics to explain the adoption of an innovation. It studies
the factors of relative advantage, compatibility, complexity, divisibility and observability to determine their effects on
the adoption of innovative products.
The TAM was introduced by Davis (1989) to provide an explanation of the determinants of technology acceptance.
The model suggests that perceived ease of use (PEOU) and perceived usefulness (PU) are the primary determinants to
explain the variance in users’ intentions to use new technology. The model hypothesises users’ intention to use is also
influenced by users’ attitudes towards using, while users’ attitude is directly affected by the PEOU and PU of new
technology.
Both TAM and IDT have been applied broadly in studies of computing, the Internet, e-commerce, online games,
Internet shopping and mobile commerce (Yeh and Li 2009; Chong 2013). Studies using IDT have focused on product
characteristics, such as relative advantage, compatibility, complexity, divisibility and observability, to determine their
effects on the adoption of innovative products (Rogers 1995; Kwon and Suh 2004). Other studies in the TAM have
investigated technology acceptance from a consumer-centric perspective by incorporating PU and PEOU (Davis 1989).
There have also been some attempts made to integrate IDT into the TAM to investigate users’ technology acceptance
(Chang and Tung 2008). Chen et al. (2002) integrated the TAM and IDT to explain consumer behaviour in virtual store
settings. Kwon and Suh (2004) used a combined model of TAM and IDT to examine the adoption of supply chain man-
agement system by enterprises. These studies suggest there is a complementary relationship between the IDT and TAM
because these two models share conceptual bases that make them ideal for complementary uses (Chen et al. 2002). The
complexity construct in IDT can be replaced by the PEOU in the TAM, while the relative advantage from the IDT is
often considered to play the same role as PU in the TAM (Liebermann and Stashevsky 2002).
By applying the TAM and IDT, this research takes an extended new perspective from which to examine Chinese
consumer adoption of home 3D printing systems. The study also investigates the moderating effects of demographic
variables on Chinese consumers’ system adoption to identify the characteristics of earlier adopters. A number of studies
(Teo 2001; Cutler, Hendricks, and Guyer 2003; Wei et al. 2009) have found that technology users tend to possess cer-
tain demographic characteristics that can moderate the adoption of technology. For example, Chong (2013) identified
the moderating effects of demographic profiles in the PU and PEOU of m-commerce.
 International Journal of Production Research 6019

3. Research model and hypotheses

Primarily based on two well-established TAM and DIT theories, our research provides both theoretical and empirical
analysis to explain the direct factors and moderating effects that determine consumers’ adoption of 3D printing systems.

3.1 Constructs from the TAM

3.1.1 Perceived ease of use
PEOU is defined as the degree to which people consider the use of a new system requires little effort (Davis 1989;
Jeyaraj, Rottman, and Lacity 2006). Gould and Lewis (1985) recommended all designers should strive for usability and
suggested that any systems designed for people should be easy to use, easy to learn and helpful to users. From the
users’ point of view, it is reasonable to expect that systems which are user friendly and easy to use are more likely to
be adopted and used. Thus, we propose the following hypothesis:

H1: PEOU is positively related to users’ intention to use home 3D printing systems.

3.1.2 Perceived Usefulness

PU is defined as the extent to which one person considers whether or not a new technology could increase their produc-
tivity and work-related performance (Davis 1989). Jeyaraj, Rottman, and Lacity (2006) suggested that PU has a
far-reaching effect on adoption of new technology. Previous research has shown that PU will have a significantly posi-
tive impact on behavioural intention (Venkatesh and Morris 2000; Jeyaraj, Rottman, and Lacity 2006). This leads to the
following hypothesis:

H2: PU is positively related to users’ intention to use home 3D printing systems.

3.2 Constructs from IDT

Tomatzky and Klein (1982) found the product characteristics of relative advantage, compatibility and complexity have
strong influences on individuals’ adoption of new products and technologies, whereas divisibility and communicability
showed inconsistent findings on their effects on the adoptions of new technologies. For this reason, divisibility and
observability were excluded in this study. We replaced the construct of complexity by PEOU and the relative advantage
construct by PU, based on previous studies (e.g. Moore and Benbasat 1991).
Perceived Compatibility (PC) refers to the extent to which an innovation is perceived to be consistent with the
adopters’ beliefs, lifestyle, existing values, experience and current needs (Moore and Benbasat 1991). PC is found to be
significantly related to users’ intention to use e-learning (Liao and Lu 2008), a specific groupware application (Lotus
Domino) discussion database (Van Slyke, Lou, and Day 2002), web-based shopping (Van Slyke, Belanger, and
Comunale 2004), innovation adoption (Tornatzky and Klein 1982) and Internet Banking Services (Tat et al. 2008). We
propose the following hypothesis:

H3: PC is positively related to users’ intention to use home 3D printing systems.

3.3 Extended construct – perceived enjoyment

3D printing systems inherently possess both practical and hedonic characteristics that are critical for innovation
adoption. The importance of perceived enjoyment (PE) has long been established in studies of workplace computing
(Webster and Martocchio 1992) and computer games (Holbrook et al. 1984). 3D printing systems have quickly become
an affordable option and they can now be purchased at a very reasonable price for end users. The potential advantage
of using a 3D printing system the capability to produce physical objects at a much lower cost and more rapidly than tra-
ditional manufacturing methods. This advantage may be likely to bring enjoyment to users, for example in terms of real-
ising their creativity in design. Hobbyists and professional users are now able to design, download and print out a broad
range of three-dimensional physical objects. College undergraduates and high school students, and even younger chil-
dren are beginning to use the technology in the realm of education (Eisenberg 2013). PE has been found to have posi-
tive and significant effects on people’s intention to use technologies, including word processing systems, mobile games,
website usage, Internet-related activities such as messaging, television commerce, e-commerce and m-commerce and
blog usage (Davis, Bagozzi, and Warshaw 1992; Teo 2001; Hsu and Lin 2008; Chong 2013). In this study, PE is
6020 Q. Wang et al.

defined as the extent to which the process of using a 3D printing system to realise design ideas into real products is
perceived to be fun or enjoyable. Hence, we posit:

H4: PE is positively related to users’ intention to use home 3D printing systems.

3.4 Moderators
3.4.1 Age
Age has been identified to have a moderating effect between technology use and perceptions (Yi, Wu, and Tung 2005).
As age increases it becomes more difficult to process complex information and to allocate attention to information (Li,
Lindenberger, and Sikström 2001). In studies of the adoption of the Internet, researchers have claimed that consumers
tend to be young people because they perceive lower risks when compared to older people (Libermann and Stashevsky
2002). Age has been found to have a negative relationship with skill level (Elder, Gardner, and Ruth 1987) and with
computer usage (Zeffane and Cheek 1993). Older workers were reported to be more likely to experience technology
stress than younger workers (Harrison and Rainer 1992). Age has also been found to have moderation effects on effort
expectancy related constructs (e.g. ease of use) on users’ intentions to make use of mobile applications in their learning
(Wang, Wu, and Wang 2009). New technology that is perceived to be easy to use will become more critical in
technology adoption for aged people (Kubeck et al. 1996). These points lead to the following hypothesis:

H5a: The relationship between PEOU and intention to use home 3D printing systems is moderated by age, such that the
relationship is stronger for older people.

Previous research suggests that younger workers are more concerned about job-related achievement, task accomplish-
ment and external rewards than older workers (Kuo and Chen 2004). PU is also closely related to outcomes such as task
accomplishment and extrinsic reward (Venkatesh and Morris 2000). Hence, we also expect the moderation effect of age
on the interaction between PU and intention to use will be more salient to younger people than to older people. It
follows that:

H5b: The relationship between PU and intention to use home 3D printing systems is moderated by age, such that the relation-
ship is stronger for younger people.

Age has been found to have significant negative effects on PE in Internet usage when PEOU is not included in the
regression, however, ‘in the presence of PEOU, the result is no longer significant’ (Teo and Lim 1999). This may sug-
gest that PEOU dominates PE in the case of Internet usage for older users. We do not intend to study PE in the isolation
of PEOU, hence, based on the previous research, we posit:

H5c: The relationship between PE and intention to use home 3D printing systems is not moderated by age.

According to Xue et al. (2012) if the elderly regard technology as being compatible with their current usage habits, then
they would perceive that using the technology would be beneficial and would have higher intentions to use the same.
Since older people are more reluctant to adopt new technologies (Wang and Sun 2016), we hypothesise that:

H5d: The relationship between PC and intention to use home 3D printing systems is moderated by age, such that the relation-
ship is stronger for older people.

3.4.2 Gender
It has been suggested that females are more fearful of using computers than males (Igbaria, Livari, and Maragahh
1995). In the studies on gender differences on Internet usage (Jackson et al. 2001), observed that females retain higher
levels of computer anxiety and lower levels of self-efficacy than males. PEOU is suggested to be more salient for indi-
viduals espousing feminine rather than masculine values (Srite and Karahanna 2006). Furthermore, PEOU has been
found to be more salient for females than for males on the adoption of information technology (e.g. Venkatesh and
Morris 2000). This leads to the following hypothesis:

H6a: The relationship between PEOU and intention to use home 3D printing systems is moderated by gender, such that the
relationship is stronger for females.
 International Journal of Production Research 6021

Venkatesh el al. (2003) confirmed the moderating role of gender in the influence of performance expectancy (usefulness)
on the adoption of new technology. As PU is closely related to one’s performance, its values are highly regarded as
masculine values (Venkatesh and Morris 2000; Venkatesh et al. 2004). Research on gender difference on using
e-learning (Ong and Lai 2006) and on using new technology (Venkatesh and Morris 2000) found that levels of accep-
tance towards a new technology were more influenced for men by their assessment of usefulness than they were for
women. Thus:

H6b: The relationship between PU and intention to use home 3D printing systems is moderated by gender, such that the
relationship is stronger for males.

PE has significantly influenced the intention to adopt the Internet for learning in higher education, for both males and
females (Macharia and Nyakwende 2011). However, this is inconsistent with the findings of Hwang (2010), who
reported that PE exerts a stronger effect on intention to use e-commerce for males than for females. This study will test
the following hypothesis:

H6c: The relationship between PE and intention to use home 3D printing systems is moderated by gender, such that the
relationship is stronger for males.

Egan and Perry’s (2001) work on the multidimensional nature of gender identity provides some initial insights into the
possible relationship between genders and compatibility. They link PC with self-perceived competency with activities
that were either stereotypical of males or females, however, it is not clear whether the moderation effect of gender on
PC is stronger for males or females. Therefore, we posit:

H6d: The relationship between PC and intention to use home 3D printing systems is moderated by gender.

3.4.3 Educational level

Adoption of new technology is inter-related to the extent of knowledge one has acquired in order to use such technol-
ogy (Rogers 1995). A higher educational level may give rise to a greater level of knowledge in new technologies,
thereby accelerating the early adoption of a new technology. People educated to a higher level were found to be more
likely to use the Internet (Rhee and Kim 2004) and to be early adopters of spreadsheet software (Brancheau and
Wetherbe 1990). Individuals who are less well educated testify that insufficient knowledge about the Internet is one of
the main reasons for not using the same (NTIA 2002). Educational level has also been found to have a significantly
positive relationship with PEOU in the study of acceptance of new information technology (Agarwal and Prasad 1999)
and a moderating effect on the relationship between PEOU and intention to use the Internet (Kripanont 2006).
Therefore, we posit:

H7a: The relationship between PEOU and intention to use home 3D printing systems is moderated by educational level, such
that the relationship is stronger for people with a higher level of education.

In a study of online music system acceptance, Suki (2011) found that the acceptance level for people with a higher edu-
cational level was more influenced by PU and PEOU than was the case for people with a lower level of education.
Therefore, we propose the following hypothesis:

H7b: The relationship between PU and intention to use home 3D printing systems is moderated by educational level, such that
the relationship is stronger for people with a higher level of education.

PE has not been reported to have a direct relationship with education level. However it is evident that education level is
positively related to usage of computers (Igbaria, Livari, and Maragahh 1995) and PE is one of the main reasons why
consumers use the Internet in terms of activities such as messaging, browsing, downloading and purchasing (Teo 2001).
Therefore, we hypothesise:

H7c: Educational level has a moderation effect on the relationship between PE and intention to use home 3D printing systems,
such that the relationship is stronger for users with a higher level of education.
6022 Q. Wang et al.

Study on the relationship between educational level and technology acceptance is limited (Chong 2013). Rhee and Kim
(2004) found that those with a higher level of education will, in fact, be more likely to use the Internet. Liebermann
and Stashevsky (2002) found that users with a lower educational level will perceive Internet use to be more risky than
users with a higher educational level. Najmul Islam (2016) suggests that PC will have an impact on the relationship
between learning and technology usage. Users with a higher level of education find e-learning systems fit their learning
styles better than those users with a lower education level and are more likely to adopt such systems. Therefore, it is
hypothesised that education level moderates the relationship between PC and intention to use home 3D printing systems,
such that the effect is stronger for higher educated users. Therefore, we hypothesise:

H7d: Educational level has a moderation effect on the relationship between PC and intention to use home 3D printing systems,
such that the relationship is stronger for higher educated users.

3.4.4 Design background

Ralph and Wand (2009) defined a designer as a person who ‘specifies the structural properties of a design object.’ In
this study, we define designers as those who are currently actively involved in design-related activities and those consid-
ering design as either their profession or as a ‘serious’ hobby. Hoskins (2013) describes how the creative industries are
now directly interfacing with 3D printing technology and how this influences the practices of designers. By using 3D
printing technology, designers can create physical objects directly from a digital model, which provides flexibility in
geometric shapes and presents an opportunity for designs to be realised in a way that was not previously possible with
traditional manufacturing technologies (Hilton and Jacobs 2000). The usefulness of a 3D printing system to designers is
becoming prominent as such systems have helped many designers in several fields, including fashion design, architec-
ture design, the automobile industry and medical equipment. A plausible reason for designers to use a home 3D printing
system is both its usefulness in helping their design work, and its PEOU compared with other systems (Ariadi et al.
2012). Compatibility is found to be significantly related to users’ intention to use innovative products (Tornatzky and
Klein 1982), and while most of our participants with a design background had some experience with 3D modelling
applications, we hypothesise that designers are more likely to possess compatible experience of working with systems
similar to a 3D printing system than would be the case for non-designers. Therefore, we postulate that designers are
more likely to use 3D printing systems than non-designers, especially those who perceive such systems to be useful,
easy and fun to use and compatible with their working experience and needs from their work. This leads to the
following hypotheses:

H8a: The relationship between PEOU and intention to use home 3D printing systems is moderated by a designer variable, such
that the relationship is stronger for designers.

H8b: The relationship between PU and intention to use home 3D printing systems is moderated by a designer variable, such
that the relationship is stronger for designers.

H8c: The relationship between PE and intention to use home 3D printing systems is moderated by a designer variable, such
that the relationship is stronger for designers.

H8d: The relationship between PC and intention to use home 3D printing systems is moderated by a designer variable, such
that the relationship is stronger for designers.

Drawing upon existing literature and prior empirical findings, the TAM and IDT were integrated to construct the
research model for our research (see Figure 1).

4. Data collection and samples

Data for this research was collected in two stages from non-designers and designers from various disciplines including
fashion designers, interior designers, advertising designers, graphic designers and design students. A pilot test was
conducted with 30 participants with a mixture of university students and designers to validate the survey instruments.
Figure 2 has been used to illustrate how the home 3D printing system works.
 International Journal of Production Research 6023

Figure 1. Research model based on the TAM and IDT.

Figure 2. The white part on the yellow tray is a component of the vase. It was printed from the model depicted on the computer
screen.

A Scenario – Using Home 3D printing to print a Vase (Translated from Chinese version)
Step 1: Produce a 3D model of the vase using computer-aided design (CAD) software;
Step 2: Convert the CAD model of the vase to the STL format (a file format developed for 3D printing Systems) by software;
Step 3: Copy the STL file to the computer that controls the 3D printer;
Step 4: Setup the 3D printer, including refilling the plastic material;
Step 5: The printer uses fused-filament fabrication. Molten plastic is extruded from a fine nozzle and laid down on a flat plate. The
plate then drops a small distance, and the next layer is added, with each layer adding to build a complete vase;
Step 6: Remove the printed vase from the printer;
Step 7: Post-process the printed vase to remove the supports.

A total of 281 participants took part in the survey (134 males and 122 females). Out of 281 returned surveys, 25
were not usable because of missing data, leaving 256 usable surveys for analysis. Tables 1 and 2 present the demo-
graphic profile of the participants of the studies.
Variable measurement. The questionnaire was designed with items validated from previous technology adoption stud-
ies adapted from previous research (Igbaria, Livari, and Maragahh 1995; Venkatesh and Davis 2000; Wei et al. 2009).
6024 Q. Wang et al.

Table 1. Demographic profiles.

Demographic profile Number Percentage

Gender
Male 134 52.34
Female 122 47.66
Age
20 years and less 39 15.23
21 to 29 years 87 33.98
30 to 39 years 47 18.36
40 to 49 years 41 16.02
50 years and above 42 16.41
Highest educational level
High school 54 21.1
Diploma/Polytechnic 67 26.2
Bachelor degree 73 28.5
Postgraduate degree or over 62 24.2
Design background
Designer 112 43.75
Non-designer 144 56.25

Table 2. Factor analysis.

No. of items Factor loadings Cronbach’s Alpha Eigen value AVE

PEOU 4 0.656–0.845 0.807 2.368 0.6367

PU 3 0.779–0.817 0.768 2.059 0.7006
PE 3 0.779–0.824 0.873 1.343 0.8256
PC 3 0.632–0.758 0.704 1.032 0.6165
ItU 3 0.697–0.798 0.786 1.078 0.7436

The construct - PC taken from IDT to be used in our study measured (i) users’ assessments on how the system fits their
work and lifestyle; (ii) users’ self-rated level of how the system fits their working experience and design interests; and
(iii) their assessment on the needs of using the 3D printing system for their work. The scales used to measure the TAM
constructs, the IDT constructs and the additional construct of PE were set on a 5 point Likert Scale ranging from 1
(strongly disagree) to 5 (strongly agree). Gender was coded as a dummy variable, which is in line with prior studies
(Venkatesh et al. 2003), while age was considered as a continuous variable (Morris and Venkatesh 2000) and educational
level was recorded as the highest educational qualification attained, coded from 1 to 4 (i.e. 1 for high school and 4 for
postgraduate degree or higher). Design background was coded as 1 for a non-designer and 2 for a designer. Participants
in this study who were regarded as designers included those who claimed to work as a designer or who have actively
(measured by frequency and within one standard deviation of lowest frequencies of professional designers) engaged in
design related activities, such as DIY, personalisation of mobile or digital applications, painting, or drawing.

5. Results and discussion

5.1 Reliability tests
The reliability of this questionnaire was tested by Cronbach’s alpha. Nunnally (1978) stipulated that if the constructs are
generally above or close to 0.70, then it can be confirmed that the item measurements of the constructs are reliable. In
Table 2, Cronbach’s alpha for the four constructs ranges from 0.794 to 0.909, thus confirming the reliability of the item
measurements of the constructs.
Similar to methods used in e.g. Teo (2001), factor analysis with varimax rotation was performed to ascertain whether
PEOU, PU, PC, PE and Intention to Use (ItU) are distinct constructs. Following Igbaria, Livari, and Maragahh (1995),
an eigenvalue of 1.0 and a factor loading of 0.50 were used as the cut-off points. For these constructs, the correspond-
ing eigenvalues are greater than 1 and their associated factor loadings are larger than 0.50, therefore, they are confirmed
to be distinct from each other (see Table 2).
 International Journal of Production Research 6025

Table 3. Correlation and squared root of the AVE.

PEOU PU PE Comp ITU

PEOU 0.798
PU 0.617 0.837
PE 0.507 0.687 0.908
PC 0.468 0.559 0.548 0.785
ITU 0.576 0.767 0.629 0.736 0.862

5.2 Validation tests

The average variance extracted (AVE) of each construct was used to confirm the measurement model. Fornell and
Larcker (1981) proposed that convergent validity is achieved if the AVE for each construct is greater than 0.50. Table 2
shows that the AVE for each of the constructs is well above 0.50 and, therefore, convergent validity is evident. To con-
firm discriminant validity, the square root of the AVE for each construct should be larger than the correlations of the
inter-constructs in the measurement model (Hair Jr. et al. 2009). The results are shown in Table 3 and it can be seen that
the discriminant validity of each construct is clearly supported.

5.3 Hierarchical regression analysis

Table 4 presents the results from a regression analysis. The combined model using constructs from both the TAM and
IDT could explain 57% of the variation (adjusted R2) in intention to use, which significantly improved what the TAM
has achieved. All motivational variables are confirmed to have a significant positive relationship with intention to use at
different significant levels. This finding is consistent with previous findings from technological adoption studies in com-
puters, the Internet, E-commerce, online games, Internet shopping and mobile commerce (Igbaria, Parasuraman, and
Baroudi 1996; Chang and Cheung 2001; Hsu, Lu, and Hsu 2007; Basyir 2009; Wei et al. 2009; Chong 2013).

5.4 Structural equation model

All hypotheses related to the moderating effects were tested using covariance-based structural equation modelling
(SEM). SEM has the strength to examine and estimate causal relationships even among latent variables that cannot be
measured directly. To test the moderation effects, we followed the procedures outlined in Hair Jr. et al. (2009) by testing
the imposed equality constraints on the path coefficients in the competing models to confirm the moderation effects of
age, gender, education level and design background on the relationship between motivational variables and intention to
use. To do so, we divided our sample into subgroups accordingly (e.g. when considering the moderation effect of
gender, we divided the sample into a male group and a female group).
Age has been found to have a highly significant positive moderating effect between PEOU and intention to use (see
Table 5). The moderation effect is stronger for older users (see Table 6). Therefore, H5a is supported. In addition, the
degree to which a 3D printing system is perceived to be easy to use is highly influential on aged people’s decisions to
reject or adopt the technology (Kubeck et al. 1996). However, age was not confirmed to have a significant moderating

Table 4. Regression analysis.

Dependent variable: Intention to Use

TAM Only TAM & IDT

Perceived ease of use 0.128** 166**
Perceived usefulness 0.383* 0.465*
Perceived enjoyment 0.174* 0.216*
Perceived compatibility 0.197*
Adjusted R2 0.456 0.576

*p<0.01; **p<0.05.
6026 Q. Wang et al.

Table 5. Moderation effect.

P values
Age Gender Education Level Design background

PEOU→ITU 0.032 0.072 0.144 ***

PU→ITU 0.148 0.167 0.715 0.002
PE→ITU 0.345 0.050 0.425 0.149
PC→ITU 0.276 0.216 0.383 0.016

Note: All the numbers presented in the table are p values.

***p<0.001.

Table 6. Moderation effect.

Moderation effect - Age Coefficient estimate

Age group 2 Age group 3 Age group 4

H5a: PEOU→ITU 0.058 (0.601) 0.211 (0.227) 0.554(*)

Moderation effect - Gender Coefficient estimate
Female Male

H6a: PEOU→ITU 0.797(*) 0.304(0.071)

H6c: PE→ITU 0.047(0.682) 0.317(0.005)
Moderation effect - Designer Coefficient estimate
Non-designer Designer

H8a: PEOU→ITU 0.040(0.498) 0.177(0.028)

H8b: PU→ITU 0.336(0.346) 0.532(***)
H8d: PC→ITU 0.158(0.268) 0.316(0.032)

Note: Numbers in brackets are the associated p values.

***p <0.001; age group 2 is between 21 and 30; age group 3 is between 31and 40; age group 4 is between 41 and 50; the bold coef-
ficient estimates are significant at certain significance levels (as shown in brackets).

effect between PU and intention to use, between PC and intention to use and between PE and intention to use. There-
fore, H5b, H5c and H5d are all rejected. These results may indicate that the degree of the effects that PU, PE and PC
have on intention to use is less sensitive to age than PEOU. Previous researchers have also stressed the highly signifi-
cant importance of ease of use for older individuals compared to younger users in terms of the acceptance of new tech-
nology (Gurtner et al. 2014), which is probably due to older users having less experience with technology as a result of
declining absorptive capacity (Sorce 1995). The moderation effect of gender is evident between PEOU and intention to
use (see Table 5). The coefficients for both females and males on the path between PEOU and intention to use are
found to be significant (see Table 6), however, for females, the effect of PEOU on their intention to use is highly signif-
icant at a 0.1% significant level, while for males, the effect is moderately significant at a 7.1% significant level (see
Table 6). This difference is further reinforced by the magnitude of the estimated coefficients reported in Table 6. Gender
has also been found to have a significant positive moderation effect between PE and intention to use (see Table 5). For
males, their intention to use new technology is found to be more influenced by their perception of enjoyment gained
from using such new technology (see Table 6) when compared to females. Hence, H6a and H6c are both supported.
Gender is not found to have significant moderation effects on the relationships between PU and intention to use. This is
contrary to the findings of Ong and Lai (2006) which found male users’ intentions to use were more influenced by use-
fulness than was the case for females. Gender is also not found to have significant moderation effects on the relation-
ships between PC and intention to use. This is consistent with Hur, Kim, and Kim (2013) who found that gender does
not have significant moderating effects on the relationship between PC and intention to use a tablet computer. Our find-
ings show that PU and PC are considered equally important for both males and females on predicting their adoption of
3D printing systems. Hence, both H6b and H6d are rejected.
Educational level is not found to be an important moderator of the key relationships with intention to use. Hence,
H7a, H7b, H7c and H7d are all rejected. These findings contradict previous studies which support the hypothesis that
highly educated people tend to adopt new technology more quickly than the less educated (e.g. Weinberg 2005).
 International Journal of Production Research 6027

Table 7. Summary of hypotheses results.

Direct Effects (Dependent variable: ITU)

Brief Explanation Decision

H1 PEOU Perceived Ease of Use is found to have a significant positive effect on ITU. A system Accept**
which is easy to use enhances users’ intentions to use such a system. This confirmed
the earlier findings on the importance of systems’ usability in influencing users’
intentions to use the same (e.g. Pavlou and Fygenson 2006).
H2 PU Perceived Usefulness is found to have a significant positive effect on ITU. A system Accept***
which is perceived to be useful enhances users’ intentions to use such a system.
H3 PE Perceived Enjoyment is found to have a positive effect on ITU. When users perceived a Accept***
3D printing system as enjoyable to use, this enhanced their intentions to use such a
system.
H4 PC Perceived Compatibility is found to have a positive effect on ITU. When users Accept***
perceived a 3D printing system as compatible with their previous experiences or needs,
this enhanced their intentions to use the same.
Moderating effects
Brief Explanation Decision

Age H5a PEOU→ITU An improved perception of ease of use can significantly increase older people’s Accept***
intentions to use a 3D printing system.
H5b PU→ITU There is no significant difference in the influence of PU on ITU across all age groups. Reject
Perceived usefulness enjoys equal importance across different age groups in the
adoption of a 3D printing system.
H5c PE→ITU No significant difference was identified in the influence of PE on ITU across all age Reject
groups in this study.
H5d PC→ITU There is no significant difference in the influence of PC on ITU across all age groups. Reject
Gender H6b PU→ITU There is no significant difference in the influence of PU on ITU across the genders. Reject
H6c PE→ITU PE exerts far more influence on males than females on their intention to use a 3D Accept***
printing system.
H6d PC→ITU There is no significant difference in the influence of PC on ITU across the genders. Reject
H7a PEOU→ITU We found there is no significant difference in the influence of PEOU, PU, PE and PC Reject
H7b PU→ITU on ITU across the educational levels.
H7c PE→ITU
H7d PC→ITU
Designer H8a PEOU→ITU Designers value perceived ease of use, usefulness and compatibility more than non- Accept**
H8b PU→ITU designers, which is probably because designers are more likely to use 3D printing Accept***
H8d PC→ITU systems if they perceived them to be useful in their design work, easy to use and
compatible with their professional experience.
H8c PE→ITU There is no significant difference in the influence of PE on ITU between designers and Reject
non-designers.

***p<0.001; **p<0.01.

However, the findings from these studies do not necessarily suggest the true causal effect of education on technology
adoption, as highlighted by Riddell and Song (2012). Our study could suggest the potential effects that education has
on technology adoption may depend on the kind of technology to be adopted.
Design background is a highly important moderator for three out of four key relationships in our model (see Tables
5 and 6). Hence, H8a, H8b and H8d are supported. PEOU, PU and PC are considered far more important for designers
than non-designers on their intention to use home 3D printing systems. The significant difference is probably due to the
differences between the skills possessed by designers and non-designers, as pointed out by Dunne, Haltiwanger, and
Troske (1997) who found that the adoption of technologies associated with design tasks is strongly related to the skills
of the workforce. Our results do not support that there is a significant difference between designers and non-designers
on the impact of PE on their intention to use a 3D printing system. However, our results (see Table 6) indicate that
designers consider the usability of the 3D printing system (i.e. how easily the 3D printing system can be used), the
functionality of the 3D printing system (i.e. how useful it can be for their work) and compatibility with their working
experience are far more important than how enjoyable it can be to use the system. Hence, H8c is rejected. We have
included summaries of the hypothetical results in Table 7.
6028 Q. Wang et al.

6. Conclusion and implications

Studies of the use of 3D printing technology among Chinese users remain sparse. This research investigates Chinese
users’ intention to use home 3D printing system. Although the TAM and IDT are two well-established theories for
explaining the diverse aspects of the adoption of technology and innovation, each individual theory has limitations in
terms of explaining adoption decisions that reflect both product and consumer dimensions. Hence, this study has
attempted to combine the established theories of the TAM and IDT to gain an understanding of technology acceptance.
We have also included an analysis of the moderating effects of user-specific variables (i.e. demographic variables) in
our model to better understand people’s acceptance and intention to use new technologies (Chen and Chan 2011).
Our integrated model based on the TAM and IDI was found to account for approximately 57% of the variance
(adjusted R2) in intention, which is a considerable improvement over the original TAM model (See Table 4). We proved
our proposed direct constructs (i.e. PEOU, PU, PC and PE) to be the main driving force behind users’ intentions to
accept 3D printing systems. These results may shed some theoretical insight into conceptualising the dual approach
(TAM and DIT) to studies of the adoption of innovative technology.
Our results also suggest that younger people are more likely to be early adopters of home 3D printing systems. Age
was found to have a moderation effect on the relationship between PEOU and intention to use, while the effects of
PEOU on intention to use become more profound as age increases. This may indicate that for older users, their intention
to use home 3D printing systems could possibly be enhanced if their perceptions of ease of use of such systems can be
improved. Furthermore, age is not found to have significant moderation effects on the other key relationships with
intention to use.
Gender was found to have significant moderation effects on the relationship between PEOU and intention to use,
and between PE and intention to use. This may suggest: (1) the perception of ease of use is more salient for females
than males. An improved perception of ease of use of 3D printing systems could help to boost female’s adoption of the
same as female users tend to have higher levels of fears of using new technology, as identified in earlier studies (e.g.
Igbaria and Chakrabarti 1990); (2) males are more influenced by their perception of enjoyment of using home 3D print-
ing systems when compared to their female counterparts, and PE exerts a much stronger influence on men than on
women with regard to their intention to use such systems.
Contrary to the findings of previous studies (e.g. Chong 2013), a moderation effect of educational level on the key
relationships with intention to use has not been established in this study. This challenges the perception that only highly
educated users will use home 3D printing systems. However, participants’ design background was found to have a sig-
nificant impact on their intentions to adopt home 3D printing systems, and participants with a design background are
more likely to adopt such systems than non-designers. PEOU, PC and PU are all found to exert more influence on par-
ticipants with a design background than those without a design background. Drawing from designers’ emphasis on a
system’s ease of use, compatibility with their previous working experience and its usefulness, a 3D printing system
should be supported with the most widely used 3D modelling software if 3D manufacturers’ main target users are those
with a design background. As non-designer users’ concerns over using a 3D printing system centre on their lack of
technical design skill, an easy to use design tool to accompany the system may foster higher levels of consumer adop-
tion of home 3D printing systems among non-designers. One suggestion is to use software that is already embedded
with licensed 3D models for various purposes as a starting point, with which users can continue building their own 3D
models. Another alternative is to build a crowd sourcing 3D model library for users to buy, download and modify. A
good example is the Thingiverse repository, initiated by Bre Pettis, for individuals to upload and share digital models
for 3D printing (Bradshaw, Bowyer, and Haufe 2010). With the input of consumers, an anticipated advantage is an
enhancement of motivation for design innovation.
The lower degree of adoption by non-design participants indicates that it is still not feasible for consumers with
non-design background to directly manufacture their products at this moment. A new collaborative approach of user par-
ticipatory design could be a potential solution, in which the artistic elements of the product design are done by
designers, while the remaining design decisions would be added by the actual users. This finding is consistent with
Campbell et al. (2011) who highlighted that the replacement of production professionals is not currently possible, and
collaborative consumer design is required.
This study of Chinese users’ adoption of home 3D printing systems may facilitate new conversations between manu-
facturers and consumers and thereby introduce new opportunities to design products in new business models for open
design platforms in China. Future research work could examine the effectiveness of 3D design tools for creating a mini-
revolution in consumer-led design in the ways products and services are designed and consumed in China. We will
examine the facilitation process in terms of the home 3D printing technology, design tools and resources that would
enable consumers to successfully conceive and produce their own designs.
 International Journal of Production Research 6029

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported from the National Natural Science Foundation of China [grant number 71401085]; the Interna-
tional Doctoral Innovation Centre (IDIC) at the University of Nottingham China. The opinions expressed are those of
the authors and do not represent views of the National Natural Science Foundation of China.

Acknowledgements
The authors also gratefully acknowledge the insightful comments from anonymous reviewers which have helped to improve
significantly on the earlier version of this paper.

References

Agarwal, R., and J. Prasad. 1999. “Are Individual Differences Germane to the Acceptance of New Information Technologies?”
Decision Sciences 30 (2): 361–391.
Ariadi, Y., R. I. Campbell, M. A. Evans, and I. J. Graham. 2012. “Combining Additive Manufacturing with Computer-aided
Consumer Design.” The 23rd International Solid Freeform Fabrication Symposium, Austin, TX.
Basyir, A. 2009. A Model of Consumers? Acceptance of Internet shopping: A Study of Technology Acceptance Model in Penang,
Malaysia. Saarbrücken: VDM Verlag Dr. Müller.
Berman, B. 2012. “3-D printing: The New Industrial Revolution.” Business Horizons 55 (2): 155–162.
Bradshaw, S., A. Bowyer, and P. Haufe. 2010. “The Intellectual Property Implications of Low-cost 3D Printing.” SCRIPTed 7 (1):
5–31.
Brancheau, J. C., and J. C. Wetherbe. 1990. “The Adoption of Spreadsheet Software: Testing Innovation Diffusion Theory in the
Context of End-user Computing.” Information Systems Research 1 (2): 115–143.
Campbell, T., C. Williams, O. Ivanova, and B. Garrett. 2011. “Could 3D Printing Change the World? Technologies, Potential, and
Implications of Additive Manufacturing.” Strategic Foresight Report.
Chang, M. K., and W. Cheung. 2001. “Determinants of the Intention to use Internet/WWW at Work: A Confirmatory Study.”
Information and Management 39 (1): 1–14.
Chang, S. C., and F. C. Tung. 2008. “An Empirical Investigation of Students’ Behavioural Intentions to Use the Online Learning
Course Websites.” British Journal of Educational Technology 39 (1): 71–83.
Chen, K., and A. H. S. Chan. 2011. “A Review of Technology Acceptance by Older Adults.” Gerontechnology 10 (1): 1–12.
Chen, L. D., M. L. Gillenson, and D. L. Sherrell. 2002. “Enticing Online Consumers: A Technology Acceptance Perspective.”
Information and Management 39 (8): 705–719.
Chong, A. Y. L. 2013. “Mobile Commerce Usage Activities: The Roles of Demographic and Motivation Variables.” Technological
Forecasting and Social Change 80 (7): 1350–1359.
Cutler, S. J., J. Hendricks, and A. Guyer. 2003. “Age Differences in Home Computer Availability and Use.” Journals of Gerontology:
Series B 58 (5): S271–S280.
Davidson, P., and A. Trentmann. 2013. “China Chasing US lead in 3-D Printing.” Accessed October 9, 2014. http://www.usato
day.com/story/money/business/2013/12/01/china-advances-3d-printing-manufacturing/3675057/.
Davis, F. D. 1989. “Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology.” MIS Quarterly
13 (3): 319–342.
Davis, F. D., R. P. Bagozzi, and R. P. Warshaw. 1992. “Extrinsic and Intrinsic Motivation to Use Computers in the Workplace.”
Journal of Applied Social Psychology 22 (4): 1111–1132.
Dunne, T., J. Haltiwanger, and K. Troske. 1997. “Technology and Jobs: Secular Change and Cyclical Dynamics.” Carnegie-Rochester
Public Policy Conference Series 46: 107–178.
Egan, S. K., and D. G. Perry. 2001. “Gender Identity: A Multidimensional Analysis with Implications for Psychosocial Adjustment.”
Developmental Psychology 37 (4): 451–463.
Eisenberg, M. 2013. “3D Printing for Children: What to Build Next?” International Journal of Child-Computer Interaction 1 (1):
7–13.
Elder, V. B., E. P. Gardner, and S. R. Ruth. 1987. “Gender and Age in Technostress: Effects on White Collar Productivity.”
Government Finance review 3 (6): 17–21.
Feiereisen, S., V. Wong, and A. J. Broderick. 2013. “Is a Picture Always Worth a Thousand Words? The Impact of Presentation
Formats in Consumers’ Early Evaluations of Really New Products.” Journal of Product Innovation Management 30 (S1):
159–173.
6030 Q. Wang et al.

Fornell, C., and D. F. Larcker. 1981. “Structural Equation Models with Unobservable Variables and Measurement Error: Algebra and
Statistics.” Journal of Marketing Research 18 (3): 382–388.
Forrest, C. 2014. “What’s Holding Back 3D Printing from Fulfilling Its Promise?” Accessed September 14. http://www.zdnet.com/
whats-holding-back-3d-printing-from-fulfilling-its-promise-7000032250/.
Gould, J. D., and C. Lewis. 1985. “Designing for Usability: Key Principles and What Designers Think.” Communications of the
ACM 28 (3): 300–311.
Gurtner, S., R. Reinhardt, and K. Soyez. 2014. “Designing Mobile Business Applications for Different Age Groups.” Technological
Forecasting & Social Change 88: 177–188.
Hair Jr., J. F., W. C. Black, B. J. Babin, R. E. Anderson, and R. L. Tatham. 2009. Multivariate Data Analysis. Delhi: Pearson
Education.
Harrison, A. W., and R. K. Rainer Jr. 1992. “The Influence of Individual Differences on Skill in End-User Computing.” Journal of
Management Information Systems 9 (1): 93–111.
Hilton, P. D., and P. F. Jacobs. 2000. Rapid Tooling: Technologies and Industrial Applications. New York: CRC Press.
Holbrook, M. B., R. W. Chestnut, T. A. Oliva, and E. A. Greenleaf. 1984. “Play as a Consumption Experience: The Roles of
Emotions, Performance, and Personality in the Enjoyment of Games.” Journal of Consumer Research 11 (2): 728–739.
Hoskins, S. 2013. 3D Printing for Artists, Designers and Makers. London: Bloomsbury.
Hsu, C. L., and J. C. C. Lin. 2008. “Acceptance of Bog Usage: The Roles of Technology Acceptance, Social Influence and
Knowledge Sharing Innovation.” Information & Management 45 (1): 65–74.
Hsu, C. L., H. P. Lu, and H. H. Hsu. 2007. “Adoption of the Mobile Internet: An Empirical Study of Multimedia Message Service.”
Omega 35 (6): 715–726.
Hur, W. M., H. Kim, and W. M. Kim. 2013. “The Moderating Roles of Gender and Age in Tablet Computer Adoption.”
Cyberpsychology, Behavior, and Social Networking 17 (1): 33–39.
Hwang, Y. 2010. “The Moderating Effects of Gender on e-commerce Systems Adoption Factors: An Empirical Investigation.”
Computers in Human Behavior 26 (6): 1753–1760.
Igbaria, M., and A. Chakrabarti. 1990. “Computer Anxiety and Attitudes Towards Microcomputer Use.” Behaviour and Information
Technology 9 (3): 229–241.
Igbaria, M., J. Livari, and H. Maragahh. 1995. “Why do Individuals Use Computer Technology? A Finish Case Study.” Information
& Management 29 (5): 227–238.
Igbaria, M., S. Parasuraman, and J. J. Baroudi. 1996. “A Motivational Model of Microcomputer Usage.” Journal of Management
Information Systems 13 (1): 127–143.
Jackson, L. A., K. S. Ervin, P. D. Gardner, and N. Schmitt. 2001. “Gender and the Internet: Women Communicating and Men
Searching.” Sex Roles 44 (5–6): 363–379.
Jeyaraj, A., J. W. Rottman, and M. C. Lacity. 2006. “A Review of the Predictors, Linkages, and Biases in IT Innovation Adoption
Research.” Journal of Information Technology 21 (1): 1–23.
Kripanont, N. 2006. “Using a Technology Acceptance Model to investigate Academic Acceptance of the Internet.” Journal of Busi-
ness Systems, Governance and Ethics 1 (2): 13–28.
Kubeck, J. E., N. D. Delp, T. K. Haslett, and M. A. McDaniel. 1996. “Does Job-related Training Performance Decline with Age?”
Psychological Aging 11 (1): 92–107.
Kuo, Y. F., and L. S. Chen. 2004. “Individual Demographic Differences and Job Satisfaction among Information Technology Person-
nel: An Empirical Study in Taiwan.” International Journal of Management 21 (2): 221–231.
Kwon, I. W. G., and T. Suh. 2004. “Factors Affecting the Level of Trust and Commitment in Supply Chain Relationships.” Journal
of Supply Chain Management 40 (1): 4–14.
Li, S.C., U. Lindenberger, and S. Sikström. 2001. “Aging Cognition: From Neuromodulation to Representation.” Trends in Cognitive
Sciences 5 (11): 479–486.
Liao, H.-L., and H.-P. Lu. 2008. “The Role of Experience and Innovation Characteristics in the Adoption and Continued use of
e-learning Websites.” Computers & Education 51 (4): 1405–1416.
Liebermann, Y., and S. Stashevsky. 2002. “Perceived Risks as Barriers to Internet and e-commerce Usage.” Technology Market
Research 5 (4): 291–300.
Lipson, H., and M. Kurman. 2013. Fabricated: The New World of 3D Printing. Indianapolis: Wiley.
Macharia, J., and E. Nyakwende. 2011. “Gender Differences in Internet Usage Intentions for Learning in Higher Education: An
Empirical Study.” Journal of Language, Technology and Entrepreneurship in Africa 3 (1): 244–254.
Moore, G. C., and I. Benbasat. 1991. “Development of an Instrument to Measure the Perception of Adopting an Information
Technology Innovation.” Information Systems Research 2 (3): 192–222.
Morris, M. G., and V. Venkatesh. 2000. “Age Differences in Technology Adoption Decisions: Implications for a Changing Work
Force.” Personnel Psychology 53 (2): 375–403.
Mugge, R., and D. W. Dahl. 2013. “Seeking the Ideal Level of Design Newness: Consumer Response to Radical and Incremental
Product Design.” Journal of Product Innovation Management 30 (S1): 34–47.
Nahm, J., and E. S. Steinfeld. 2014. “Scale-up Nation: China’s Specialization in Innovative Manufacturing.” World Development 54
(C): 288–300.
 International Journal of Production Research 6031

Najmul Islam, A. K. M. 2016. “E-learning System Use and its Outcomes: Moderating Role of Perceived Compatibility.” Telematics
and Informatics 33 (1): 48–55.
National Telecommunications and Information Administration (NTIA). 2002. A Nation Online: How Americans are Expanding their
use of the Internet. Washington, DC: US Department of Commerce.
Noorani, R. 2005. Rapid Prototyping: Principles and Applications. New Jersey: Wiley.
Nunnally, J. C. 1978. Psychometric Theory. 2nd ed. New York: McGraw-Hill.
Ong, C. S., and J. Y. Lai. 2006. “Gender Differences in Perceptions and Relationships among Dominants of e-learning Acceptance.”
Computers in Human Behavior 22 (5): 816–829.
Pavlou, P. A., and M. Fygenson. 2006. “Understanding and Predicting Electronic Commerce Adoption: An Extension of the Theory
of Planned Behaviour.” MIS Quarterly 30 (1): 115–143.
Ralph, P., and Y. Wand. 2009. “A Proposal for a Formal Definition of the Design Concept.” In Design Requirments Workshop, edited
by K. Lyytinen, J. Loucopoulos, and W. Robinson, 103–136. Berlin: Springer-Verlag.
Ratto, M., and R. Ree. 2012. “Materializing Information: 3D Printing and Social Change.” First Monday 17 (7), accessed Oct 21,
2014. available at: http://firstmonday.org/htbin/cgiwrap/bin/ojs/index.php/fm/article/viewArticle/3968.
Rhee, K. Y., and W. B. Kim. 2004. “The Adoption and Use of the Internet in South Korea.” Journal of Computer-Mediated
Communication 9 (4): 35–42.
Riddell, W. C., and X. Song. 2012. “The Role of Education in Technology Use and Adoption: Evidence from the Canadian
Workplace and Employee Survey.” IZA Discussion Papers 6377, Institute for the Study of Labor.
Rogers, E. M. 1995. Diffusion of Innovations. 4th ed. New York: The Free Press.
Sorce, P. 1995. “Cognitive Competence of Older Consumers.” Psychology and Marketing 12 (6): 467–480.
Srite, M., and E. Karahanna. 2006. “The Role of Espoused National Cultural Values in Technology Acceptance.” MIS Quarterly 30
(3): 679–704.
Suki, N. M. 2011. “Gender, Age, and Education: Do they Really Moderate Online Music Acceptance?” Communications of the
IBIMA 2011, http://dx.doi.org/10.5171/2011.959384.
Tat, H. H., K. M. Nor, E. T. Yang, K. J. Hney, L. Y. Ming, and T. L. Yong. 2008. “Predictors of Intention to Continue Using Internet
Banking Services: An Empirical Study of Current Users.” International Journal of Business and Information 3 (2): 233–244.
Teo, T. S. H. 2001. “Demographic and Motivation Variables Associated with Internet Usage Activities.” Internet Research 11 (2):
125–137.
Teo, T. S. H., and R. Y. C. Lim. 1999. “Intrinsic and Extrinsic Motivation in Internet Usage.” OMEGA. International Journal of Man-
agement Science 27 (1): 25–37.
Tornatzky, L. J., and K. J. Klein. 1982. “Innovation Characteristics and Innovation Adoption-implementation: A Meta-analysis of
Findings.” IEEE Transactions on Engineering Management 29 (1): 28–45.
Van Slyke, C., F. Belanger, and C. Comunale. 2004. “Factors Influencing the Adoption of Web-based Shopping: The Impact of
Trust.” Database for Advances in Information Systems 35 (2): 32–46.
Van Slyke, C., H. Lou, and J. Day. 2002. “The Impact of Perceived Innovation Characteristics on Intention to Use Groupware.”
Information Resources Management Journal 15 (1): 5–12.
Venkatesh, V., and F. D. Davis. 2000. “A Theoretical Extension of the Technology Acceptance Model: for Longitudinal Field Stud-
ies.” Management Science 46 (2): 186–204.
Venkatesh, V., and M. G. Morris. 2000. “Why don’t men ever Stop to Ask for Directions? Gender, Social Influence, and their Role in
Technology Acceptance and Usage Behaviour.” MIS Quarterly 24 (1): 115–139.
Venkatesh, V., M. G. Morris, G. B. Davis, and F. D. Davis. 2003. “User Acceptance of Information Technology: Toward a Unified
View.” MIS Quarterly 27 (3): 425–478.
Venkatesh, V., M. G. Morris, T. Sykes, and P. L. Ackerman. 2004. “Individual Reactions to New Technologies in the Workplace: The
Role of Gender as a Psychological Construct.” Journal of Applied Social Psychology 34 (3): 445–467.
Wang, Q. F., and X. Sun. 2016. “Investigating Gameplay Intention of the Elderly Using an Extended Technology Acceptance Model
(ETAM).” Technological Forecasting and Social Change (Accepted for publication).
Wang, Y. S., M. C. Wu, and H. Y. Wang. 2009. “Investigating the Determinants and Age and Gender Differences in the Acceptance
of Mobile Learning.” British Journal of Educational Technology 40 (1): 92–118.
Webster, E. J., and J. J. Martocchio. 1992. “Microcomputer Playfulness: Development of a Measure with Workplace Implications.”
MIS Quarterly 16 (2): 201–226.
Wei, T. T., G. Marthandan, A. Y. L. Chong, K. B. Ooi, and S. Arumugam. 2009. “What Drives Malasian m-commerce Adoption? An
Empirical Analysis.” Industrial Management & Data Systems 109 (3): 370–388.
Weinberg, B. 2005. Experience and technology adoption. Mimeo: Ohio State University.
Xue, L., C. Yen, L. Chang, H. C. Chan, B. C. Tai, S. B. Tan, H. B. L. Duh, and M. Choolani. 2012. “An Exploratory Study of Age-
ing Women’s Perception on Access to Health Informatics via a Mobile Phone-based Intervention.” International Journal of
Medical Informatics 81 (9): 637–648.
Yeh, Y. S., and Y. M. Li. 2009. “Building Trust in m-commerce: Contributions from Quality and Satisfaction.” Online Information
Review 33 (6): 1066–1086.
6032 Q. Wang et al.

Yi, Y. D., Z. Wu, and L. L. Tung. 2005. “How Individual Differences Influence Technology Usage Behaviour? Toward an Integrated
Framework.” Journal of Computer Information Systems 46 (2): 52–63.
Zeffane, R., and B. Cheek. 1993. “Profiles and Correlates of Computer Usage: A Study of the Australian Telecommunications
Industry.” International Journal of Computer Applications in Technology 6 (1–2): 122–134.
Zhao, M., S. Hoeffler, and D. W. Dahl. 2009. “The Role of Imagination-focused Visualization on New Product Evaluation.” Journal
of Marketing Research 46 (1): 46–55.
This article was downloaded by: [202.40.157.7] On: 20 August 2023, At: 01:19
Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA

Organization Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org

Wakes of Innovation in Project Networks: The Case of

Digital 3-D Representations in Architecture, Engineering,
and Construction
Richard J. Boland, Jr., Kalle Lyytinen, Youngjin Yoo,

To cite this article:

Richard J. Boland, Jr., Kalle Lyytinen, Youngjin Yoo, (2007) Wakes of Innovation in Project Networks: The Case of Digital
3-D Representations in Architecture, Engineering, and Construction . Organization Science 18(4):631-647. https://
doi.org/10.1287/orsc.1070.0304

Full terms and conditions of use: https://pubsonline.informs.org/Publications/Librarians-Portal/PubsOnLine-Terms-and-

Conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact [email protected].

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.

Copyright © 2007, INFORMS

Please scroll down for article—it is on subsequent pages

With 12,500 members from nearly 90 countries, INFORMS is the largest international association of operations research (O.R.)
and analytics professionals and students. INFORMS provides unique networking and learning opportunities for individual
professionals, and organizations of all types and sizes, to better understand and use O.R. and analytics tools and methods to
transform strategic visions and achieve better outcomes.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
 Organization Science informs ®

Vol. 18, No. 4, July–August 2007, pp. 631–647 doi 10.1287/orsc.1070.0304

issn 1047-7039
eissn 1526-5455
07
1804
0631 © 2007 INFORMS

Wakes of Innovation in Project Networks: The Case of

Digital 3-D Representations in Architecture,
Engineering, and Construction
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

Richard J. Boland, Jr., Kalle Lyytinen

Department of Information Systems, Case Western Reserve University, 10900 Euclid Avenue,
Cleveland, Ohio 44106 {[email protected], [email protected]}

Youngjin Yoo
Fox School of Business, Temple University, 1810 North 13th Street,
Philadelphia, Pennsylvania 19122, [email protected]

C hanges in the technologies of representation in a heterogeneous, distributed sociotechnical system, such as a large
construction project, can instigate a complex pattern of innovations in technologies, practices, structures, and strategies.
We studied the adoption of digital three-dimensional (3-D) representations in the building projects of the architect Frank
O. Gehry, and observed that multiple, heterogeneous firms in those projects produced diverse innovations, each of which
created a wake of innovation. Together, these multiple wakes of innovation produce a complex landscape of innovations with
unpredictable peaks and valleys. Gehry’s adoption of digital 3-D representations disturbed the ecology of interactions and
stimulated innovations in his project networks by: providing path-creating innovation trajectories in separate communities of
practice, creating trading zones where communities could create knowledge about diverse innovations, and offering a means
for intercalating innovations across heterogeneous communities. Our study suggests that changes in digital representations
that are central to the functioning of a distributed system can engender multiple innovations in technologies, work practices,
and knowledge across multiple communities, each of which is following its own distinctive tempo and trajectory.
Key words: innovation; diffusion of innovation; innovation theory; IT-induced innovation; wakes of innovation; digital
3-D representation; intercalated innovations; trading zones; Frank Gehry; path creation; path dependency; distributed
systems; architecture; engineering and construction

Introduction Frank Gehry’s accomplishments as an architect are

We studied the adoption of digital 3-D representations well known, and he has received many distinguished,
in the design and construction projects of the architect, international awards in architecture, including the pro-
Frank O. Gehry, and observed that his adoption of digi- fession’s highest accolade, the Pritzker Prize. However,
tal 3-D models as a primary representation gave rise to his ability to spawn wakes of technology and process
multiple and diverse wakes of innovation throughout his innovations in the tradition-bound architecture, engi-
project networks. We use the image of a wake to depict neering, and construction (AEC) industries is another
an innovation as emerging in and traveling across an important part of his legacy. Consider, for example, the
innovation space, much as a wake travels through water. following sample of the innovations during the design
Research on technology innovation and diffusion has put and construction of the Peter B. Lewis Building in
considerable effort into studying the wake of a single Cleveland, Ohio (Figure 1).
technology innovation in a typically homogeneous envi-
In that one project:
ronment (Damanpour 1991, Swanson 1994, Zmud 1984,
(a) the structural engineering firm invented a new
Rogers 1995, von Hippel and von Krogh 2003, Bijker
1995, Latour 1997, Williams and Edge 1996). However, method for designing a steel roof with dramatically
in our study, firms involved in Frank Gehry’s project curved surfaces, which won an industry award for engi-
networks produced multiple and distinct innovations— neering innovation;
each of which became a different wake of innovation. (b) the drywall contractor invented multiple patent-
These multiple wakes overlapped on, intruded on, and able ways to frame undulating wall surfaces, and began
interacted with each other in ways that formed a tur- a new line of business, consulting on high-profile con-
bulent, self-propagating system of innovations—a com- struction projects;
plex, undulating surface of diverse innovations that fed (c) the Cleveland fire marshal developed new tech-
back into the project, stimulating further innovations in niques for modeling smoke evacuation, which were then
a staccato fashion. presented at their national training academy;
631
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
632 Organization Science 18(4), pp. 631–647, © 2007 INFORMS

Figure 1 The Peter B. Lewis Building 5. the precise tolerances increased the risk of loca-
tion errors, leading the construction manager instead of
the contractors to measure the position of all pipes, win-
dows, ducts, etc.; and
6. the precisely curved interior surfaces made the
traditional techniques of acoustic control less desir-
able, stimulating a search for alternative construction
materials.
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

Note how each innovation responded to the evolving sur-

face of wakes created by prior innovations, and set the
stage for future ones by adding to the turbulence of the
innovation space.
In this paper we seek to develop an explanation of
these wakes of innovation and the recursive quality
of the innovation surface that characterized the Frank
Gehry projects we studied. The innovation turbulence
(d) the specialty metal contractor invented a water- began in Gehry’s studio, as it moved from being com-
proof shingling system that dramatically reduced the puter phobic to being a world-renowned center of digital
thickness and cost of the roofing; 3-D architecture and construction in less than a decade.
(e) the construction manager expanded its scope of In the early 1990s, while other architects were busy
work and begin providing location measurements to con- adopting 2-D computer-aided design (CAD) software,
tractors, increasing its own risk, but reducing construc- Frank Gehry refused to use 2-D CAD tools in his stu-
tion time and construction errors; and dio, in part because he disliked computers, and in part
(f) the drywall contractor became the first American because his design method relies heavily on working by
company to license a soundproof plaster system from hand with physical models. However, by 2000, Gehry
Swiss developers, which created a new line of specialty Partners LLC was regarded as the most advanced user of
business for them. 3-D digital tools in architecture (Lindsey 2001). Behind
that transformation is a fascinating network of innova-
For all these innovations, there exist alternative, tra-
tions in the distributed sociotechnical systems of AEC
ditional methods that the contractors, fire marshal, or
projects.
construction manager could have employed. However,
Gehry Partners’ adoption of digital 3-D technolo-
instead they invented a better way (less costly, more
gies disrupted the ecology of practices that had evolved
elegant, or more reliable) to accomplish their part of
among AEC project teams, allowing professional and
the overall project. Also, these innovations were car-
craft communities to mindfully deviate from their estab-
ried forward into subsequent, non-Gehry projects by lished practices of design and construction, and to create
the innovating firms, thereby strengthening each individ- new sets of self-reinforcing relationships between other
ual organization, as well as the AEC industries. Each actors (Garud and Karnoe 2001). Each firm attended to
innovation can be thought of as creating a wake in their disrupted ecology with reasoning grounded in their
their respective community of practice, but in the Lewis own facts, needs, and specifics (Swanson and Ramiller
Building project, they also served to influence each other 2004). They did not act with a global logic, but with the
as interacting wakes of innovation. Consider the follow- local logic of their own situation and concerns (Boland
ing examples: 2003). Gehry Partner’s adoption of digital 3-D represen-
1. the new roof structure enabled the exterior and inte- tations also altered traditional communication patterns
rior surfaces of the building to directly coincide with through which actors learned about each other’s plans
each other—an unusual condition in a complexly shaped and coordinated their work. Their individual mindful-
building; ness (Weick and Sutcliffe 2006) and new communication
2. the roofing shingle system enabled that parallelism practices allowed innovations that occurred in one com-
between the interior and exterior surfaces to be carried munity to occasionally be inserted into, or intercalated
out with extremely tight precision; with, the unique development process of another com-
3. the precision requirement challenged the interior munity (Galison 1997), which pushed each to expand its
surface contractors to invent new, closer tolerance, own innovation further, while at the same time strength-
curved surface construction techniques; ening interactions among them (Brown and Duguid
4. the curved interiors created a new challenge for the 2000).
fire smoke evacuation studies, stimulating the invention In the remainder of our paper, we first provide an
of new models of air flows for extremely complex inte- overview of the traditional practices in an AEC project
rior surfaces; network. Next, we synthesize a theoretical framework
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
Organization Science 18(4), pp. 631–647, © 2007 INFORMS 633

that describes how wakes of innovation within AEC research on innovation. Recent research on innovation,
networks can interact with one another as one firm’s tra- however, is beginning to recognize the importance of
jectory of innovation becomes open to discovery, nego- networks and the limitations of studying innovations in
tiation, and knowledge creation with another’s. We then a single, vertically integrated firm. Next, we will exam-
use the proposed framework to analyze our case study ine the emerging body of innovation literature that forms
data, showing how innovations with digital 3-D repre- the basis of the theoretical framework of our study.
sentations were deployed in designing and constructing
Frank Gehry’s buildings, and how they generated multi-
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

ple other innovations in the AEC system of design and Theoretical Framework of the Study
construction. Our paper concludes by posing implica- Studies of IT innovation have often explored how certain
tions for developing theories of innovation in distributed, factors push (Damanpour 1991, Swanson 1994, Zmud
heterogeneous, sociotechnical networks. 1984) or pull (Rogers 1995) innovations through mar-
kets. More recent work focuses on the user’s active
Architecture, Engineering, and role in pulling innovations, as in continuous innovation
theories (von Hippel 1998, von Hippel and von Krogh
Construction as Context for the Study of 2003), or on the social processes shaping and stabilizing
Innovation an innovation, as in theories of the social construction
Complex building projects serve as good models for of technology (Bijker 1995, Latour 1987, Williams and
the network forms of organizing that are increasingly Edge 1996). They are primarily studies of a particular
found in today’s economy. AEC projects are distributed wake of innovation in relatively homogeneous adopter
(designed and constructed by multiple, autonomous groups.
actors), heterogeneous (composed of communities with Recently, a growing body of innovation research has
distinct skills, expertise, and interests), and sociotech- shifted its focus from a single innovator to a net-
nical (requiring trust, values, and norms, as well as IT work of heterogeneous actors (Tuomi 2002, Van de Ven
capabilities and complex fabrication processes). For ease 2005). When innovation is viewed as being a distributed
of expression, we will simply refer to complex building
phenomenon, it is characterized by network effects,
projects as “distributed systems” or “project networks.”
messiness, ambiguity, and combinability (Lyytinen and
The traditional practice of a standard, AEC design/bid
Damsgaard 2001, Tuomi 2002, Van de Ven 2005).
project begins with an architect working with a client to
Because the network approach assumes an innovation
create a set of drawings (blueprints) and specifications,
space to be dynamic and volatile, it includes political
which indicate the intention and form of what is to be
and institutional influences across the network’s multiple
built. Standard contracts assume that these drawings will
be paper based and two dimensional (Stein et al. 2001). communities. Studies of the communal and networked
These drawings do not show every detail required in features of innovation (Brown and Duguid 2000, von
construction, but instead indicate typical ways in which Hippel 2005) show that increased network diversity pro-
construction elements are to be handled. The architect’s motes new combinations, fosters learning, and enables
drawings leave it to the contractors and subcontractors to faster diffusion (Tuomi 2002), but at the same time,
identify a suitable method of constructing the intended it can build thicker boundaries that inhibit their spread
elements in a workmanlike way. Contractors take the (Ferlie 2005). These scholars stress that networked inno-
architect’s drawings and use them as a basis for cre- vations are associated with new types of interactions
ating their own documents for the architect’s approval. among innovators (Van de Ven 2005), which are enabled
In these “shop drawings,” they show how they intend and mediated by a wider variety of IT artifacts (Carlile
to fabricate and install their part of the building. In 2002). Thus, IT is not just a target of innovation, but is
preparing their shop drawings, contractors ask archi- also an engine of innovation because it increases knowl-
tects questions by submitting RFIs (requests for infor- edge permeability across boundaries, thereby enabling
mation) or RFCs (requests for clarification). Even in new knowledge to be created, and distributing it in
projects of modest complexity, RFIs and RFCs are to be new ways.
expected, and contracts specify the format and timing Traditionally, the pace of innovation in networks is
that is required for making the requests and providing explained by network topologies (Ahuja 2000, Obst-
responses. Within this framework of 2-D representa- feld 2005), which themselves can be transformed by IT
tions and communications, responsibilities for errors are capabilities (Butler 2001). The new emerging patterns
traced to either the architect or the engineer for having of interaction enable further knowledge creation among
specified an inadequate design, or to the contractor for the network’s communities (Tuomi 2002). Indeed, we
having failed to apply the correct workmanlike expertise. observed that new interactions among heterogeneous
The dynamics of these project networks are distinct actors was the modus operandi in shaping the wakes
from the operations of a single firm, even though sin- of innovations that followed Gehry Partners’ introduc-
gle firms have frequently been the site for much prior tion of 3-D representations in their construction projects.
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
634 Organization Science 18(4), pp. 631–647, © 2007 INFORMS

An emphasis on the interactions and relationships among with respect to the alternatives available in their environ-
heterogeneous actors is particularly important in study- ment, and that they follow reinforced patterns of feed-
ing IT innovations, because IT is more than just the tools back relations in choosing their courses of action and
that are deployed in a project. IT is an environment—a targets of learning (Arthur 1989, March 1991). Path cre-
“web” of equipment, techniques, applications, and peo- ation, in contrast, emphasizes human agency and the
ple that creates a social context, including the history way that actors can and will mindfully deviate from
of commitments that formed that web, the infrastructure what appears to be the common expectation (a tradition-
that supports its development and use, and the social ally reinforced pattern of action) in order to sample new
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

relations and processes of its use (Kling and Scacchi experience, explore new forms of practice, and create
1982). new resources (Garud and Karnoe 2001).
Recent studies on network-based innovation have ex-
amined interactions and knowledge sharing within at Trading Zones and Intercalation of Innovations
most two communities or one firm (Bechky 2003). Although path creation highlights an actor’s entrepre-
Moreover, these studies primarily concentrated on the neurial deviation from established patterns of practice as
structural features of social networks (Ahuja 2000, a key source of innovation, it neither explains how such
Obstfeld 2005), paying little attention to the role of tech- deviations are sustained and expanded, nor how they
nology in facilitating new forms of interactions across relate to other actors’ path-creating activities. In Gehry
networks that span multiple professional communities Partners’ building projects, there was no lone actor devi-
(Boland and Tenkasi 1995) or entrepreneurial activities ating mindfully. Quite the contrary, multiple actors were
of individual actors who are embedded in such struc- deviating mindfully in response to challenges posed
tural arrangements. Thus, in order to fully account for by Gehry Partners’ digital 3-D representations—often
the complex patterns of innovations that we observed in unexpected and unrelated communities. These com-
in our study of Gehry’s digital 3-D projects, we need munities invented novel uses of 3-D representations,
a theoretical language that allows us to explain diverse crafted new knowledge and skills with them, enacted
actors’ individual innovations, as well as how those new construction practices, and generated new organi-
actors and their activities are interrelated and produce zational forms (Yoo et al. 2006). What we observed
wakes of innovation spreading beyond the boundaries of was a series of abrupt and unexpected breakdowns in
their communities. the established systemic relationships that had evolved
In summary, in order to understand and analyze the over many centuries within and across AEC communi-
wakes of innovation induced by the adoption of digital ties. The AEC industry typically follows long-standing
3-D technologies, we need to understand: (1) how sep- traditions and well-established structural arrangements
arate communities integrate and adopt digital 3-D and in their project governance, in which each element in a
other innovations into their work practices, and (2) how project’s system of relations is interrelated and interde-
and under what conditions such innovations can traverse pendent with others. A significant change in one element
heterogeneous communities. We will address the first in the AEC system can disturb the established pattern
question with the concept of path creation (Garud and of relations among other elements (Churchman 1968).
Karnoe 2001), and the second question with the notions This is especially true when the change affects the prin-
of trading zones and intercalation (Galison 1997), which ciple forms of representation among project actors, dis-
will be explained in the next sections. turbing their ability to communicate and coordinate. The
wakes of innovations that we saw were an expanding
Innovation as Path Creation trial-and-error search for new configurations of relation-
The concept of path creation (Garud and Karnoe 2001) ships among the actors, ideas, and technologies in AEC
is exemplified in Gehry’s willful intention to push the projects. We saw these new relationships take shape in
boundaries of architecture and construction. The theoret- sporadic patterns as actors’ paths crisscrossed at criti-
ical construct of path creation is a reaction against the cal junctures of a project, propelled by the expanded
limitations of path dependence as used in evolutionary deployment of 3-D representations for a specific design
economics (David 1985, Arthur 1989). Although path challenge, or a local construction task.
dependence brings a historical systems view to economic The image of wakes captures the spreading patterns of
explanations of technological innovation, and recognizes mutually dependent and reinforcing deviations that fol-
that temporality and dynamic adaptation to unanticipated lowed engagements with 3-D representations in specific
events are important in understanding how technolog- contexts. However, this image of a sociotechnical system
ical innovations emerge and are adopted over time, it resetting its dynamic equilibrium—no matter how vivid
does not normally include agency as a creative force that it was for us—did not explain why it was unfolding in
makes history by establishing new cycles of positively front of our eyes. We felt a pressing theoretical need
reinforced feedback relations. Path dependence assumes to delve more deeply into this phenomenon. In mak-
that human actors are generally passive or conservative ing sense of the staccato-like spurts of innovation that
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
Organization Science 18(4), pp. 631–647, © 2007 INFORMS 635

we observed, we drew upon Galison’s (1997) seminal These three concepts—path creation by entrepreneurial
study of the history of microphysics in the 20th century. actors, intercalated innovations across communities, and
Galison reveals the dynamics of concurrent and inde- the intensity of interactions in trading zones at perme-
pendent paths of innovation associated with at least three able boundaries—will be used to explain how and why
communities necessary to do microphysics. These com- wakes of innovations across AEC communities followed
munities are concerned with instruments (measurement the adoption of digital 3-D representations in Gehry
and observation), experiments (design and execution), Partners’ projects.
and rival theories. He shows that each of these commu-
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

nities innovates at different rates and generates its own

(path-dependent) innovation trajectory. Innovation shifts Research Setting and Method
within and between communities are not simultaneous— According to Yin (1994, p. 40), a single case study can
a shift in theory is not necessarily accompanied by a be useful when the purpose of the study is revelatory
synchronic shift in instrumentation or experimentation. in nature. In our case studies of Gehry Partners’ build-
He refers to this staccato pattern as an intercalated pro- ing projects, we sought: (a) to examine how patterns
cess of change, in which multiple innovations that occur of diverse innovations unfolded within and across dis-
at different rates and times in the different communities tributed systems; and (b) to analyze the mechanisms by
are inserted into the project network. which such innovations take place and interact in their
Galison’s idea of intercalated innovations across mul- project networks. With a single set of cases, we sought
tiple communities offered us a fruitful way to make to firmly ground our theoretical accounts with empir-
sense of the heterogeneous and nonlinear innovation pro- ical observations (Eisenhardt 1989, Martin and Turner
cesses we observed. Like microphysics, the AEC indus- 1986). Overall, we expected these theoretical accounts
tries consist of many distinct communities, each with to offer faithful explanations of the wakes of innovation
a unique professional identity and its associated exper- in the distributed systems of the large AEC projects we
tise, skills, tools, and technologies. These communities studied.
chart their own innovation paths, which follow their During our study, we conducted interviews with the
own logic at their own tempos, and produce intercalated architects, engineers, construction managers, and sub-
changes as their paths cross into each other’s innovation contractors who worked on a range of Gehry Partners’
trajectories. The organization and structure of commu- projects. We visited the home offices of participants
nity boundaries, which either enable or inhibit cross- to review their internal processes, to examine construc-
boundary learning and knowledge sharing, are critical in tion documents with them, and to observe their design
shaping innovation trajectories of each community and and construction practices. The objectives of these inter-
the set of communities as a whole. In particular, we views were to identify: (a) the ways in which the various
need to explore what makes boundaries permeable so actors experienced the digital 3-D representations used
that dynamic learning can take place. by Gehry Partners to be different from their usual doc-
Galison (1997) observes that structural features such umentation and information-sharing practices (i.e., was
as formal organization charts, responsibilities, and con- there a change in routines and knowledge related to their
trol structures, as well as social networks and informal use of digital 3-D representations indicating changes in
communications defined by friendships, mentor-student their trading zones?), (b) the ways in which the actors
relationships, memberships in committees, or traffic adapted to those differences (i.e., how these new repre-
patterns of buildings, all affected the permeability of sentations and boundary exchanges related to their own
boundaries and interaction of knowledge needed to invention of new routines in construction and design
do microphysics. He observed that when community practices), (c) the difference it made to their practice
boundaries cross or overlap during a project, trad- (i.e., did these deviations lead to new and reinforced
ing zones can emerge (Kellogg et al. 2006). In these modes of experiential learning and innovation?), and (d)
trading zones, rudimentary cross-specialty languages can the ways in which they communicated or shared their
develop, allowing formerly distinct knowledge to flow in innovation experiences with other communities (i.e., did
both directions. Galison sees this as similar to the pidgin learning take place in trading zones?).
and creolized languages that emerge when different cul- In collecting the data, we were particularly interested
tures come into sustained close proximity. The stronger in how the deployment of digital 3-D leads to the gen-
and richer the trading zone, the easier it is for learning eration of new knowledge, acquisition of new technolo-
to travel from one community into another, and for com- gies, retraining of employees, hiring of employees with
munities to coproduce new knowledge. Trading zones new skills, restructuring of organizations, changes in
are thus both physical and cognitive arenas for commu- routines, changes in how actors perceived and enacted
nities with separate innovation trajectories to negotiate, their environment, changes in how actors related to
collaborate, and learn through mutual perspective mak- other actors, and ultimately, changes in an actor’s iden-
ing and perspective taking (Boland and Tenkasi 1995). tity. We first focused on how Gehry Partners changed
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
636 Organization Science 18(4), pp. 631–647, © 2007 INFORMS

their work practices following their adoption of digital Gehry Partners LLP played a central role in innovating
3-D representations. Thereafter, we examined how their with 3-D technologies, and the complexity of Gehry’s
clients, contractors, engineers, and regulators appropri- designs, along with the use of 3-D representations, cre-
ated digital 3-D tools. To accomplish our research goals, ated the trading zones in which the wakes of innovations
we asked Gehry Partners associates (partner architects, occurred. We will next describe the main uses of 3-D in
designers, project architects, and CATIA™ operators) Gehry’s office and the nexus of contexts in which other
how designing and constructing recent buildings (and the actors applied and appropriated 3-D technologies.
Peter B. Lewis building in particular) differed from their
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

earlier projects, and what mindful deviations took place. Gehry Partner’s Use of Digital 3-D Technologies
Second, we asked contractors and regulators involved Over the last decade the use of digital 3-D representa-
in Gehry projects to compare and contrast that expe- tions such as CATIA™ and Rhino™ software has become
rience with other projects they have had in the past more commonplace in AEC due to the lowered cost and
years. We also asked them about projects subsequent to improved functionality of 3-D tools. In many ways, dig-
their involvement in the Gehry project in order to assess ital 3-D is a potentially disruptive innovation because
ways in which innovations from the Gehry project were it uses enhanced computer capabilities to fundamentally
carried forward into later projects. We videotaped and change the representational systems (or “genres”) by
photographed models and drawings of buildings, and which knowledge about a building is represented, com-
followed demonstrations of how they used 3-D and other municated, manipulated, and shared. It meets the defini-
digital technologies and what changes in their work tion of radical innovation (Zaltman et al. 1973) in that it
practices occurred to help understand how knowledge is a “significant departure from existing practices” cog-
was generated and shared. We started with a set of ini- nitively, materially, and relationally. In what follows we
tial interviewees for each project obtained from sponsors highlight the main functionalities of the 3-D software
or Gehry Partners. We updated and expanded the list tools that have been adopted in Gehry’s office between
through snowballing as we traced the network of rela- 1992–2004, and the benefits they offered (see Lindsey
tions associated with each construction project. 2001 for detailed description of their use of 3-D CAD
A total of 83 interviews—48 interviewees from 16 in their design practices).
organizations—were conducted.1 Most interviews were Digital 3-D representations are more interactive and
transcribed and analyzed verbatim. We also contacted intelligent than 2-D representations (Shih 1996), and
some interviewees afterwards to clarify issues raised in allow more daringly shaped structures to be built cost
the interviews and to collect additional documents from effectively. Three-dimensional tools allow for a full
them. Our data collection branched out from an initial visualization of designs in actual scale, and support
focus on three Gehry Partners’ projects (the Peter B. simulation as well as integration and coordination of
Lewis Building, the MIT Stata Center, and the Princeton detailed design information for digital manufacturing.
Science Library) as we traced the involvement of key When fully developed, a digital 3-D representation is a
actors in earlier and later Gehry Projects. Following complete digital prototype of a building that acts like
these actors forward and backward in time led us to the actual building. A digital 3-D prototype can reflect
study aspects of the following additional Gehry projects: all elements of the design and construction process and
the “Fish Sculpture” at the Barcelona Olympics, the permits different tasks and elements of the building to
Guggenheim Museum in Bilbao, Spain, the Experience
be dynamically interconnected, so that multiple actors
Music Project in Seattle, the Millennium Band Shell in
can coordinate and share their knowledge more openly
Chicago, and the Bard College Center for the Perform-
(Lacourse 2001). Digital 3-D also enhances designers’
ing Arts. Following the recommendations of Eisenhardt
cognition by allowing multiple aspects of a building
(1989), Boyatzis (1998), and Yin (1994), we then repeat-
to be displayed in a single representation and explored
edly read the transcripts and interview notes and viewed
interactively (Baba and Nobeoka 1998). It also enables
the photos and video streams to identify key themes and
designers to visually move back and forth between views
topics in each interviewee’s report of their use of 3-D
of detail and views of the whole in a hermeneutic pro-
representations and their participation in the construction
cess of learning (Boland et al. 1994). In addition, digital
project.
3-D affects how design information is used by project
stakeholders, making previously tacit design knowledge
Innovating with Digital 3-D Representations explicit (Yap et al. 2003) and enabling richer, more
by Gehry Partners detailed knowledge transfer and interactions. This fea-
Due to the distributed nature of the AEC system, path- ture was critical for Frank Gehry because he strives to
creating innovations did not always begin with Gehry fully engage his clients, consultants, and contractors dur-
Partners. Sometimes it was the software developers, ing the design process.
sometimes the contractors, and sometimes the fabrica- Another particularly valuable aspect of 3-D represen-
tors who initiated a new wake. However, we note that tations is that they integrate structural elements of the
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
Organization Science 18(4), pp. 631–647, © 2007 INFORMS 637

building design (such as size, material properties, and (1997), they were sporadic and opportunistic—forming
cost) with its 3-D geometry (or solid modeling). The and moving between firms as each firm followed its own
combined geometric and structural information can be local logic and innovation trajectory.
stored, modified, and later reused through a repository
(Toupin 2001). Designers can easily manipulate complex In the Beginning: Intercalation of 3-D from
surface geometries, more freely experiment with non- Aerospace into Frank Gehry’s Design Practice
traditional shapes, and quickly explore their costs and Frank Gehry’s design method is based on building phys-
structural implications, which aligns with Gehry’s exper- ical, three-dimensional models by hand, which, ironi-
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

imental design style (Lindsey 2001). Changes to sections cally, facilitated his appropriation of 3-D technologies.
of an integrated 3-D model can be automatically prop- His method is to always begin with a free-form, impres-
agated (Argyres 1999), showing consequent changes to sionistic sketch intended to convey the spatial and emo-
adjacent sectors of the building, and helping to manage tional sense he is trying to achieve with the building,
complex design processes (Greco 2001). For all these which he calls a “dream image.” (Frank Gehry, Decem-
reasons, 3-D representations have the potential to dra- ber 6, 2004). After that, he and his colleagues build
matically reduce the cost, effort, and error rates of design models with wood, paper, metal, cloth, plastic, or what-
and construction projects (Koutamanis 2000). ever is at hand in an effort to explore the design ideas
Global positioning systems can be combined with 3-D latent in his original sketch. From the earliest massing
tools to change the traditional process of measurement models, to rough study and design models, to the more
at a construction site by creating accurate location data refined final models, there is no drafting on paper and
and managing it centrally as part of a complete digi- no 2-D CAD involved. As one of the staff expressed it:
tal prototype (Gragg 1999). Three-dimensional represen- We persisted much longer in Frank’s office where if you
tations of the building show every point on a line or had a computer, then you weren’t a real architect. You
surface as an x y z Euclidian coordinate relative to know, you were some kind of drafting support team. (Reg
an established absolute X = 0, Y = 0, Z = 0 point at Prentice, November 13, 2003)
the building site. Instead of locating building elements During a single design project, hundreds of physical
in relation to each other with a tape measure, special- models are constructed to explore design options and
ized surveyors locate sets of absolute points within a alternative forms. They include models of the site and
3-D space,2 thus reducing the possibility of propagating its surroundings, models of interior spaces, models of
measurement errors. Three-dimensional representations the entire building, and models of especially important
can also be used to generate accurate data for guiding elements of the building. The design team works at mul-
downstream fabrication. As a result, Gehry partners have tiple scales simultaneously (usually three) because each
intensively collaborated with a new breed of contractors, scale reveals different problems and possibilities with
like A. Zahner & Co. and Permasteelisa, who are lead- the evolving design. Other architects use physical mod-
ers in the use of advanced digital fabrication techniques els, but not as their primary design and thinking tools
(Koutamanis 2000). as Frank Gehry does. They use physical models primar-
In the next section, we summarize our main findings ily as presentation devices with their clients, not as a
under four headings that emerged from our analysis of tool for thinking during design. In an important sense,
the interviews and documentary data. First, we present Gehry’s emphasis on physical models and lack of expe-
how Gehry Partners adopted 3-D technologies from the rience with 2-D CAD tools meant lower learning barri-
aerospace industry. Second, we examine a series of path- ers in adapting to digital 3-D. When they began using
creating events within Gehry Partners that surrounded 3-D tools, Gehry’s staff did not have to unlearn practices
the initial adoption of 3-D. Third, we explore the uses of that centered on using digital 2-D. In contrast, they were
digital 3-D technologies within the heterogeneous net- able to build upon their physical 3-D design practice by
work of actors in construction projects. These findings digitizing their most precise physical models.
show how trading zones emerged among participating A fortuitous event that enabled Frank Gehry’s adop-
communities and how intercalation of innovations took tion of 3-D came in the late 1980s, when he won a
place. Finally, we discuss how the multiple path-creating competition for the design of the Disney Concert Hall
activities within Gehry Partners’ projects resulted in a in Los Angeles, which he worked on in several phases
stream of innovations among participating firms at dif- during the 90s until the hall finally opened in the fall
ferent times and generated the wakes of innovation we of 2003. Disney Concert Hall was much larger than any
observed. These wakes result from concurrent path cre- of the projects Gehry had undertaken to that date. To
ation events in technologies, work arrangements, con- successfully carry out its design and construction, Frank
struction methods, and forms of knowledge among a Gehry recruited Jim Glymph to join his office as a senior
project’s architects, engineers, construction managers, partner. Jim had extensive experience and was known for
contractors, and trades. These multiple innovations were his design ability, his deep technical knowledge of con-
never synchronized, but similar to Galison’s analysis struction, and his project management skills. Jim firmly
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
638 Organization Science 18(4), pp. 631–647, © 2007 INFORMS

believed that computer-based tools would be necessary improve the quality of outcomes. The software capabili-
to support the design and construction of the increas- ties of CATIA were just one element in the many inno-
ingly complex surface geometries imagined in Gehry’s vations that emerged across the sociotechnical network
studio. He began to explore computer tools that would of the Fish Sculpture project. The unique work practices
satisfy some or most of their needs, but concluded that that they enacted, with architects, fabricators, and con-
none of the available tools for architectural design (e.g., structors together in one trailer, and the new knowledge
AutoCAD) were suitable. base provided by the 3-D building model, along with
The architectural commission that triggered adoption intense communication protocols, fewer management
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

of digital 3-D technology at Gehry Partners was the oversight restrictions, and a thoroughly collaborative
design of a pavilion for the 1992 Barcelona Olympics. problem-solving approach, were necessary additional
After completing the Pavilion, Frank Gehry proposed innovations. The Fish project effectively created new
construction of a separate Fish Sculpture on the Olympic trading zones among architects, engineers, and fabrica-
grounds. He had long been fascinated by fish shapes, tors. They learned that the use of 3-D representations
and had developed a physical model of his sculp- allowed project members to participate in frequent trad-
ture, but no detailed plans, when the proposal was ing of expertise and coproduction of knowledge during
accepted. The Fish Sculpture had an extremely tight, all phases of design and construction. This intense level
six-month timetable and a limited budget. At this highly of continuous collaboration and invention has never hap-
constrained moment, Glymph renewed his search for pened again, although it has continued to serve as an
suitable software tools, and found that the CATIA™ sys- ideal “organizing vision” for present and future projects.
tem, which had originally been developed to design the
French Mirage fighter jet, filled many of their needs and Path Creation: Integration and Expansion of
could offer a possible solution. In addition to adopting 3-D into Frank Gehry’s Design Practice
the CATIA™ software, Glymph proposed that they work The experience of the Fish Project taught Gehry Part-
closely with a small team of contractors who would also ners that 3-D tools could be effectively integrated into
use the tool, and they could attempt to do the Fish Sculp- their firm in ways that complemented and empowered
ture project “paperless.” their existing design practice. The 3-D representations
These unique events—the Disney Concert Hall com- offered higher levels of precision and flexibility, gener-
petition, the recruitment of Jim Glymph, and the accep- ated 2-D drawings as a by-product, and also allowed the
tance of the Fish Sculpture project—were, in retrospect, continued use of physical models as the primary design
turning points that set the stage for a significant shift in medium. Digital 3-D tools also enabled them to scale
the trajectory of Frank Gehry’s architectural practice in small physical models up to larger prototypes without
losing the subtleties of a design. As a result, 3-D rep-
ways that were not evident at the time. The Fish project
resentations became a central element of the emerging
enabled them to gain positive experience in using 3-D
digital design practice at Gehry Partners. Two significant
representations through an entire construction project,
steps in developing their digital practice were the design
and to modify their design practice to benefit from 3-D
and construction of the Guggenheim Museum in Bilbao,
tools.
Spain (1991–1997), and the Experience Music Project
During the Fish Sculpture project, Gehry and the con-
(EMP) in Seattle, Washington (1995–2000). However,
tractors deviated from several traditional building prac-
Gehry Partners have remained committed to the use of
tices in order to realize the benefits of using digital 3-D
physical models as the primary medium for exploring
models. As Jim Glymph recounts it:
design ideas:
The first project (using 3-D) was the most successful we When we’ve got a very free-form gestural model, which
may have ever had.    And what convinced us was we has all this energy in it, one of the things is to capture it.
got the job done a month early. Everybody made money, And if you go to substandard Euclidian geometry and you
there were no extras – that never happens. That convinced cut sections, and you build this thing up, and you get nice
us that “wait a minute, what we just did because we arcs that are close and everything just seems so close. But
had to, maybe we should be doing this all the time, it when you get the model put back together it’s dead. And
worked.” But the other thing that happened was a lot the measurable differences are tiny and    even I was sur-
of (management) rules were suspended in order to allow prised.    he (Frank Gehry) tried to create an illusion in
that to happen    . (Jim Glymph, November 9, 2002) this building (Bilbao) and it’s a very subtle curvature that
The Fish construction turned out to be a striking you would think would make no difference. The building
without it is completely dead. So there are these gestu-
success—a daring sculpture completed ahead of sched-
ral things, which I think, are why people have emotional
ule and on budget. The project significantly increased reactions to Frank’s buildings. There’s something about
Gehry Partners confidence in the use of digital 3-D tech- building something with your hands that leaves some-
nology, as well as their belief that if they could effec- thing we can all recognize, maybe not consciously, but
tively break from traditional practices within the industry, subconsciously. It humanizes it.
they could create better collaboration on projects and (Jim Glymph, November 9, 2002, p. 4)
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
Organization Science 18(4), pp. 631–647, © 2007 INFORMS 639

I don’t believe that a computer model will ever be able projects, enabled design and construction knowledge to
to tell me what a physical model can. flow more freely in both directions, and provided oppor-
(Craig Webb, Design Partner, Gehry Partners, Decem- tunities for active trading among the different commu-
ber 22, 2003) nities. Within these newly invigorated trading zones,
The Guggenheim Museum in Bilbao, Spain was their alternative beliefs and knowledge could be negotiated
first use of CATIA™ to support documentation and com- without homogenizing the architect’s intention for the
munications among contractors during a large-scale con- building, or the unique practices within the different
communities. (Boland and Tenkasi 1995).
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

struction project, although 3-D technologies were used

only for the most difficult and critical parts of the The traditional 2-D based process in AEC resembles
exteriors and main gallery designs. In the EMP build- Manning’s (1979) concept of a semiotic loose coupling
ing, Gehry Partners expanded the use of CATIA™ and that characterized interactions between a police com-
employed 3-D representations as the primary contract munication system and people making emergency calls.
documents for the whole building and all its parts. Dur- Like Manning’s police communication system, the stan-
ing EMP, they began to use layered CATIA™ models to dardized 2-D blueprints enabled actors in diverse com-
communicate separately with different trades and con- munities to “conventionalize” their interactions, creating
tractors. During the Peter B. Lewis Building project an adequate basis for coordination (Stinchcombe 1959,
(1997–2002), they used CATIA™ ’s capabilities during Bechky 2006). Actors in AEC projects know how to
design to identify double curved surfaces in complex encode, classify, decode, and transform ideas with the
exterior and interior geometries in order to reduce con- 2-D code shared by all participants. When such 2-D
struction cost. By the time Gehry Partners were com- information crosses boundaries among communities dur-
pleting work on the Lewis Building, their architects had ing a project, it: (a) maintains the integrity, consis-
begun using CATIA™ to create fast prototypes during the tency, and autonomy of each participating actor; and
initial design phase of their projects in order to develop (b) loosens couplings between actors by hiding details of
early and accurate cost estimates. their construction practices that cannot be readily com-
Gehry Partners use of 3-D representations to commu- municated through 2-D drawings.
nicate with contractors and clients throughout design and Such loose coupling is well entrenched in most con-
construction has not decreased their reliance on physi- struction practices, as the project architect for the Peter
cal models—instead, it has added new routines to their B. Lewis building explained to us:
design practice, whereby they analyze constructability, With typical projects, many architects have a standoff
cost, and even maintenance- or energy-related aspects of position from contractors. They just basically enforce
their architectural designs earlier than previously possi- their documents and their specifications and criticize.
ble. During their appropriation of 3-D technologies by And they expect that the contractor knows how to do
Gehry Partners, the horizon of what they feel confi- everything. They don’t talk about process - they just talk
dent to design and build has expanded, and the agency about results. Architects run around and measure. Mea-
they can jointly mobilize by using digital 3-D represen- sure if the deviation is in acceptable limits and write up
reports     (Gerhard Mayer, September 20, 2002, p. 11)
tations has become more powerful and pervasive. This
has resulted in a positive feedback cycle (Levitt and However, when 2-D blueprints were replaced with dig-
March 1988, March 1991) in which learning how to ital 3-D representations at the AEC community bound-
use 3-D technologies throughout the construction pro- aries, the traditional contract language and associated
cess has enabled more daring structural forms to be suc- principles of loose coupling did not provide actors with
cessfully built. The growing sense of confidence is well sufficient ability to understand and negotiate their new
echoed by comments one of the lead architects made roles, or with a sufficient understanding of their risk.
about designing the Peter B. Lewis Building: This is the case even though digital 3-D representa-
   CATIA™ makes the whole process and the whole tions carry much more information about the building
method to build it somewhat predictable. And that’s what than the 2-D images do, because the standard contrac-
all our processes are looking for.    Without CATIA™ , it tual relations and traditional communication patterns in
would have not been possible to do it (the Lewis Build- the construction industry are inseparable from the use
ing). (Gerhard Mayer, September 20, 2002, p. 8) of 2-D representations at the community borders (Stein
et al. 2001). They may have more information with 3-D
Building Trading Zones with 3-D Tools representations, but they also have less ability to trace
By adopting 3-D representations in their design practice, the cause of an error, or even to agree on what infor-
Gehry Partners disrupted many traditional patterns of mation the representation contains, because the number
interaction that are usually taken for granted within the of perspectives it makes available is indefinitely large.
AEC system, because they shared more information with Thus, the use of 3-D representations at community bor-
more project members earlier. This increased boundary- ders on an AEC project challenged the key organiz-
spanning and knowledge-trading activities during the ing practices and encouraged new interactions in trading
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
640 Organization Science 18(4), pp. 631–647, © 2007 INFORMS

zones, where knowledge creation across communities conditions before fabricating the next element to be
could take place. installed. In fitting the windows, this would mean mea-
In these trading zones, the architects, engineers, con- suring the actual sizes of the openings in the concrete
struction managers, contractors, and subcontractors could walls before preparing the metal frames and glazing
continually redefine and mutually adjust their relation- for installation. At the Lewis Building project, however,
ships through the use and interpretation of digital 3-D because of the extremely complex geometries of the
representations as “boundary objects” (Star and Griese- walls and the windows, manufacturing of metal frames
and glazing required more lead time than the traditional
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

mer 1989). Such changes made project management

an interdependent and mediated form of organization, approach of field measurement would allow. The con-
displaying what Thompson (1967) calls coordination struction manager recounted this incident as follows:
through “mutual adjustment.” These new project net- So, [the window and glazing subcontractor] came to me
works acted more like tightly coupled systems, in which and says, I can field measure but you won’t get the win-
actors are less independent, with identities that become dow for 16 weeks, and so therefore you can’t put the
blurred as their tasks affect each other continuously, sig- brick in. I can’t do that. I said, you go make it per
nificantly, and directly. CATIA, I’m waiving your requirement to field measure,
A dramatic example of how these new trading zones I’m guaranteeing that concrete opening will be where it’s
operated comes from GQ Contractors, the plaster and supposed to be, and that that window will fit provided
drywall subcontractor on the Lewis Building project. you build it the way the architect designed it     (Dan
Seib, March 4, 2003, p. 42)
Their chief operating officer reported that in his 20 plus
years of working in the industry, he had spent fewer In the future, 3-D representations are expected to en-
than eight hours in architects’ offices. Normally, they gender even more open and intense trading zones as
would receive drawings, analyze them to make a bid, the models become more fully parametric and con-
rework them as shop drawings if they were awarded tain cost, construction time, and other information for
the contract, and do the work. On the Lewis Building each element of the building. Theoretically, such 3-D
project, however, he spent the equivalent of 17 weeks representations will form one central model with all
in the architect’s office, working with 3-D modelers on the information needed by all the contractors and
detailed CATIA™ models to plan how the framing and consultants—information that contractors previously had
drywalling could be done. to create by themselves from the architect’s 2-D
blueprints. A subcontractor commented explicitly about
I have never, ever spent more than an hour in an archi-
tect’s office prior to this job. And I spent 22 trips, 4 and this increasingly open and intense trading within his own
5 days at a time in their office. And I spent some days firm:
where I was in there at 8:00 in the morning and I didn’t It used to be this problem that Zahner had: this con-
get out of there until 10 or 11 at night, working on this crete wall separating the shop from the office,    Well
frame. (Ed Seller, September 18, 2002, p. 8) then they built the engineering into the shop. And they’re
going on and moving into the shop now. Then all of a
Similarly, in these newly shaped trading zones, the sudden all the people in the shop are coming in here
temporal organization of work deviated significantly the office
mixing in with engineering people and mix-
from that of traditional construction work. The construc- ing people from the trade and so it’s kind of like, tear
tion manager and key contractors were brought into the that wall down. (Bill Zahner, May 17, 2002, p. 22)
process at the very beginning stages of the design, with a
“design assist” contract. During this phase, key contrac- Generating Wakes of Innovation
tors were involved in developing the 3-D models in order The use of 3-D during the design and construction of
to help invent better ways to build parts of the building the projects that we studied expanded and strengthened
and to reduce cost. The design assist contract creates Gehry Partners’ architectural practice. However, con-
a special trading zone for mutual learning, negotiation, structing Frank Gehry’s designs could not have been
and invention between architect and contractor during accomplished without myriad other innovations emerg-
the design process, without a commitment to award the ing from multiple actors in trading zones between the
actual construction contract. different communities that collaborated with them.
Finally, the knowledge created in trading zones can These involved intersecting wakes of innovations in
change the temporal structure of a project. Some activi- a community’s tools, technologies, work practices, and
ties that had been traditionally carried out in a sequen- organization structures and strategies. The multiple inno-
tial order could now be done concurrently, tightening vations and the trading zones that enabled them to
their interdependencies. The window fitting at the Lewis flow between communities emerged from each firm’s
Building provides a vivid example of this. The long- response to the potential of 3-D technologies in light
standing method in construction is a loose coupling of their own unique history and strategy in the project
technique of making “field measurement” of “as built” setting. Not all firms involved in the projects made
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
Organization Science 18(4), pp. 631–647, © 2007 INFORMS 641

significant adjustments or innovations in response to the not submitting to an altruistic common good. Their
disruption posed by 3-D. Some, like a concrete con- push to innovate was shaped by their professional
tractor and a steel fabricator that we studied, had little identity, unique vocabularies, and craft-specific knowl-
interest in changing their work practices or technologies, edge, rather than by shared identity, common vocabu-
because they believed their traditional practices were laries, and mutual understanding. Innovations in these
sufficient. Others innovated deeply, like A. Zahner and separate communities followed their own temporal logic
Co. (metal fabricators), DeSimone Structural Engineer- and structure rather than aligning with a coherent and
ing, and Columbia Wire and Iron, who were already far synchronized plan. Still, underneath this seemingly ran-
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

along in their use of 3-D, and pushed even further. Still dom phenomenon, we found systemic and structural pat-
others, agnostics like the construction manager (Hunt terns of interrelations among the innovations generated
Construction) were more cautious in embracing innova- by distinct actors.
tions, accommodating 3-D technology at the work site to Path-creating agency by one actor sets an innovation
the extent that they could effectively work and trade with trajectory for that firm. However, its trajectory can cross
Gehry and other contractors. As a result, they pragmat- those of other actors in an AEC project if specific con-
ically incorporated some lasting innovations into their ditions for trading knowledge are present. Such encoun-
firms, but they did not make any large-scale practice ters can become occasions for change in both actors’
changes. paths, setting up multiple wakes of innovation, while
Across this spectrum of engagement in path cre- each actor continues traversing their own, but now newly
ation, firms also moved at different temporal paces. At shaped, path. The primary example of path-creating
GQ Contractors, for instance, there was an immediate, activity traversing communities in our study is the adop-
early embrace of 3-D technology, some changes in their tion of digital 3-D representations by Gehry Partners.
work practices during construction (including the inven- This change opened a series of new trading zones for
tion of new, patentable framing techniques), and later innovations in tools, techniques, management, and gov-
changes in their structures and strategies. At Hoffman ernance within Gehry Partners’ construction projects,
Construction, we saw a strong commitment to 3-D tech- and reshaped the innovation trajectories of many project
nology at the very beginning of the Experience Music actors. Furthermore, subsequent innovations by those
Project, when they put CATIA™ stations in all subcon- actors brought about changes in Gehry Partners’ own
tractor offices and insisted that they learn and use 3-D. use of 3-D tools, as well as their design and managerial
At the end of the project, they were pleased to say: practices—making the innovation process reflexive and
“even the ditch diggers were using 3-D.” They subse- chaotic. Just as multiple wakes create unpredictable pat-
quently created a new division dedicated to 3-D based terns of interactions when they collide with one another
projects. Examples of the wakes of innovation that we on an otherwise placid surface, we found innovations
saw appearing at different paces in trading zones across from different communities producing nonlinear patterns
communities are summarized below in Table 1. Sim- of interactions.
ilar to the example from the Lewis Building that we
presented in the introduction, many of the innovations Dynamics of Wakes of Innovation
were conditioned by the use of 3-D technologies and Our research suggests that studying IT-induced innova-
were related to other innovations during the project. For tions in distributed systems is not only a question of why
instance, the new form of professional liability insur- and how a particular piece of technology is developed
ance developed for the MIT Stata Center project was and adopted by some users or populations. It is also a
adopted to encourage the free flow of information in question of identifying and accounting for a multiplicity
trading zones, by reducing the threat of litigation from of innovations and their associated cognitive and mate-
using 3-D representations; Hunt’s use of workstations rial processes as they unfold within and across diverse
in their construction site trailer engaged all parties in a communities. This characterization differs from the tra-
more intensive sharing of knowledge; and Hunt’s cen- ditional attempts to explain the diffusion of innovations,
tralized control of measurements offered transparency which have focused on understanding the adoption of
and better understanding of each contractor’s work sta- a single technology application like enterprise resource
tus and interdependencies, reducing location errors and planning systems, or computer-aided software engineer-
rework. ing tools within an organization (Cooper and Zmud
1990, Fichman and Kemerer 1997, Orlikowski 1993),
or among homogeneous firms (Kauffman et al. 2000).
Discussion It also differs from the argument that garnering com-
Throughout our field study we were struck by the stac- plementary assets is necessary to innovate effectively
cato spurts of innovation that followed Gehry Part- (Teece 1987). Although we agree that radical innova-
ners’ use of 3-D technologies. Actors on his projects tions, like the use of digital 3-D representations in AEC,
were pursuing their own economic interests and were will be accompanied by ancillary innovations across
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
642 Organization Science 18(4), pp. 631–647, © 2007 INFORMS

Table 1 Intercalated Innovations in Three Gehry Partners’ Projects

1. Hunt Construction, Indianapolis, Indiana. (Lewis Building)

(a) Hunt brought CATIA™ workstations into their construction trailer.
(b) Hunt later began using them as a representational tool for morning meetings of subcontractors.
(c) They expanded their scope of work to include providing x y  z location services for most contractors, who are usually
responsible for their own location measurements.
(d) They adopted the x y  z measurement system on other projects, bringing the survey work for all location measurements into their
construction management responsibilities.
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

2. A. Zahner and Co., Kansas City, Missouri. (Lewis Building; MIT Stata Center; Experience Music Project)
(a) For the EMP project, they transformed their fabrication facility for mass customization of metal panels, using 3-D software
to drive machinery.
(b) They invented a new self-draining steel shingle for the Lewis Building roof.
(c) They invented new fasteners to precisely assemble exterior wall panels before welding that were used in the MIT Stata Center.
(d) They invented a way to create double-curved surfaces at comparable cost to single-curved ones, thus freeing a major
design restriction for architects.
3. GQ Contractors of Cleveland, Ohio. (Lewis Building)
(a) They used 3-D workstations at Gehry Partners studio to prepare their shop drawings, but did not bring the technology into
their shop floor.
(b) They invented new, patentable framing techniques for interior walls with curving surfaces.
(c) They developed capabilities with new sound-deadening plaster materials that lead to a new line of specialty business.
(d) They redirected their business strategy toward high-end high-quality projects and expanded into consulting
services with other drywall firms.
4. The Cleveland Fire Inspection Office. (Lewis Building)
(a) They employed new techniques for modeling and evaluating smoke behavior in complex interior spaces.
(b) They presented their Lewis Building experience to their national training academy for teaching new fire inspection techniques.
5. DeSimone Structural Engineers, New York. (Lewis Building)
(a) They invented two new methods for framing structural steel.
(b) They won a national award from their profession for the most innovative structural engineering solution of the year.
6. Columbia Wire and Iron of Portland, Oregon. (Experience Music Project)
(a) They invented a new machine for making double bends in large steel beams, enabling construction of the Experience
Music Project.
(b) They have subsequently become the major source of exotically shaped steel components.
7. Hoffman Construction, Seattle. (Experience Music Project)
(a) They placed CATIA™ workstations in all subcontractors offices and expected their use on the project.
(b) In a subsequent project for the Seattle Central Library by the leading Dutch architect, Rem Koolhaas, they adopted
complimentary 3-D imaging technologies to digitally compare the design in the CATIA models with the “as built”
surfaces of the building, for fine-tuning of precise surface locations.
(c) They have now created a new construction management company to work solely on high-profile, 3-D based
projects with demanding structures and geometries.
8. MIT, Cambridge, Massachusetts. (MIT Stata Center)
(a) The project team obtained a unique insurance policy for professional liability that put all actors on the project team under one
policy, thereby reducing incentives to search for a party to “blame” for errors.
(b) The university adopted a unique payment method that guaranteed contractors a funds disbursement within 24 hours of submitting
an approved invoice (as opposed to 45 days), thereby reducing the incentive to argue over small “change orders,” which could
delay payment by months.

communities, our analysis does not support the claim subsequently shaped and reshaped by multiple, rever-
that such assets can be reliably identified, garnered, and berating innovations in which new knowledge, technolo-
mobilized, as economic analyses would suggest. IT inno- gies, and practices were constantly generated in newly
vations involve a complex ensemble of tools, practices, built and expanding trading zones. The resulting wakes
beliefs, commitments, and history (Kling and Scacchi of innovation go beyond coercive, mimetic, and norma-
1982, Orlikowski and Iacono 2001). Our study shows tive explanations of innovation found in institutional the-
that innovations involving an IT ensemble are sporadic, ory (DiMaggio and Powell 1983). The firms in these
messy, rife with ambiguity, and situated in the local log- networks were not coerced by institutional legitimacy,
ics and trajectories of each project participant. and they did not imitate one another by following others’
We also found that the adoption of 3-D representa- “best practices.”
tions in Gehry Partners’ projects was a complex pro- Our study also portrays how uncertain the diffu-
cess that connected diverse innovations from multiple sion of a complex and radical IT innovation can be.
firms of each project network. The initial IT innovation Whereas the traditional model of diffusion of innova-
of adopting 3-D representations in Gehry’s practice was tion adopts an image of linear progression, with social
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
Organization Science 18(4), pp. 631–647, © 2007 INFORMS 643

forces pulling and pushing innovations through popula- for new ways of working together on a different set of
tions of adopters (Zmud 1984), an image of wakes and tasks. The more porous boundaries and active trading
intersecting paths paints a different landscape. In this among communities spawned multiple wakes of inno-
landscape it is difficult to pinpoint who is pulling and vations along each community’s innovation trajectory,
who is pushing. In one sense, we can argue for a cen- spreading out from their projects. Whether this more
tralized push, because nothing in this network moved tightly coupled and communicatively intense arrange-
before the challenging designs of Frank O. Gehry were ment will persist, or will become newly conventionalized
embedded in 3-D representations. However, in another around 3-D representations, remains to be seen.
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

sense, Frank Gehry is not the only highly creative archi-

tect designing buildings with challenging geometries. Driving Forces of Wakes of Innovations
Other architects could have played a similar innovator Our study suggests that the forces behind such processes
role. In addition, we can see a decentralized and dis- are a combination of design vision and mindfulness,
tributed pull as well, because Frank Gehry (or any other together with the dynamics generated in new trading
architect) would not have been able to innovate with zones.
such complex geometries without drawing on relevant
resources that became available outside their practice Design Vision and Mindfulness. Our interviewees
(e.g., CATIA™ 3-D modeling from aerospace, Com- often referred to the criticality of a mindful creation
puter Numerical Control–based fabrication technologies of the new that was shared at some level by all actors
from manufacturing, etc.), and enrolling networks of involved in the network. They were mindful of others
actors willing to experiment across communities. This and of the opportunities for invention that were open to
landscape of innovation can be thought of as a set of them in their interactions (Weick and Sutcliffe 2006).
autonomous, vibrating nodes that create complex sets For Gehry Partners, the source of mindfulness was the
of connections from relatively simple interactions taking architectural vision of Frank Gehry himself, which con-
place in their trading zones (Van de Ven et al. 1999). tinually expands and creates new forms of architectural
Instead of the project networks of AEC becoming more language, exploring new materials, techniques, and tech-
homogenized from the “diffusion” of an innovation, we nologies to fulfill his vision. There is an excitement in
see the landscape of innovations becoming more color- being mindful in this way, which keeps the possibilities
ful and rugged as new actors join the network, with each for innovation open during the construction process:
actor animated by its own unique character and voice at    we always involve, we always break the problem
their intersections. down, make sure the building works, what that means,
Studies of single innovations within unique adopter spend a massive amount of time on making it work before
groups are useful in understanding implementation prob- there’s any form giving. People always think we have
lems associated with a specific innovation. However, some idea of what it’s gonna look at the end, and we
in order to understand how IT innovations substan- never do, they make sure we don’t. So, you know, it
tially transform the practices and industrial organization is a process of not knowing where you’re going, you
of distributed systems, they need to be amended with know, you have a few basic things you do about the pro-
cess itself. Probably easier to show you at the office, but
“ecological” studies that trace intercalated innovations
they’re very minimal and beyond that, you’re reacting to
across heterogeneous networks, such as the AEC. Such
the people that you’re, the new client. So every project’s
an ecological approach to IT innovations is necessary new. (Jim Glymph, November 9, 2002)
for studying the increasingly complex technologies that
rapidly digitize wider ranges of tools, activities, pro- The experience of the Fish project provided a glimpse
cesses, and places (Zuboff 1988). In this regard, we of how the managerial and work practices of an AEC
need to examine the means by which innovations emerge project could be accomplished in a more integrated way
and are traded, the likelihood of transferring innovations by using 3-D technologies. It opened a vista of imagin-
across boundaries, the structural properties of these com- ing and experimenting with radical, digital architecture,
munities, and the way that their interrelationships affect and construction practices. As Gehry Partners has tra-
the pace, direction, and scope of innovation. versed its own innovation path over the past decade, this
Our analysis shows a systemic process of innovation initial vision has become more concrete through experi-
in which an initial (path-dependent) condition of loosely ments in multiple construction projects that have helped
coupled, dynamic stability in the distributed system of them garner an expansive suite of competencies.
AEC project networks was disturbed. The convention- We observed the contractors to be mindful of the
alized use of 2-D boundary representations to support strategic question of how their firm could become a
a loose coupling with sufficient communication for the leader or change agent in their craft. Many project par-
distributed system to complete design and construction ticipants, from architects to engineers to craft work-
projects became characterized by tighter coupling in ers, share a path creation impetus and exercise mindful
trading zones, as a subset of the communities searched control over their activities. They are, or can become,
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
644 Organization Science 18(4), pp. 631–647, © 2007 INFORMS

intrigued by the possibilities of the new, and consistently as a medium for communication and coordination with
push to expand their expertise. This reflects a con- contractors, engineers, builders, and fabricators. Even
scious desire to create something different and better though many leading architects now use advanced 3-D
than what has gone before. Their innovation trajectories tools to generate detailed structural and surface models,
were stimulated by the demands and opportunities of they rarely share these models with contractors and fab-
high-profile projects and the state of development within ricators “due to the concerns related to litigation, risk
each firm. Sometimes they were more advanced in the sharing, or intellectual property rights” (Dale Stenning,
use of technologies, sometimes more collaborative in its Hoffman Construction, December 16, 2003). As a result,
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

project organization, or sometimes more experimental their use of 3-D tools, no matter how innovative it may
with materials and work practices. be internally, does not have second-order effects in gen-
Mindfulness, then, became a kind of propelling force erating spurts of innovation in the same way we saw in
that kept innovation rippling out from each part of the Gehry Partners’ case.
network—especially for those who progressed further Our study supports the position that IT innovation
in the adoption of 3-D technologies. With a state of does not take place through a single heroic innovator
mindfulness, everyone on the project had the potential or company (Swanson and Ramiller 1997). The fact
to become an innovator. One constructor commented on that Frank Gehry does not use computers and dislikes
this spirit as follows: their screen images shows that 3-D technology adoption
Because the format in which these documents was put at Gehry Partners cannot be explained by his personal
together on this job [it] did not complete the design, and needs or goals. However, Frank Gehry’s role in shaping
we found ourselves completing the design. So being that this innovation is critical because it is his architectural
this was a magnified case of that, now we’re paying more vision and collaborative practices that challenge tradi-
attention to the design of the other jobs to see where tional ways of organizing project work and open trading
the architect left off so that we can address design issues zones where actors can exchange ideas and invent meth-
early on, because a big problem here was the incom- ods for building. He has triggered a socially and tempo-
plete design. The shape was there, we always knew the
shape, but we didn’t know how we were going to create it
rally distributed transformation of artifacts (Tuomi 2002)
because the documents didn’t fulfill all, they didn’t cover over a 10-year period and has reconfigured structural
all the problems. And that’s what they ran into in the relations across a diverse network of communities in the
field, was that they had this set of documents that they AEC industries. Our study also shows how seemingly
typically would use and it would tell them everything separate innovations actually connect in time and space.
where they didn’t have it. And so even the carpenters We did not observe an immediate and direct “balancing”
in the field became part of the design, because they had of such innovations within or across firms in the form of
to figure out ways to do things     Actually, they had instant “gestalt shift.” Instead, we saw the emergence
to help complete some of the design as far as structural of pockets of productive communication within trading
elements. (Ed Sellers, November 11, 2002, p. 13)
zones, where communities who operated within different
Trading Zones and Intercalated Innovations. Al- innovation trajectories learned to negotiate, argue, invent
though design vision and mindfulness provide psycho- and thereby innovate.
logical and cognitive explanations for the path-creating
innovations we observed, they do not provide a structural
explanation as to how such individual innovations came Limitations
together to form wakes of innovations. We found that There are several limitations related to this study. Our
answer in the notions of trading zones and intercalation. findings cannot be generalized across all innovation pro-
Different actors’ innovation paths were interwoven with cesses in the construction industry, because we focused
each other, and innovations from one community were on a leading and early innovator with radically new IT
inserted into another’s innovation path, thus opening a capabilities. First, such early innovators operate in high-
trading zone and changing the trajectory of both paths end markets that can attract unique resource pools (in
(Galison 1997). In addition, certain innovations became terms of architectural, engineering, and IT talent), and
mutually dependent, providing a propelling force for can invest in technology trials and design assist con-
pushing the innovation forward. Without richer trading tracts. Second, our study reports on innovation in a lux-
zones generated by 3-D technologies, a path-creating ury market by a “super star.” It is not clear how fast
innovation would most likely have remained within its and under what conditions such innovations could be
originating community. Thus, it is not only the individ- adopted by other parts of the high-end building mar-
ual actors’ path-creating activities, but also the way they ket, or in other market segments in the AEC industry.
interrelate through trading zones that creates wakes of This would require a better understanding of both cost
innovation. and demand factors related to different types of building
It is also important to note the critical role played projects, and of how the intensive use of 3-D representa-
by Gehry Partners’ explicit intention to use 3-D tools tions can reduce cost across the value chain in different
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
Organization Science 18(4), pp. 631–647, © 2007 INFORMS 645

pockets of the industry. Finally, our study focuses on of a single wake of innovation, with a network-centric
AEC, which is a distributed industry made up of many perspective that emphasizes multiple, diverse wakes of
independent subtrades, each with their own unique com- innovation emerging in a sporadic and nonlinear way
puting and information practices. The pattern of innova- from trading zones in organizational networks. The out-
tions we observed in our study may be quite different come of such wakes is not a uniform world full of
from the patterns of innovation diffusion in more con- like-minded actors and averages. Instead, it is a vibrant
centrated and centralized industries with a few dominant and interesting world, filled with distinctive colors and
firms. voices, producing everchanging combinations of patterns
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

of technologies, routines, and cognition.

Directions for Future Research
Our findings suggest several avenues for future research. Acknowledgments
A first research question is when and how do 3-D tech- The authors are indebted to the senior editor and the reviewers
nologies or other novel representations constitute effec- for their very helpful suggestions and patient guidance of this
tive boundary objects in trading zones (Henderson 1991, manuscript.
Star and Griesemer 1989). In particular, we wonder how
Gehry Partners’ representational practices differ from Endnotes
1
those of other highly accomplished architects who do not We interviewed members of the following organizations,
use digital 3-D representations, or use them differently among others: Gehry Partners (Los Angeles), Hunt Construc-
to influence innovation and learning. tion Company (Indianapolis), Hoffman Construction Com-
pany (Seattle), A. Zahner Company (Kansas City), Mariani
Second, 3-D modeling tools are just one of many
Steel Fabricators (Toronto), Cleveland Fire Prevention Bureau,
information technologies that mediate the coordination Cleveland Building Department, Desimone Engineering (New
and communication in distributed systems. It would York), Donnelly Concrete (Cleveland), Spark Steel (Toronto),
be interesting to investigate the generative capabilities Columbia Wire and Iron (Portland), GQ Contractors (Cleve-
of a broad range of digital information technologies, land), Skanska Construction, USA (Boston), Coop Him-
including RFID, product-lifecycle tools, or scientific col- melblau Architects (Vienna), Etalliers Jean Nouvel (Paris),
laborations for enabling wakes of innovation in other Westbrook Architects (Cleveland), and Magnusson Klemencic
distributed networks. Associates (Seattle).
2
Third, our study suggests the opportunity for using With complex buildings like Frank Gehry’s, this increases
other methods in examining wakes of innovations in dis- exponentially the volume and complexity of measurements.
tributed systems. In particular, we anticipate that future For example, in the Peter B. Lewis building the amount of
measures was close to 500,000 points, which were all managed
analyses of wakes of innovation could use formal and
centrally in a separate (Excel) data base.
dynamic approaches such as multimodal network analy-
sis or simulation, which would allow the examination of
References
different patterns of innovation between heterogeneous Ahuja, G. 2000. Collaboration networks, structural holes, and inno-
communities (Monge and Contractor 2003). vation: A longitudinal study. Admin. Sci. Quart. 45 425–455.
Finally, the processes of path creation and interca- Argyres, N. 1999. The impact of information technology on coordi-
lated innovations should be generalized and validated nation: Evidence from the B-2 “Stealth” bomber. Organ. Sci.
in other industries where IT innovation forms a central 10(2) 162–180.
locus for triggering business transformation. Increasing Arthur, W. B. 1989. Competing technologies, increasing returns, and
digitization of products and business processes allows lock-in by historical events. Econom. J. 89 116–131.
for innovations at boundaries by linking heterogeneous Baba, Y., K. Nobeoka. 1998. Towards knowledge-based product
communities in networks. These innovations can cross development: The 3-D CAD model of knowledge creation. Res.
the boundaries of multiple communities, and therefore Policy 26(6) 643–659.
the network perspective followed in our study can be Bechky, B. A. 2003. Sharing meaning across occupational communi-
helpful. Such studies could help identify organizational ties: The transformation of understanding on a production floor.
conditions and processes for generating wakes of inno- Organ. Sci. 14(3) 312–330.
vation. For example, mobile multimedia, biotechnology Bechky, B. A. 2006. Gaffers, gofers, and grips: Role-based coordina-
or advanced logistic systems can offer interesting fields tion in temporary organizations. Organ. Sci. 17(1) 3–21.
to examine. Bijker, W. E. 1995. Of Bicycle, Bakelites, and Bulbs: Toward a Theory
of Sociotechical Change. The MIT Press, Cambridge, MA.
Boland, R. J. 2003. An ecology of distributed mediated cognition.
Conclusion C. Ciborra, C. Angelou, eds. Information Systems and Social
We hope to stimulate a shift in thinking about diffu- Theory. Cambridge University Press, Cambridge, MA, 119–128.
sion of IT-induced innovation in networked contexts. Boland, R. J., R. V. Tenkasi. 1995. Perspective making and per-
We want to augment the individual and firm-centric spective taking in communities of knowing. Organ. Sci. 6(4)
perspectives on innovation that focus on the diffusion 350–372.
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
646 Organization Science 18(4), pp. 631–647, © 2007 INFORMS

Boland, R. J., R. V. Tenkasi, D. Te’eni. 1994. Designing information Levitt, B., J. G. March. 1988. Organizational learning. Annual Rev.
technology to support distributed cognition. Organ. Sci. 5(3) Sociol. 14 319–340.
456–475.
Lindsey, B. 2001. Digital Gehry: Material Resistance, Digital Con-
Boyatzis, R. 1998. Transforming Qualitative Information: Thematic struction. Birkhauser, Basel, Switzerland.
Analysis and Code Development. Sage, Thousand Oaks, CA.
Lyytinen, K., J. Damsgaard. 2001. What’s wrong with the diffusion
Brown, J. S., P. Duguid. 2000. The Social Life of Information. Harvard of innovation theory: The case of a complex and networked
Business School Press, Boston, MA. technology. M. A. Ardis, B. L. Marcolin, eds. Diffusing Software
Butler, B. S. 2001. Membership size, communication activity, and Product and Process Innovations. Kluwer Academic Publishers,
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

sustainability: The internal dynamics of networked social struc- Norwell, MA, 173–190.
tures. Inform. Systems Res. 12 346–362. Manning, P. 1979. Producing drama: Symbolic communication and
Carlile, P. R. 2002. A pragmatic view of knowledge and bound- police. Symbolic Interaction 5 223–241.
aries: Boundary objects in new product development. Organ. March, J. G. 1991. Exploration and exploitation in organizational
Sci. 13(4) 442–455. learning. Organ. Sci. 2(1) 71–87.
Churchman, C. W. 1968. The Systems Approach. Dell, New York.
Martin, P. Y., B. A. Turner. 1986. Grounded theory and organizational
Cooper, R. B., R. W. Zmud. 1990. Information technology imple- research. J. Appl. Behavioral Sci. 22(2) 141–157.
mentation research: A technological diffusion approach. Man-
Monge, P. R., N. S. Contractor. 2003. Theories of Communication
agement Sci. 36(2) 123–139.
Networks. Oxford University Press, New York.
Damanpour, F. 1991. Organizational innovation: A meta-analysis of
effects of determinants and moderators. Acad. Management J. Obstfeld, D. 2005. Social networks, the tertius iungens orienta-
23(3) 555–590. tion, and involvement in innovation. Admin. Sci. Quart. 50(1)
100–130.
DiMaggio, P., W. Powell. 1983. The iron cage revisted: Institutional
isomorphism and collective rationality in organizational fields. Orlikowski, W. J. 1993. CASE tools as organizational change: Investi-
Amer. Sociol. Rev. 48(2) 147–160. gating incremental radical changes in systems development. MIS
Quart. 17(3) 309–340.
Eisenhardt, K. M. 1989. Building theories from case study research.
Acad. Management Rev. 14(4) 532–550. Orlikowski, W. J., C. S. Iacono. 2001. Desperately seeking the “IT” in
IT research: A call to theorizing the IT artifact. Inform. Systems
Ferlie, E. 2005. Conclusion: From evidence to actionable knowledge?
Res. 12(2) 121–134.
S. Dopson, L. Fitzgerald, eds. Knowledge to Action? Evidence-
Based Health Care in Context. Oxford University Press, New Rogers, E. 1995. The Diffusion of Innovation, 4th ed. The Free Press,
York, 182–197. New York.
Fichman, R. G., C. F. Kemerer. 1997. The assimilation of soft- Shih, N.-J. 1996. A study of 2D and 3D oriented architectural drawing
ware process innovations: An organizational learning perspec- production methods. Automation Construction 5 273–283.
tive. Management Sci. 43(10) 1345–1363.
Star, S. L., J. R. Griesemer. 1989. Institutional ecology, “translations”
Galison, P. 1997. Image and Logic: A Material Culture of Micro- and boundary objects: Amateurs and professionals in Berkeley’s
physics. The University of Chicago Press, Chicago, IL. Museum of Vertebrate Zoology, 1907–1939. Soc. Stud. Sci. 19
Garud, R., P. Karnoe. 2001. Path creation as a process of mindful 387–420.
deviation. R. Garud, P. Karnoe, eds. Path Dependence and Cre- Stein, S. G., R. E. Alexander, C. L. Noble. 2001. AIA general condi-
ation. Lawrence Earlbaum Associates, New York, 1–38. tions in the digital age. Construction Contracts Law Rep. 25(25)
Gragg, R. 1999. Museum design tests Hoffman’s learning curve. The 3–20.
Oregonian. Portland, OR. Stinchcombe, A. 1959. Bureaucratic and craft administration of pro-
Greco, J. 2001. CATIA Version 5 Release 7. Cadence 16(12) 26–31. duction: A comparative study. Admin. Sci. Quart. 4 168–187.
Henderson, K. 1991. Flexibile sketches and inflexible data bases: Swanson, E. B. 1994. Information systems innovation among organi-
Visual communication, conscription devices, and boundary zations. Management Sci. 40(9) 1069–1092.
objects in design engineering. Sci., Tech. Human Values 16
Swanson, E. B., N. C. Ramiller. 1997. The organizing vision in infor-
448–473.
mation systems innovation. Organ. Sci. 8(5) 458–474.
Kauffman, R. J., M. James, Y.-M. Wang. 2000. Opening the “black
Swanson, E. B., N. C. Ramiller. 2004. Innovating mindfully with
box” of network externalities in network adoption. Inform. Sys-
information technology. MIS Quart. 28(4) 553–583.
tems Res. 11(1) 61–82.
Kellogg, K. C., W. J. Orlikowski, J. Yates. 2006. Life in the trading Teece, D. 1987. Profiting from technological innovation: Impli-
zone: Structuring coordination across boundaries in postbureau- cations for integration, collaboration, licensing, and public
cratic organizations. Organ. Sci. 17(1) 22–44. policy. D. J. Teece, ed. The Competitive Challenge. Ballinger,
Cambridge, MA, 285–305.
Kling, R., W. Scacchi. 1982. The web of computing: Computer tech-
nology as social organization. Adv. Comput. 21 1–90. Thompson, J. D. 1967. Organizations in Action. McGraw-Hill,
New York.
Koutamanis, A. 2000. Digital architectural construction. Automation
Construction 9 347–360. Toupin, L. A. 2001. Knowledgeware makes reusing existing designs
a snap. Design News 56(20) 42–53.
Lacourse, D. 2001. CATIA V5 r6 serves up robust tools for mechan-
ical engineers. CADalyst 18(12) 32–36. Tuomi, I. 2002. Networks of Innovation: Change and Meaning in the
Age of the Internet. Oxford University Press, New York.
Latour, B. 1987. Science in Action: How to Follow Scientists
and Engineers Through Society. Harvard University Press, Van de Ven, A. H. 2005. Running in packs to develop knowledge-
Cambridge, MA. intensive technologies. MIS Quart. 29(2) 368–378.
 Boland, Lyytinen, and Yoo: Wakes of Innovation in Project Networks
Organization Science 18(4), pp. 631–647, © 2007 INFORMS 647

Van de Ven, A. H., D. E. Polley, R. Garud, S. Venkatraman, eds. 1999. Yap, A. Y., O. Ngwenyama, K. M. Osei-Bryson. 2003. Leverag-
The Innovation Journey. Oxford University Press, New York. ing knowledge representation, usage, and interpretation to help
von Hippel, E. 1998. Economics of product development by user: reegineering the product development cycle: Visual computing
Impact of sticky local information. Management Sci. 44(5) and the tacit dimension of product development. Comput. Indust.
629–644. 51 89–110.

von Hippel, E. 2005. Open source software projects as user innovation Yin, R. K. 1994. Case Study Research: Design and Methods. Sage,
networks—no manufacturer required. J. Feller, B. Fitzegeral, Thousand Oaks, CA.
S. Hissam, K. Lakhani, eds. Perspectives on Free and Open Yoo, Y., R. J. Boland, K. Lyytinen. 2006. From organization design
Downloaded from informs.org by [202.40.157.7] on 20 August 2023, at 01:20 . For personal use only, all rights reserved.

Source Software. MIT Press, Cambridge, MA, 267–278. to organization designing. Organ. Sci. 17(2) 215–229.
von Hippel, E., G. von Krogh. 2003. Open source software and the Zaltman, G., R. Duncan, J. Holbeck. 1973. Innovations and Organi-
“private-collective” innovation model: Issues for organiation sci- zations. Wiley, New York.
ence. Organ. Sci. 14(2) 208–223. Zmud, R. W. 1984. An examination of “Push-Pull” theory applied to
Weick, K. E., K. M. Sutcliffe. 2006. Mindfulness and the quality of process innovation in knowledge work. Management Sci. 30(6)
organizational attention. Organ. Sci. 17(4) 514–524. 727–738.
Williams, R., D. Edge. 1996. The social shaping of technology. Res. Zuboff, S. 1988. In the Age of the Smart Machine: The Future of
Policy 25 865–899. Work and Power. Basic Books, New York.
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

Bibliometric study of the reverse salient concept

Ozgur Dedehayir

Tampere University of Technology (FINLAND)

[email protected]

Received September 2009

Accepted December 2009

Abstract: In the development of technological systems the focus of system analysis is

often on the sub-system that delivers insufficient performance – i.e. the reverse salient –
which subsequently limits the performance of the system in its entirety. The reverse salient
is therefore a useful concept in the study of technological systems and while the literature
holds numerous accounts of its use, it is not known how often, in which streams of
literature, and in what type of application the concept has been utilized by scholars since
its introduction by Thomas Hughes in 1983. In this paper we employ bibliometric citation
analysis between 1983 and 2008, inclusively, to study the impact of the reverse salient
concept in the literature at large as well as study the dissemination of the concept into
different fields of research. The study results show continuously growing number of
concept citations in the literature over time as well as increasing concept diffusion into
different research areas. The analysis of article contents additionally suggests the
opportunity for scholars to engage in deeper conceptual application. Finally, the
continuing increase in the number of citations highlights the importance of the reverse
salient concept to scholars and practitioners.

Keywords: technological system, reverse salient, bibliometric study, citation life cycle

1 Introduction

Studying technology through a systemic view offers scholars unique advantages

that are inaccessible in studying technology as a unitary phenomenon. The

Bibliometric study of the reverse salient concept 569

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

perspective of technology as a hierarchically structured entity that comprises

different levels of interconnected systems and sub-systems (Hughes, 1983;
Murmann & Frenken, 2006) enables scientists and practitioners to reveal the
interaction of sub-systems within the collective system. Consequently, the
researcher increases his/her understanding of the technology as well as the
technology’s change over time.

One profound characteristic that emerges from the systemic study of technologies
is the state of inequality that exists at any point in time between different
technological sub-systems, brought about by the complexity of the system as well
as its dynamic nature. Subsequently, the performance level of a particular sub-
system is inevitably lower than the levels reached by other sub-systems. This state
of performance unevenness carries significance in scientific analyses of
technological systems because the loci of interest are often the factors which, due
to their own limitations, constrain collective system development. Scientific
literature studying technological systems has therefore utilized different concepts
which describe this phenomenon, albeit with slight nuances; including bottleneck
and technological disequilibrium. One such concept, the reverse salient, is
introduced by Thomas Hughes in his 1983 seminal work, ‘Networks of Power:
Electrification in Western Society, 1880-1930’, referring to the under-performing
systemic member, which acts as a retardant and hampers the development of the
overall system performance. Due to conceptual difference with respect to other
concepts, the reverse salient has received attention from scholars who have
appropriated its use in the study of technological systems in certain context. While
the literature holds numerous accounts of its use, the importance of the concept to
the scientific community is unknown. Moreover, it is not known how often, in which
streams of literature, and in what type of application the concept has been utilized
by scholars since its introduction.

The purpose of this paper is to study the literature in order to gain greater insight
into the use of the reverse salient concept, first introduced by Thomas Hughes in
1983. For this purpose we firstly perform bibliometric citation analysis to observe
the temporal change in the number of citations of the reverse salient concept in the
literature at large. In turn, we assess the dissemination of the concept by first
analyzing the number of different journals that cite the concept, and second,
analyzing the number of different contexts of study that cite the concept, over
time. In our analyses we consider the life cycle of concept citations as an indicator
Bibliometric study of the reverse salient concept 570

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

of the concept’s importance to the literature and subsequently assess the

significance of the reverse salient concept in the study of technological systems.
Finally, we analyze the content of the articles citing the reverse salient concept and
determine the depth of conceptual application employed by the authors.

The structure of the paper is as follows. We begin by providing a theoretical as well

as an historical perspective of the concept together with accounts of its application
in various contexts and research streams. In the following section, the paper
formulates the employed methodology, comprising a bibliometric study that covers
97 journal articles collected from different electronic citation databases. The study
results offer the temporal change in the number of concept citations in the
literature at large, then more specifically in different journals and different contexts
of study, and finally with respect to depth of conceptual use, between the years
1983 and 2008, inclusively. These results are subsequently evaluated to identify
potential expansion of the scope of research in which the reverse salient concept
can be applied.

2 Theoretical background

2.1 The technological system

In contrast to studying technology as a unitary phenomenon, scholars have

recognized certain advantages in taking a systemic view of technology. As a
system, technology has been analyzed as a hierarchically structured entity that
comprises different levels of interconnected systems and sub-systems (see for
example: Henderson & Clark, 1990; Murmann & Frenken, 2006; Rosenberg, 1976;
Rosenkopf & Nerkar, 1999) and as an actor network that constitutes different
although interrelated components (see for example: Callon, 1987). Most
importantly, these perspectives advocate a meta-level view and analysis of
technology, through which the interaction of sub-systems or actors is brought
forth. The result is the increase in explanatory power of the scholar’s analysis of
technology and its evolution.

The adoption of a systemic perspective in the study of technological evolution, as

also demonstrated in biological, socio-cultural, and organizational studies of
evolution in earlier research, enhances understanding of the evolutionary
mechanisms which are at work both within hierarchical levels as well as across

Bibliometric study of the reverse salient concept 571

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

hierarchical levels (Rosenkopf & Nerkar, 1999). This means that individual sub-
systems do not merely evolve, but rather, they co-evolve interdependently with
other sub-systems. Furthermore, the systemic view recognizes evolutionary
trajectories of technology which are simultaneously observed at every level of the
system hierarchy subsequent to variation, selection, and retention processes
(Rosenkopf & Nerkar, 1999). As a result it is possible to identify the location of a
punctuated, discontinuous development at, say, a lower-order sub-system, which
may translate into incremental development at a higher-order level of the system
echelon. Therefore, the systemic view could go some way into opening the black
box of technology, thereby reducing the ambiguity in the nature of technological
change by considering the hierarchical structure and complexity of technological
systems.

On the one hand technological systems have been treated as a make-up of purely
technical sub-systems. In this manner, an aircraft can be considered as a system
that has some major technical sub-systems including the engine, which can be
decomposed into smaller technical sub-systems, such as the compressor,
combustion system, and turbine, which are subsequently divisible yet again into
smaller parts (Constant, 1987; Murmann & Frenken, 2006). Similarly, the study of
complex product systems (CoPS) (see for example: Hobday, 1998; Prencipe, 2000)
focuses on the technical components (high cost products, systems, and networks)
which constitute technological systems. A technical emphasis allows scholars to
study the system as an entire whole by analyzing the relationship and interaction
between its technical sub-systems, for example in studying industrial change
subsequent to innovations within the technological system (Henderson & Clark,
1990) or in studying the historical developments of technological systems
(Rosenberg, 1976). Hence, social factors such as the firms which produce the
technical sub-systems are seen as external to the system.

Alternatively, technological systems have been treated as socio-technical systems

that include social components in addition to the technical. The socio-technical
perspective allows the analysis of technological systems in the context of not only
the technical artifacts but also the end-users of these artifacts as well as the
makers of the artifacts. This view has been greatly advocated by the social
constructivist approach to understanding technology systems (Bijker, Hughes, &
Pinch, 1987), although it does not necessarily belong only to this area of
investigation. The socio-technical perspective is clearly advantageous when
Bibliometric study of the reverse salient concept 572

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

scholars wish to emphasize the impact of social groups on the technology system
and its change over time. One of the most influential works that illustrates the
socio-technical approach is Thomas Hughes’ Networks of Power: Electrification in
Western Society, 1880-1930. In his seminal work, Hughes (1983) traces the
development of the electric technology system from its birth until the early periods
of the twentieth century. He assumes a socio-technical standpoint in his book by
alluding at the very outset that throughout its progress, the electric system had not
only inflicted change upon its social environment but had also been altered by it.

2.2 The socio-technical system and its evolution

As part of his systemic framework, where change causality is sourced bilaterally

from technical and social factors, Hughes (1983) provides at least two mechanisms
for systemic evolution. The first mechanism is the interaction of the technological
system with its environment, which consists of factors not in the system’s control
and thus present uncertainties. This is exemplified by the dependence of the
electric system on an environmental factor such as the supply of fossil fuels.
Technological systems thus carry incentives to evolve by internalizing their
environments over time and so minimize potentially harmful ramification of
uncertainties. A second explanation of technological system evolution is the
system’s endeavor to solve problems or fulfill goals (Hughes, 1987).

Hughes proposes that technological systems pass through particular phases during
the system’s evolution. The first of these phases sees the invention and
development of the system, owed greatly to the efforts of inventors and
entrepreneurs. Considering the initial phase of the electric technological system,
for example, Hughes highlights the imperative role played by individuals such as
Thomas Edison. The second stage is the era of technological transfer from one
region or society to others. Hughes exemplifies such technological transfer through
the dissemination of Edison’s electric system from New York City to other cities
such as London and Berlin. The third phase of systemic evolution is marked by a
period of system growth. This is the time of expansion when the technological
system strives to improve its performance, for instance with respect to economic
outcomes or output efficiency. Hughes explicates that in this developmental phase,
technological systems are dependent on the satisfactory evolution of all system
components’ performance. Within the hierarchically nestled system, the
development of technological systems is therefore reliant on the reciprocated and

Bibliometric study of the reverse salient concept 573

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

interdependent cause and effect processes amongst social and technical

components. More accurately, the modus operandi of such developmental change
may be described as co-evolutionary, where the balanced co-evolution of system
components carries significance in establishing desired system progress (Hughes,
1983). In other words, components of the system which evolve at a sufficient pace
contribute positively to the collective development. Conversely, a sub-system
which does not develop sufficiently prevents the technology system achieving its
targeted development. Hughes (1983) names these problematic sub-systems
reverse salients.

A reverse salient is the inverse of a salient that depicts the forward protrusion
along an object’s profile or a line of battle (Hughes, 1987). Hence, reverse salients
are the backward projections along similar, continuous lines. With respect to
technological system development, reverse salients refer to the components of that
system which have strayed behind the advancing performance frontier of the
system (Hughes, 1983). They are therefore the underperforming components
which hamper the progress or which prevent the fulfillment of potential
development of the collective system. In line with the socio-technical standpoint,
reverse salients can be technical elements such as motors and capacitors of an
electric system, or social elements such as organizations or productive units.

Because reverse salients establish the limited pace of system development, it is

imperative that they are corrected, where correction is attained through
incremental or radical inventions (Hughes, 1987). Consequently, the reverse
salient forms a nexus or focusing device (Rosenberg, 1969) for technological
system stakeholders, in particular innovators and organizations, which congregate
around the retardant and strive to remove it through innovation processes. If the
reverse salient is not able to be corrected within the bounds of the existing
technological system through incremental innovations, then only radical
innovations can successfully resolve the problematic sub-system. The outcome of
radical innovations is the creation of a new and different technological system, as
witnessed through the innovation of the alternating-current system which
overcame the low cost distribution hurdle of the electric technological system,
where the direct-current system could not (Hughes, 1983).

Bibliometric study of the reverse salient concept 574

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

2.3 The reverse salient concept

The reverse salient is a useful concept for both scholars and practitioners in
analyzing technological system evolution. This is because quite often the loci of
interest in scientific analyses of technological systems are the factors which limit
system development. Furthermore, these limiting factors may be more than merely
technical components, and at times, quite importantly, may be social components
as well. In this way the conceptualization of system performance hindering factors
as reverse salients can be more applicable in certain contexts in contrast to similar
or overlapping concepts such as bottleneck and technological imbalance
(Rosenberg, 1976).

Although the reverse salient and bottleneck can be used interchangeably in

particular contexts (see for example: Fransman, 2001; Geels, 2006; Geyer &
Davies, 2000; Markard & Truffer, 2006) often the reverse salient refers to the sub-
system that not only curbs the performance or output of the collective system but
also requires correction because of its limiting affect. This is not necessarily the
case with bottlenecks. For instance, a particular system’s output performance may
be compromised due to a bottleneck sub-system but the bottleneck will not require
improvement if the overall system’s current output performance is satisfactory. If,
on the other hand, higher performance would be required of the same system, the
bottleneck may emerge as a reverse salient that holds the system back from
attaining that higher output performance.

A similar overlap exists between the reverse salient and technological imbalance
concepts (see for example: den Hond, 1998; Fransman, 2001; Takeishi & Lee,
2005). Rosenberg’s (1976) imbalance arises from the comparison of technological
components of complex machines or operations, which reveal that at any point in
time machine component parts will vary in their ability to exceed their existing
performance levels, attributable to some limiting component. This condition results
in the disequilibrium of technologies within the same system. Rosenberg illustrates
this disequilibrium effect through the example of a machining operation in the
manufacturing of bicycle hubs. In this example, despite performance improvements
which quicken the sub-process of hub forming, the overall pace of hub production
remains unchanged. Benefits from the forming process improvement are not
realized until the interdependent drilling sub-process quickens its speed of
production as well. A condition of disequilibrium therefore emerges between two

Bibliometric study of the reverse salient concept 575

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

processes, embedded within the same complex operation. Hence, the technological
imbalance described by Rosenberg can equally be seen as a state of reverse
salience in a technological system when studying only its technical components.
However, the reverse salient concept differs in that it can be used to study
imbalances not only between different technical components, but also social
components.

Historical developments in a variety of technological systems have been used in the

literature to illustrate reverse salience. In his seminal work, Hughes (1983) gives
account of the development of Edison’s direct-current electric system towards the
objective of supplying electricity within a defined region of distribution. Sub-
systems such as the direct-current generator were identified as reverse salients
and corrected. However, perhaps the most notable limitation of the direct-current
system was its low voltage transmission distance. The reverse salience in this case
was the cost of distributing electricity beyond a certain range. To reduce costs,
Edison introduced a three-wire system to replace the previously installed two-wire
alternative and trialed different configuration of generators, as well as the usage of
storage batteries. While these had a positive impact, they did not correct the
reverse salient completely. The satisfactory resolution of the problem of costly
transmission and distribution of low voltage electricity was eventually provided by
the radical innovation of the alternating-current system (Hughes, 1983).

Numerous authors have since offered their accounts based on the analysis of
different technological systems. MacKenzie (1987) has pinpointed the gyroscope
sub-system as a technical reverse salient in the ballistic missile technological
development, where the systemic objective has been to increase missile accuracy.
With the objective of proliferating mobile music throughout the end-user market,
Takeishi and Lee (2005) have argued that music copyright managing institutions
have acted as a social reverse salient in the evolution of the mobile music
technology system in Japan and Korea. According to Mulder and Knot (2001), the
development of the PVC (polyvinyl chloride) plastic technology system has been
sequentially hampered by several states of reverse salience, including: difficulty to
process PVC material, quality of manufactured products, health concerns for
individuals exposed to effluent from PVC manufacturing facilities, and finally the
carcinogenetic nature of vinyl chloride.

Bibliometric study of the reverse salient concept 576

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

3 Data and methods

3.1 The reverse salient concept in the literature

The reverse salient has emerged as a useful concept since its first introduction in
the literature in 1983. Notwithstanding, it is not known how often, in which
streams of literature, and in what type of application the concept has been utilized
by scholars since its introduction. To understand the value of the concept as seen
by scholars and at the same time predict future avenues of conceptual use it is
necessary to investigate the concept’s utilization in the literature thus far. This
understanding is important for both scholars and practitioners because their
analysis of technological systems requires the selection of conceptual tools, such as
the reverse salient, whereby the value afforded by previous scholars on the
conceptual tool establishes its potential selection among other concepts.

On the premise that the number of citations of a concept indicates the impact of
that concept on the literature, we derive several propositions as to the expected
adoption of the reverse salient concept in the literature from the time of its
inception. Here, we treat a concept as an intellectual idea or information which
experiences similar dissemination pattern within scientific fields as the adoption of
innovations (Ley, 2003). Using the number of citations as a proxy for concept
dissemination, we predict the emergence of an S-curve pattern (Cano & Lind,
1991; Lariviere, Archambault, & Gingras, 2007; Ley, 2003; Schwartz & Ibaraki,
2001; Schwartz & Fang, 2007) that is marked by rapid increase in the cumulative
number of citations early on, followed by a diminishing rate of increase as the
concept’s value in the eyes of scholars reduces and the concept nears
obsolescence. As the results of Cano and Lind’s (1991) citation life cycle analysis
for seminal works indicate, the temporal change in the number of citations may
follow either a sharp rise early after concept introduction (i.e. in the first four to
seven years) and gradually reduce thereafter, or increase gradually for some
period after initial introduction (i.e. for the first six years) and then begin a period
of take-off.

We firstly focus on the citations of the reverse salient in the literature at large.
After its introduction in 1983, we expect some time to elapse before scholars
become aware of the concept and its potential applicability. Cano and Lind (1991)
have for example identified a three year lapse from the time of a work’s

Bibliometric study of the reverse salient concept 577

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

introduction and its first citation. In time, however, the number of adopters is
expected to increase as the growing number of conceptual applications in the
literature provides greater visibility and illustrate the concept’s application potential
and importance. This leads us to derive the following proposition:

• Proposition 1: The number of citations of the reverse salient concept will

increase in time after its introduction.

The reverse salient concept is widely applicable in different contexts despite

originating in the study of the electric technological system analyzed by Hughes.
Hence, we anticipate that the concept is likely to disseminate into different
researched fields and appear in different publication outlets (in this article we focus
on journals as publication outlets). Moreover, the pattern of concept diffusion is
expected to follow an S-curve (see for example: Hargens, 2000), although the due
to the relatively recent introduction of the concept, it is uncertain whether
sufficient time has elapsed for the manifestation of the complete S-curve.
Nevertheless, we predict that the concept will initially receive interest from only
limited fields of research as awareness among the scientific community remains
limited. Over time however, we believe that the concept will receive attention from
a growing number of both researched fields (i.e. technological systems that have
been analyzed) and journals. This leads us to state the following propositions:

• Proposition 2: The number of journals citing the reverse salient concept will
increase in time after its introduction.

• Proposition 3: The number of research fields citing the reverse salient

concept will increase in time after its introduction.

In addition to quantitative indicators, the mode of application of the reverse salient

concept is also interesting to study. This qualitative aspect informs of the depth of
conceptual engagement the authors make as they analyze a particular
technological system. We estimate a change in the degree of conceptual
application to take place over time, as scholars become more familiar with the
concept and increase their knowledge. Therefore, soon after the concept’s
introduction in 1983, we expect awareness and knowledge to remain limited,
consequently resulting in shallow conceptual utilization. In other words, in this
early phase, scientific work that makes use of the reverse salient concept will
remain shy of comprehensive application of the concept in the analysis of

Bibliometric study of the reverse salient concept 578

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

technological systems. However, in time, we expect conceptual understanding to

increase, especially with growing visibility, leading to greater depth of conceptual
usage and in turn further development of the concept. This argument leads to the
following proposition:

• Proposition 4: The number of shallow conceptual applications will reduce,

while the number of deep conceptual applications will increase over time.

3.2 The data

The objective of this paper is to investigate the impact of the reverse salient
concept on the general body of literature – in other words its utilization by scholars
– since its introduction by Hughes in 1983. The research method employed here is
firstly a bibliographic citation analysis, which describes the impact of one research
on another (Brown & Gardner, 1985). Moreover, citations of a particular concept
indicate its innovativeness or usefulness as scholars consider them to be worthy of
reference and application in their own research. Citation analysis is also
advantageous because of its objective nature thus rendering it free of perceptions
of the researcher, contrary to other forms of similar research (e.g. opinion surveys)
(Chandy & Williams, 1994).

In this study we considered only journal articles as the outlet for scholarly work,
based on the argument that articles are more competitive than other modes of
communication in scholarly discipline (Chandy & Williams, 1994) and therefore
more comprehensive. We collected all journal articles citing the exact phrase
“reverse salient”, “reverse salients”, and “reverse salience”, between 1983 and
2008, inclusively. For this process we accessed journals that were listed on a
number of databases: JSTOR, ScienceDirect, EBSCOhost, ProQuest, Sage Premier,
ISI Web of Knowledge, IEEE Xplore, and SpringerLink.

Our initial search results included journal introductions, conclusions, editor

comments, commentaries, responses to critique, and book reviews. We decided to
filter these out of the scope of our analysis because our study focuses on the
utilization of the reverse salient concept in the analysis of a technological system,
and often the aforementioned works did not analyze a technological system. We
also excluded articles which either falsely referred to the concept (e.g. “adverse
salients” rather than “reverse salients”) or where the search term (e.g. “reverse
salient”) appeared in the reference list but not at all in the main text of the article.

Bibliometric study of the reverse salient concept 579

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

Furthermore, we left out of our study any articles which were currently in press, as
it remained equivocal as to when they would be published and hence, whether they
would fall within the timeframe of our study. Finally, although we looked for
citations from 1983 onwards, we could not access articles earlier than 1995 (nine
articles in all) in the ScienceDirect database and were therefore forced to neglect
these works.

All in all we acquired 97 journal articles to be included in our study and listed these
in chronological order. By adopting a diachronous approach, thereby counting the
number of citations made to a single concept in a longitudinal study (Cano & Lind,
1991), we analyzed the temporal change in the number of reverse salient citations
in the literature at large. It should be noted that in this longitudinal study, the data
has not been corrected for population growth, for instance with respect to active
researchers or the number of journals that may have increased over the years.

Next, to understand the adoption of the concept within different contexts, we

collected the articles firstly in groups of the technological systems that had been
studied (e.g. we collected all articles that had studied the aviation technological
system into one group), and secondly in groups of the journals where the article
had been published (e.g. we collected all articles that had appeared in the
Research Policy journal into one group). We then analyzed the change in the
number of different technological systems that had applied the reverse salient
concept over time, and the change in the number of different journals that had
cited the concept over time, respectively.

Finally, we analyzed the content of the articles (Kassarjian, 1977) in order to

derive qualitative understanding of the reverse salient concept’s utilization in the
literature. Here, the collected articles were analyzed in the depth of conceptual
application employed by the authors in their studies. We grouped articles into the
shallow category if the articles had not integrated the reverse salient concept in a
comprehensive manner. Typically, the articles which were allocated to this category
had merely referred to the reverse salient concept in theory building or in the
theoretical discussion of the article. Conceptual articles, which had subsequently
not studied a technological system, were also allocated to the shallow category
(while it can rightfully be argued that conceptual papers have significant depth of
analysis, in this study we focused on the depth of conceptual use in the analysis of
technological systems only, following the seminal work Hughes). On the contrary, if

Bibliometric study of the reverse salient concept 580

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

the authors had analyzed a specific technological system using the reverse salient
concept we felt that they had engaged in deeper conceptual application. We
classified these latter articles within the deep category. Typically, the articles which
were judged to belong to this category had, as a minimum requirement, identified
a reverse salient in the analyzed context (i.e. the technological system). Some of
these works had further extended the discussion by considering the consequence
and resolution of reverse salients. While such content analysis can be subjective,
we tried to reduce the degree of subjectivity in our analysis by minimizing the
number of categories into which the articles could be allocated with respect to the
depth of conceptual application (only two categories were used). After this
classification, we analyzed the change in the number of articles belonging to the
shallow and deep categories over time, respectively.

4 Results

The temporal change in the number of citations of the reverse salient concept in
the literature at large is shown in Figure 1. This graph indicates a rising number of
citations per year from 1990 (the first journal article citing the concept) onwards.
Hence, the cumulative number of citations shows an almost exponential growth for
the timeframe of analysis, commensurate with our first proposition.

Figure 1. “Citations of the reverse salient concept in the literature at large”.

The growing number of citations in the literature suggests that the reverse salient
concept continues to hold relevance to the works of modern day scholars such that

Bibliometric study of the reverse salient concept 581

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

they feel it beneficial to use it in their work. Subsequently, the concept has been
cited in a number of different periodicals. Figure 2 displays the journals that have
cited the reverse salient two times or more.

Figure 2. “Most frequently citing journals”.

Research Policy accounts for 14 out of the 97 total citations, significantly more than
any other journal. The journals Science, Technology & Human Values, and
Technology Analysis & Strategic Management have both published six articles each
and rate second in the overall frequency listing. It is additionally interesting to note
that nine out of the total of 59 different journals (i.e. 15%) which have cited the
reverse salient concept over time have been policy oriented journals. This finding
may indicate the utility of the concept in policy formulation or the call for policy
changes, especially with respect to new technological systems that face
performance limiting factors. Further, the change in the number of different
journals which cite the concept over time is shown in Figure 3.

Since the first journal that cited the reverse salient concept in 1990, the number of
journals which have published the concept has grown steadily over the timeframe
of analysis. This finding confirms our second proposal and therefore indicates the
broadening of the fields of research that have assigned value to the concept.

A second indicator of conceptual dissemination is found in the contexts, in other

words the technological systems, which have been analyzed by scholars. Figure 4

Bibliometric study of the reverse salient concept 582

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

shows the frequency of citation of the reverse salient concept in the study of
different research fields.

Figure 3. “Citations of the reverse salient concept in journals”.

Figure 4. “Most frequently citing research fields (studied technological systems)”.

Figure 4 illustrates that the most frequently citing articles are conceptual and
theoretical (12 in all) in their nature and have not analyzed a particular
technological system. The technological system which has been studied most
frequently with respect to the reverse salient is the railroads and railways (four
citations), followed by a number of other systems (three citations) including
aviation, and information and communication technological systems. Interestingly,

Bibliometric study of the reverse salient concept 583

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

the research fields listed in Figure 4 are quite often either historical systems (e.g.
telegraphy, railroads and railways) which have necessitated retrospective analysis
of reverse salients, or new systems (e.g. photo-voltaic energy, electric vehicle)
which require prospective analysis. The evolution of the number of research fields
citing the reverse salient concept over time is shown in Figure 5.

Figure 5. “Citations of the reverse salient concept in research fields”.

Once again there appears to be a gradual increase in the total number of different
research contexts which have incorporated the concept over time. This result is in
line with our third proposal and again confirms our expectation of the pattern of
concept dissemination in scientific research.

Finally, we compare the depth of conceptual application in citing articles over time
in Figure 6.

Our study revealed 46 articles that had applied deep conceptual analysis in
comparison to 51 articles which had employed shallow analysis. Figure 6 further
indicates that the number of citations in both deep and shallow analysis has
remained fairly similar across time. Nevertheless, there appears to be a persisting
lag in the early period until 2001, where the deep analysis articles trail the shallow
analysis articles in number of citations. This outcome is consistent with our fourth
proposal which stipulates a period of ferment early after the concept’s inception,
due to limited knowledge of the concept and its potential. However, and as also
indicated by Figure 6, the deep analysis citations show increase at a later period,

Bibliometric study of the reverse salient concept 584

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

albeit short lived, overtaking the shallow citations in number. This result yet again
supports our fourth proposal, although the strength of this support is arguable,
considering in particular the most recent period where shallow analysis citations
once again outnumber the deep analysis citations.

Figure 6. “Depth of conceptual application in citing articles”.

5 Discussion and conclusions

The life span of a particular scientific work is an indication of the significance of

that work in the literature. Seminal works or concepts are likely to receive high
number of citations from the scientific community for many years, while works
carrying less value are likely to experience short-lived and limited adoption after
their inception. Cano and Lind (1991) have observed that seminal works can
display pronounced increase in citations within the first four to seven years after
being introduced, followed by a period of decrease in citations. While their study
has focused on the medicine and biochemistry disciplines, their findings
nevertheless offer some guideline as to the expected life of important scholarly
contributions in other disciplines as well. Thus, using Cano and Lind’s work as a
frame of reference, we find that Hughes’ reverse salient concept has continued to
be perceived as valuable and has subsequently received a growing number of
citations in the literature from its inception in 1983 until and including 2008.
Furthermore, the dissemination pattern of the concept in periodicals as well as the
number of studied contexts indicates similar increase. These results reaffirm once
again the reverse salient concept’s perceived value to scholars. These findings have
Bibliometric study of the reverse salient concept 585

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

important implications for the scientific community as well as the community of

practitioners who select and employ conceptual tools in their analysis of
technological systems, whereby the apparent scientific value of the concept
increases its scope for adoption in future studies in academia and practice.

The reverse salient concept has been adopted and cited by scholars within a wide
scope of research fields. For example, despite its initial application in 1983 within
the context of the electric technological system, the reverse salient has been used
in fields as far away as agricultural science, banking, urban systems, musical
instrument technology, and government or public policy. The socio-technical
characteristic of the concept affords flexibility, which has clearly been manipulated
by scholars within literature at large. Not surprisingly, the concept has also
continued to be used within a close radius of its origin, exemplified in particular by
a number of articles covering the field of production and energy systems.

There are other interesting observations which can be made with respect to the
research areas where the concept has been highly cited as well as areas where it
has not been cited at all. Firstly, the reverse salient concept appears to be utilized
in the analysis of alternative energy technological systems, which is presently a
topic of high interest. A good example is Christiansen and Buen’s (2002) article,
which provides a comprehensive application of the reverse salient concept in
assessing the case of photovoltaic and wave power development in Norway. We
forecast an increase in the number of studies which apply the reverse salient
concept in this and similar research fields, where the development of technological
systems is often limited by both technical and social components of the system. In
this case, the reverse salient provides a very fitting conceptual tool in the analysis
of these systems.

Another general research discipline which cites the concept frequently is that of
telecommunications, both in the modern and historical contexts. The articles
dedicated to modern telecommunication cover topics such as wireless services and
information and communication technology, which due to complex socio-technical
systemic structures are predicted to continue publishing works citing the concept.
It is also not surprising to observe the association of the reverse salient concept
with the retrospective analyses of the telecommunication technological system. The
telephone and telegraph technological systems have been analyzed in historical

Bibliometric study of the reverse salient concept 586

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

context, following Hughes’ initial analysis of the closely related electric

technological system’s development.

Notwithstanding, there are relatively unexplored fields of research in which the

reverse salient concept has potential to be incorporated – corresponding to
research areas which have either cited the concept sparingly or not at all. One such
arena is strategic management. The present study shows limited citation of the
reverse salient concept inside strategic management circles. For instance, the
Technology Analysis & Strategic Management journal cites the reverse salient
concept in only six out of the 97 articles considered in this study. Hence, we
believe that the study of technological systems within the strategic management
literature has potential in incorporating the reverse salient concept and has
therefore not yet been exploited. Moreover, such application will require the
increasing number of deep conceptual use, which currently accounts for
approximately half of the cases studied.

Although some noteworthy conclusions have been drawn from the study presented
here, some of the limitations of the study need to be brought forward to the
reader’s attention. Firstly, the citation list constructed from the research activity
account only for the articles which appear in journals. Consequently, citations in
books, monographs, working papers, and conference proceedings are not taken
into account. This circumstance may have led to the exclusion of some sources
where the reverse salient has been cited and therefore pinpoints one avenue of
extending the present study. Notwithstanding, focusing on research within journals
holds some merit in the sense that these publications are arguably more
comprehensive than other modes of communication in scholarly discipline (Chandy
& Williams, 1994). Journal articles therefore may provide a more suitable source of
literature for content analysis as utilized in this paper, where focus is on
researching the idiosyncrasies of a concept’s use in addition to its citation alone.
Our employed method of citation analysis has additionally received critique as an
imperfect measure of research impact. This is because although some research
may have impact or prove to be excellent scientific work it can still remain
sparingly cited. Conversely, some research may receive high number of citations
despite having less impact and overall value. Hence, while citation analysis remains
widely used in bibliometric studies, its true indication of the impact of a seminal
work or concept in the literature needs to be assessed with care.

Bibliometric study of the reverse salient concept 587

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

Another limitation of our study is that the initial pool of journals researched in this
study has been constrained to those which have been within access. The inclusion
of additional article databases in conducting this research, such as Google Scholar
(see for example: Clarke, 2008), can enlarge the array of articles that are
garnered. This however, remains a limitation of the present paper which can be
address in future work. Additionally, future studies can assess the life cycle of the
reverse salient concept by fitting curves (e.g. S-curve) on the attained cumulative
citation graphs, which can illuminate the future citation behavior of the concept
(Schwartz & Fang, 2007). Other supplementing research, using citation and
content analyses of related concepts such as technological imbalance (Rosenberg,
1976) and bottleneck can help ascertain how academic literature perceives the
reverse salient concept on a relative scale.

In summary, it is proposed that literature holds two sets of potential concerning

the use of the reverse salient concept which is found to be seminal in technology
system analysis. Firstly, there lies an opportunity for a higher ratio of scholars to
derive deeper conceptual utility in their work. Secondly, and related to the first, the
application of the reverse salient concept remains poised to incorporate a more
strategic approach to technological system analysis undertaken thus far in
literature. We believe that the realizations of these potentials will benefit the
literature concerned with the study of technological system in general and at the
same time further develop the reverse salient concept.

References

Bijker, W. E., Hughes, T. P., & Pinch, T. J. (Eds.). (1987). The social construction of
technological systems. USA: The MIT Press.

Brown, L. D., & Gardner, J. C. (1985). Applying citation analysis to evaluate the
research contributions of accounting faculty and doctoral programs. The
Accounting Review, 60(2), 262-267.

Callon, M. (1987). Society in the making: The study of technology as a tool for
sociological analysis. In W. E. Bijker, T. P. Hughes & T. P. Pinch (Eds.), The social
construction of technological systems (pp. 83-106). USA: The MIT Press.

Cano, V., & Lind, N. C. (1991). Citation life cycles of ten citation classics.
Scientometrics, 22(2), 297-312.

Bibliometric study of the reverse salient concept 588

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

Chandy, P. R., & Williams, T. G. E. (1994). The impact of journals and authors on
international business research: A citational analysis JIBS articles. Journal of
International Business Studies, 25(4), 715-728.

Christiansen, C. A., & Buen, J. (2002). Managing environmental innovation in the

energy sector: The case of photovoltaic and wave power development in Norway.
International Journal of Innovation Management, 6(3), 233-256.

Clarke, R. (2008). An exploratory study of information systems researcher impact.

Communications of the Association for Information Systems, 22, 1-32.

Constant, E. W. (1987). The social locus of technological practice: Community,

system, or organization? In W. E. Bijker, T. P. Hughes & T. P. Pinch (Eds.), The
social construction of technological systems (pp. 233-242). USA: The MIT Press.

den Hond, F. (1998). The 'similarity' and 'heterogeneity' theses in studying

innovation: Evidence from the end-of-life vehicle case. Technology Analysis and
Strategic Management, 10(4), 529-543.

Fransman, M. (2001). Analysing the evolution of industry. Economics of Innovation

and New Technology, 10(2), 109-141.

Geels, F. W. (2006). Co-evolutionary and multi-level dynamics in transition: The

transformation of aviation systems and the shift from propeller to turbojet (1930-
1970). Technovation, 26(9), 999-1016.

Geyer, A., & Davies, A. (2000). Managing project-system interfaces: Case studies
of railway projects in restructured UK and German markets. Research Policy,
29(7), 991-1013.

Hargens, L. L. (2000). Using the literature: Reference networks, reference contexts,

and the social structure of scholarship. American Sociological Review, 65(6), 846-
865.

Henderson, R. M., & Clark, K. B. (1990). Architectural innovation: The

reconfiguration of existing product technologies and the failure of established
firms. Administrative Science Quarterly, 35(1), 9-30.

Bibliometric study of the reverse salient concept 589

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

Hobday, M. (1998). Product complexity, innovation and industrial organisation.

Research Policy, 26, 689-710.

Hughes, T. P. (1983). Networks of power: Electrification in western society, 1880-

1930. USA: The John Hopkins University Press.

Hughes, T. P. (1987). The evolution of large technological systems. In W. E. Bijker,

T. P. Hughes & T. P. Pinch (Eds.), The social construction of technological systems
(pp. 51-82). USA: The MIT Press.

Kassarjian, H. H. (1977). Content analysis in consumer research. The Journal of

Consumer Research, 4(1), 8-18.

Lariviere, V., Archambault, E., & Gingras, Y. (2007). Long-term variations in the
aging of scientific literature: From exponential growth to steady-state science
(1900-2004). Journal of the American Society for Information Science and
Technology, January, 288-296.

Ley, D. (2003). Forgetting postmodernism? recuperating a social history of local

knowledge. Progress in Human Geography, 27(5), 537-560.

MacKenzie, D. (1987). Missile accuracy: A case study in the social processes of

technological change. In W. E. Bijker, T. P. Hughes & T. J. Pinch (Eds.), The social
construction of technological systems (pp. 195-222). USA: The MIT Press.

Markard, J., & Truffer, B. (2006). Innovation processes in large technical systems:
Market liberalization as a driver for radical change? Reseach Policy, 35(5), 609-
625.

Mulder, K., & Knot, M. (2001). PVC plastic: A history of systems development and
entrenchment. Technology in Society, 23, 265-286.

Murmann, J. P., & Frenken, K. (2006). Toward a systematic framework for research
on dominant designs, technological innovations, and industrial change. Research
Policy, 35, 925-952.

Prencipe, A. (2000). Breadth and depth of technological capabilities in CoPS: The

case of the aircraft engine control system. Research Policy, 29, 895-911.

Bibliometric study of the reverse salient concept 590

O. Dedehayir
doi:10.3926/jiem.2009.v2n3.p569-591 ©© JIEM, 2009 – 2(3): 569-591 - ISSN: 2013-0953

Rosenberg, N. (1969). The direction of technological change: Inducement

mechanisms and focusing devices. Economic Development and Cultural Change,
18, 1-24.

Rosenberg, N. (1976). Perspectives on technology. Cambridge: Cambridge

University Press.

Rosenkopf, L., & Nerkar, A. (1999). On the complexity of technological evolution:

Exploring coevolution within and across hierarchical levels in optical disc
technology. In J. A. C. Baum, & B. McKelvey (Eds.), Variations in organization
science (). USA: Sage Publications, Inc.

Schwartz, F. W., & Fang, Y. C. (2007). Citation data analysis on hydrogeology.

Journal of the American Society for Information Science and Technology, 58(4),
518-525.

Schwartz, F. W., & Ibaraki, M. (2001). Hydrogeological research: Beginning of the

end or end of the beginning? Ground Water, 39(4), 492-498.

Takeishi, A., & Lee, K. (2005). Mobile music business in Japan and Korea:
Copyright management institutions as a reverse salient. Journal of Strategic
Information Systems, 14, 291-306.

©© Journal of Industrial Engineering and Management, 2009 (www.jiem.org)

Article's contents are provided on a Attribution-Non Commercial 3.0 Creative commons license. Readers are
allowed to copy, distribute and communicate article's contents, provided the author's and Journal of Industrial
Engineering and Management's names are included. It must not be used for commercial purposes. To see the complete
license contents, please visit http://creativecommons.org/licenses/by-nc/3.0/.

Bibliometric study of the reverse salient concept 591

O. Dedehayir
 Long Range Planning 50 (2017) 310e321

Contents lists available at ScienceDirect

Long Range Planning

journal homepage: http://www.elsevier.com/locate/lrp

The Dynamics of Open Strategy: From Adoption to Reversion

Melissa M. Appleyard, Henry W. Chesbrough

Innovation has become more open in recent years. Yet the decision to become more open and the challenge of sustaining that openness
are not well understood. This is the concern of the “content” branch of Open Strategy, defined as the branch that addresses an organi-
zation’s open innovation strategy. We examine the initial motivations to adopt an open strategy, and then consider when organizations
choose to maintain that open strategy or revert to a more proprietary approach. Similarly, we examine motivations to open up a pre-
viously proprietary strategy. We find that these dynamics depend on the organization’s desire to either foster greater growth (which
favors a more open strategy) or secure greater control and profit directly from the innovation (which favors a more proprietary strategy).
Crucially, these choices can shift over the life cycle of a market and are dependent on the competencies amassed by the organization. In
early phases, when there are relatively few legacy customers and many new arrivals, open strategies attract customers at a faster rate. In
later phases, as the market matures and new arrivals have slowed, there are few new customers to attract with an open strategy and
reversion to a more proprietary strategy becomes quite attractive. This suggests that the longevity of open initiatives might be curtailed as
organizations opt for value capture over cooperative value creation.
© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

Introduction

If we are to make strategic sense of innovation communities, ecosystems, networks, and their implications for
competitive advantage, we propose that a new approach to strategy d open strategy d is needed.
d Chesbrough and Appleyard, 2007: 58
Since the publication of Chesbrough and Appleyard (2007) nearly a decade ago, there has been significant interest in Open
Strategy but limited attention has been paid to the dynamics that characterize a firm’s pursuit of this new approach to
strategy. Two branches of Open Strategy have emerged: a “content” branch that examines the ability of organizations to
sustain themselves economically with an open approach to innovation (Chesbrough and Appleyard, 2007); and a “process”
branch that explores the systems that can enhance strategy formulation by furthering participation of both internal and
external actors and improving transparency inside and outside of the firm (Whittington et al., 2011). While other papers in
this special issue address this latter branch, this research provides a framework for examining the content branch of Open
Strategy from adoption to d under certain circumstances d its abandonment. In our context, we define abandonment as
reversion to a proprietary strategy. We have witnessed these dynamics in industries ranging from life sciences to consumer
packaged goods, and in this paper, we particularly focus on examples from the information and communications technology
(ICT) and software industries, such as the evolution of the Android operating system. The analysis draws on existing literature
pertaining to the role of knowledge accumulation, network effects, and ecosystem development in fueling cross-firm
cooperation when a new market is emerging, complemented by interviews from leading firms that contribute extensively
to open innovation initiatives including IBM and Intel. We find that the decision to either open up a previously closed project
(i.e., abandon property rights protection) or close off a previously open project (i.e., revert to a proprietary strategy) hinges on
the evolving realities of the market including rate of product adoption and the emergence of a supporting ecosystem, as well
as changes in the competencies accumulated by the cooperating firms.
Prior research has observed that open innovation is leading to new empirical phenomena that do not fit well with Por-
terian theories of business strategy (Chesbrough and Appleyard, 2007; Lerner and Tirole, 2002; West and Gallagher, 2006).
The primary tension between Open Strategy and traditional business-level strategy rests with the need to secure an economic
return in the face of relinquishing control over critical strategic assets and capabilities (Barney, 1991; Dwoskin, 2016; Peteraf,
1993; Wernerfelt, 1984). The reconciliation of this tension is how we define Open Strategy: a firm’s justification for

http://dx.doi.org/10.1016/j.lrp.2016.07.004
0024-6301/© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).
 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321 311

participating in an open initiative, including its ability to capture value from the initiative. In our application of Open Strategy,
an open initiative is characterized by: a) the reliance on assets outside of the firm’s boundaries (inclusion), and b) the (free)
access to project results by outsiders (transparency). By way of example, in the setting of an open source software (OSS)
project, contributions to the code base might span hundreds of software developers who are freelancers, affiliated with non-
profit organizations, or employed by companies. The compilation of their individual contributions leads to a new software
product, e.g., the well-known example of Linux, which is published in the public domain. This pooling of assets externally
leads to a high-quality, freely accessible product, and because of this open access, the participating firms then have to
construct business models to capture value elsewhere in the value chain (Chesbrough and Appleyard, 2007; Perr et al., 2010).
The participating firms also need to decide their level of commitment to openness over the life of the project, and the drivers
behind this decision are the focus of this paper.
The remainder of the paper is organized as follows. In the next section, we introduce a conceptual framework for
considering the range of strategic possibilities between open and closed innovation initiatives and provide examples. We then
analyze the foundations of value creation during open innovation through a producer and consumer surplus lens. The
subsequent section details the basis for collaboration and the factors that might lead to a shift in strategic choice, including
the possibility of reversion. The following section goes more deeply into internal competencies that traditional firms have
grappled with to effectively support open initiatives. Included is a consideration of how successful cultivation of compe-
tencies, the technology prowess competency in particular, might actually hasten reversion back to a closed strategy. The
triggers of reversion have not received extensive treatment in the literature, and they help to inform the limits of open
innovation. In the final sections, we discuss the findings and conclude with possible paths for future research.

A dynamic view of open strategy

The primary research contributions of this paper are: 1) to demonstrate how a firm’s commitment to Open Strategy may
be far from static; and 2) to identify the underlying forces driving the migration of strategic choice. In this section, we
introduce an intertemporal framework for understanding the range of strategic choices with illustrative examples. Typically,
past researchers have emphasized the strategic decision to own and control intellectual property in order to earn a revenue
stream to pay for research and development (Grossman and Helpman, 1994; Nordhaus, 1969; Romer, 1990). Or they have
emphasized a perpetual commitment to open, user-led innovation (Baldwin and von Hippel, 2011; Shah and Tripsas, 2007;
Swann, 2014; von Hippel, 2005), where one’s ability to benefit directly from the use of an innovation obviates any need to
appropriate economic value from that innovation. However, these otherwise highly contrasting approaches share a common
characteristic: once the strategic choice is made to be open or to be proprietary, that decision is implicitly treated as fixed
thereafter. Recent examples reveal that this is an incomplete reflection of reality, where some projects that began as pro-
prietary have moved to become open, while other projects that began as open have subsequently become proprietary. The
choice of openness thus is not static, but dynamic. An examination of firms that have moved along the OpeneClosed strategy
continuum can help illustrate the dynamics. Figure 1 shows how companies and open projects might start with an Open or

Cisco: OpenDaylight Linux

Open IBM: PC Apache
Sun: Java Drupal
Mozilla
Ex Post

IBM: Mainframe Google: Apps

Closed Microsoft: Windows Google: Android OS
Oracle Amazon: Kindle Fire

Closed Open

Ex Ante

Figure 1. Dynamics behind the OpeneClosed Strategic Choice

312 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321

Closed strategy (“Ex Ante”) and remain in that initial state over time (“Ex Post”). These two strategic choices constitute the
main diagonal of Figure 1.
The literature provides extensive treatment of the ClosedeClosed strategy (Porter, 1980), where the organizations pur-
suing this strategy wall off their innovation initiatives from external contributors, both at the outset and over time. Following
this strategy, companies like IBM, Microsoft, and Oracle have enjoyed lucrative profit streams from their proprietary products
and services.
Similarly, the literature is replete with examples of the phenomenon of OpeneOpen initiatives (Allen, 1983; Baldwin and
von Hippel, 2011; von Hippel, 2005; von Hippel and von Krogh, 2003). Three enduring open source software (OSS) projects
represent the OpeneOpen approach in Figure 1: Linux, Apache and Drupal. They also constitute a breed of innovation where
(advanced) users are themselves the innovators (Franke and Shah, 2003; Lakhani and von Hippel, 2003; Lerner and Tirole,
2002; O’Mahony and Bechky, 2008; Raymond, 2001). They have managed to sustain their commitment to openness over
time, something we worried greatly about in previous work (Chesbrough and Appleyard, 2007).
A novel contribution of this paper is to focus on the “off-diagonals” of Figure 1, for the main diagonal in the figure does not
represent all of the possible options firms face. In the off-diagonal quadrants, projects can migrate from Closed to Open or
from Open to Closed. As an example of the former, in 2013, Cisco coordinated the launch of the OpenDaylight initiative. This
open source software project is in effect taking some of the intelligence out of Cisco’s proprietary network hardware and
putting it into an openly available software layer. IBM’s PC business similarly started out as initially closed, but rapidly
developed into a de facto open standard, driven in large part by the actions of Intel and Microsoft supporting “IBM compatible”
PC systems (Chesbrough and Teece, 1996). The history of Java also followed this pattern, where the shepherds of Java at Sun
switched Java’s license to a more open one in 2006 (Fisher, 2006). Mozilla, the free web browser, resulted from the aftermath
of Netscape’s battle with Microsoft in the mid-1990s in the so-called Browser Wars.
The reverse flow also arises. Two examples of the Open to Closed migration are Amazon’s build-out of the Kindle Fire
ecosystem and Google’s management of the Android operating system and related apps. To launch its Kindle tablet, Amazon
adapted, or “forked,” a version of Android to create its Fire operating system and developed its own application programming
interfaces (APIs) that would dictate how applications would connect to the operating system (Pon et al., 2014). This allowed
Amazon to control the specifications of the apps that the Kindle and other Amazon devices could use. This control ensured
that Amazon would be able to integrate its range of services into apps sold in its Appstore, allowing customers to access not
only the digital content that they purchased through Amazon but the expanding range of services like cloud storage and
streaming media (Pon et al., 2014). Because of the very small marketshare of Amazon Fire devices relative to the whole
Android universe, only one-tenth the number of apps have been developed for Fire relative to Google’s app store (Pon et al.,
2014).1
While Amazon pursued a closed Fire ecosystem after adapting an open operating system, the closing off of Android under
Google’s leadership has been more oblique. After Google acquired Android in 2005, it opened up the source code to draw in a
broad developer base to propel adoption (Open Handset Alliance, 2007), consistent with the positive feedback effects of
Shapiro and Varian (1999). Initially, the open approach succeeded in attracting a large and vibrant ecosystem to Android to
catch up to Apple’s iOS operating system, as it was intended to do. That is not the end of the Android story, however. Over
time, as Android has surpassed Apple’s operating system in unit market share in the smartphone arena (Business Wire, 2014),
Google has progressively closed off Android through a variety of means. Google has inserted two layers of proprietary control
(Amadeo, 2013). The first is Google’s migration of open source-derived apps to closed source versions provided and main-
tained exclusively by Google, in the form of Google apps, such as the Calendar and Camera. Google will continue its now-
proprietary apps into the future, without any further sharing back to the initial open source versions.
The second way Google is closing off the Android platform is by extracting functionality out of the operating system and
placing it in a new proprietary “app” called Play Services (Amadeo, 2013; Pon et al., 2014). This is a classic envelopment
strategy, in the language of Eisenmann et al. (2011). Play Services provides systems-level capabilities like account syncing and
Googleþ connectivity required for the smooth functioning of Google’s now-proprietary apps found in the Play Store. This
means that Google can update Play Services frequently without having to rely on infrequent, major Android operating system
updates. It also means that companies like Samsung, who has tried to mimic the functionality of Google’s proprietary apps via
its own suite of apps, may have difficulty “inventing around” this new systems-level app, thereby becoming even more
beholden to Google. It further means that individual volunteer contributors have no future role to play in this integration of
services.
While Android is indeed licensed through the Apache Software License (Android, 2014), which is an open source license,
Google has in effect closed off Android without much public notice or outcry. While these methods are most likely perfectly
legal, in our view they violate the spirit of open source development, turning Android into “more of a ‘look but don’t touch’
kind of open” (Amadeo, 2013). They also qualify some of the more extravagant claims of open source proponents of
“democratizing innovation” (e.g., von Hippel, 2005). The reversion to a proprietary strategy suggests that the initial open

1
As Pon et al. (2014) analyze, their very different strategies around welcoming apps into their respective stores reflects the different business models
pursed by Amazon (retail sales) and Google (advertising). What remains to be seen is whether Google’s strategy of shifting functionality out of Android and
into Play Services as described below will hobble Amazon in the long-run when Amazon wants to upgrade Fire, or whether “OS-agnostic” apps like those
based on HTLM5 might lessen the importance of the operating system.
 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321 313

approach was a strategy to gain rapid adoption, subject to later revision if circumstances warranted it. Open need not imply a
commitment to remain open in perpetuity.

The underlying value created by open and closed strategies

The question becomes, why does such migration away from the main diagonal pattern occur? Why do initially Closed
strategies shift to Open, and why do initially Open strategies shift to Closed? This section presents the underlying gains that
accrue to producer surplus or consumer surplus…or both from the Open or Closed strategic choice. As these underlying gains
shift, so too can the optimal choice of strategy.
In the ClosedeClosed cell of Figure 1, firms enjoy the (temporary) monopoly profits afforded by intellectual property
rights, along with the high barriers to competition that drive the Five Forces analysis of Michael Porter (Porter, 1980). The
producer surplus is increased by having control over price and quantity. In the OpeneOpen cell, the user-innovators
capture a share of the returns from both increased producer and consumer surplus because they are not only the crea-
tors but also typically the consumers of the output (Baldwin and von Hippel, 2011; Baldwin et al., 2006; Swann, 2014). They
can also benefit from providing complementary goods and services which need not be open, a topic we discussed
extensively in our earlier paper (Chesbrough and Appleyard, 2007). In the Close-Open and Open-Closed cells, producer and
consumer surplus evolve with the growth of the market and drive the inter-temporal migration of strategy in ways we
discuss below.

Producer surplus in open initiatives

Specifically considering open initiatives, the contributions to producer surplus can be characterized in terms of cost,
quality, speed, and coordination. The momentum surrounding an open project can draw in vast numbers of contributors as
seen in many of the successful open source software projects.2 This can drive down costs of production while enhancing
quality because of the sheer volume of contributors and their diversity. The often reduced and shared cost of development can
lead to a greater number of experiments running in parallel, which can shorten development cycles (Baldwin and Clark,
2000), and because transparency is a hallmark of open projects (Whittington et al., 2011), results can be disseminated in a
more timely and broader fashion. As Google has demonstrated through the initial open build-out of the Android ecosystem, a
further benefit of openness is, in quick order, to coordinate a critical mass of players to launch a new platform. This allowed
Google to overcome a late start even in a setting with strong network effects at a time when first Microsoft’s and then Apple’s
ecosystem had threatened to dominate the smartphone market (Pon et al., 2014; West and Gallagher, 2006).3

Consumer surplus in open initiatives

From the perspective of consumer surplus, openness generates value for consumers through price, features, and self-
determination. Price generally is lower than proprietary alternatives when the output is freely available, like freely down-
loadable open source software, and when value can be captured elsewhere in the value chain. Android can be viewed as an
example of where this latter downward force on price has occurred. Google does not need to charge for use of its Android
operating system, because the company can rely on downstream revenues from a complementary good, paid search.4
Regarding features, user-developers can make contributions to the open project to address their specific needs. As Shapiro
and Varian (1999) and Swann (2014) demonstrate, a boost to consumer surplus can be realized when users are freed from
the lock-in of proprietary systems in favor of open regimes of self-determination. Open projects thus maximize consumer
surplus by allowing customers to enjoy desired (or even tailored) features at a lower price while avoiding lock-in to a
particular company’s ecosystem.
In sum, when deciding in which cell to be, companies need to weigh the monopoly rents from ClosedeClosed versus the
“double-agent” returns of the OpeneOpen cell (where the innovator also is the user) (David, 2001), which are influenced by
the value created through producer and consumer surplus. But as the value shifts over time due to the factors analyzed below,
firms can find their strategies drifting across the diagonals of Figure 1.

The factors driving strategic choice

The payoffs to consumer and producer surplus can precipitate the migration of strategy, and this section examines how
this occurs through user-driven, firm-driven, and market-driven forces. The analysis also introduces the notion that firms

2
Crowdsourced innovation projects in a variety of sectors also have attracted a considerable number and variety of contributors (Hautz et al., 2010).
3
This desire to fuel adoption of a new technology standard lays at the heart of Tesla’s announcement about opening up its patent portfolio (Musk, 2014).
This is another setting with strong network effects, and Tesla, as a late entrant, is partnering with established firms Nissan and Toyota to build out a joint
charging network in the United States.
4
As some observers in Silicon Valley put it: “If you are not paying for the product, you ARE the product.” See Chesbrough and Appleyard (2007) for more
examples of how firms can profit from open approaches.
314 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321

might actually pursue a hybrid strategy that exhibits features of both a Closed and Open strategy reflecting the interaction
between user and firm preferences.
Our first consideration is the rationale behind moving from a Closed to Open strategy and how the benefits accruing to
users can help spur such movement. Firms following Closed strategies based on competitive barriers such as intellectual
property rights, by and large do a good job of capturing value via higher prices and a larger producer surplus (Shapiro and
Varian, 1999), which explains why many firms choose a closed strategy initially. However, closed strategies need to
generate enough consumer surplus to entice customers to buy their offering. Knowing that the offering is closed and thus
must be purchased “off-the-shelf,” customers must derive considerable value to motivate them to buy. The obvious rationale
to open up an initially closed strategy is that the offering failed to generate enough value to attract enough customers to
sustain the closed strategy. This was the case with Mozilla, for example. Microsoft’s bundling of its Explorer browser with the
Windows operating system created an insuperable barrier for Netscape Navigator. The latter had to be downloaded and
installed in order to be used, while Microsoft’s offering was already installed, and offered for free.
In response, Netscape eventually decided to release an open-source version of its Navigator browser, called Mozilla. While
it could no longer sustain the further development of this browser on its own in competition with Microsoft, it reasoned that
an open community might want an extensible browser that users and developers could modify as they wished. Mozilla
remains a viable browser option to this day, twenty years after it was put into open source. So Netscape’s calculation proved to
be prescient.
This example of Closed to Open migration reflects how the payoffs captured by “user-innovators” can lead to strategic
change. The increased payoffs to users derive from improved features, self-determination, and a reduced price can accrue as
follows:

(1) The benefits attributable to “self-service,” i.e., allowing users to tailor the product to their needs (Chesbrough and
Appleyard, 2007), can offset relative costs penalties d a higher cost of installation in the case of Mozilla.
(2) An additional benefit of an open source alternative is the elimination of software licensing fees. The pressure by lead
users for an open source alternative to bring down such fees helped persuade Cisco to initiative the OpenDaylight
project (Duffy, 2013), another Closed-to-Open example in Figure 1. Many of these lead users subsequently contributed
code to the project.

The less-studied case is the movement from Open to Closed in Figure 1. The factors influencing such movement can be
grouped into firm-driven and market-driven categories.
The firm-driven factors depend on the strategic calculus of the firm. A firm contemplating abandonment of an open
initiative might weigh the following, which reflect the dynamics of producer surplus related to achieving product quality,
coordination across the ecosystem, speed of deployment, and threats to profitability:

(1) Has the firm accumulated enough internal technology prowess to guarantee quality and carry the initiative forward
without the help of the community? This factor will be considered in greater detail below.
(2) A major part of making most open initiatives successful is the emergence of a supporting ecosystem. Has the firm’s
ability to manage all of the associated touch-points with the open initiative, i.e., its “Architectural Management,” been
honed enough? For example, in the Android case above, as Google reverts to a proprietary approach, it needs to manage
not only the Android operating system, but also the relationships across the ecosystem spanning the independent apps
developers, the handset makers, and the telephony and Wi-Fi service providers. Have the ecosystem partners
continued to invest in and support the offering? If not, maybe they are not adding much value anymore. Or if they have,
might there be the risk that they actually take control of the offering (as occurred with Intel and Microsoft with the IBM
PC)?
(3) Another critical player in the ecosystem is the customer. If a firm were to close off a project, has its Architectural
Management team forged a strong and credible relationship with the user base, such that there will not be mass
defection if they fully take over the project, close it off, and move it in-house? When considering the customer
response, the customer base can be usefully separated into legacy customers (who have already invested substantial
time and money into the offering) and newly arriving customers (who have not).
(4) Has there been a dissipation of interest within the sponsoring firms, because the results from the open project have
fulfilled their goals or have fallen short of the goals (Schweik and English, 2012)?
(5) Has the success of an open initiative prompted a firm to acquire it and close it down? This can occur if the open
initiative threatens the profitability of a main line of business of the potential acquirer, as in the example of Oracle’s
acquisition of Sleepycat, after which Oracle changed the Sleepycat software license to make adoption of the open
version much less attractive (Phipps, 2013).

The market-driven factors address the evolution of the customer base and the preservation of the customer experience.
They blend the payoffs to consumer and producer surplus:
 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321 315

(1) Firms involved in an open initiative must develop business models that balance value creation with value capture. The
arrival rate of new customers into the market addressed by the open initiative drives value creation. Therefore when
markets are small and arrivals are high, there are benefits to an open approach from the customers’ perspective,
because they value not having to “lock-in” to a proprietary architecture (Shapiro and Varian, 1999) and appreciate
expanded product choice. Open accelerates adoption faster than Closed, because consumers are benefiting from self-
determination while producers are benefiting from coordination and speed of adoption.
However, when markets become large and the arrival of new customers slows, this reduces the further benefits of
remaining open. The strategic emphasis shifts from value creation to value capture. The motivations behind such a shift
parallel those behind the migration from exploration to exploitation in high-risk environments (March, 1991; Raisch et al.,
2009; Rothaermel and Deeds, 2004; Smith et al., 2010; Volberda et al., 2001).
(2) As markets mature, customers also care about the stability of the product platform, when either there is a steep
learning curve to use the product or the need for interoperability with complementary goods and services. The con-
sumer desires feature stability and the producer wants the ecosystem to achieve a seamless customer experience.
These desires are particularly salient when a large proportion of the customer’s employees need to be trained to use the
product or related integrated system (Shapiro and Varian, 1999). The customer’s desire for stability is a conceivable
argument behind Google’s decision to close off the Android operating system.
Closing off Android means that Google decreases the likelihood of the source code being “forked” into multiple versions
for which they could not guarantee quality or compatibility. The argument is that excessive fragmentation could
complicate the customer experience to the point that customers defect for a more stable platform such as Apple’s.

These effects can be synthesized along two dimensions: the “LiberalizationeControl” continuum, which reflects the firm’s
assessment of the firm-driven factors; and the “VarietyeStability” continuum, which reflects the market-driven factors
influenced by customer behavior (Figure 2). The interactions of the two dimensions can add subtleties to the firm’s strategic
choice. The “Closed” quadrant in Figure 2 reflects the situation described in the market-driven factor (2) above when cus-
tomers would be willing to lock into a stable, closed system like Apple’s iOS to ensure the customer experience does not
appreciably change over time. In contrast, customers might favor feature variety d and even the ability to create their own
features d so a project with these characteristics like Linux would be in the “Open” quadrant.

Customer Preference
Variety

Internal Versioning Open

(e.g., MicrosoŌ: Xbox) (e.g., Linux)

Lead Innovator’s
Preference
Control LiberalizaƟon
Closed External Versioning
(e.g., Apple: iOS) (e.g., Intel: Dual Source)

Stability

Figure 2. The Interaction between Innovator and Customer in Strategy Creation

To understand the off-diagonal quadrants, we consider strategic choices that are composites of open and closed char-
acteristics. These quadrants demonstrate the possibility that firms may choose a more nuanced strategy by relaxing the
binary choice between Open and Closed. An example of a strategy that ensures stability but loosens intellectual property
rights ownership occurs when a firm guarantees a second source for its product (Shapiro and Varian, 1999), which again
reflects a customer demanding stability. The requirement of a second source is common in critical applications, e.g., tech-
nologies used in the national defense arena. Companies that supply such technologies, like IBM, can require that their parts
suppliers, like Intel, license their chip technologies to a second source of supply, like AMD.5 Unlike the example of stability
leading to a Closed strategy, this quest for stability would instead require Intel to relax, or “liberalize,” its control over its
product by licensing its chip design to AMD. This customer requirement would force Intel to open up and move into the
hybrid “External Versioning” quadrant of Figure 2.

5
IBM also obtained a manufacturing license to Intel’s microprocessor design, so that it could act as its own second source.
316 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321

The dynamics behind the movement from the Closed quadrant to “Internal Versioning” could be driven by pressure from
the customers for new features, or greater variety, rather than greater stability. Internal Versioning allows a firm to maintain
control over much of its intellectual property portfolio but open up enough to allow others to develop complements that
would work with a firm’s core technology. An example can be drawn from the video gaming industry. In this industry, new
generations of consoles allow for enhanced graphics capabilities and engagement with game play. To enter this industry,
superior technology prowess is required to make a new console so compelling that the barriers to entry from network effects
are broken and gamers are willing to switch to the new system. But over subsequent generations of the console, the cus-
tomers would want to be able to use their library of games, so the console maker would need to balance backward
compatibility with new features. This balance is also affected by the number of newly arriving customers relative to legacy
customers.
Internal Versioning is defined as this ability of incumbents to release multiple generations of propriety systems to satisfy
customers’ desire for variety and feature improvement. Microsoft demonstrated this ability to marshal internal technology
prowess to enter the console industry with the Xbox (Takahashi, 2011), and exhibited sensitivity to backward compatibility in
subsequent generations. Microsoft also understood the need to relax control to attract complementors namely independent
game developers, thus embracing some horizontal movement in Figure 2 toward Liberalization. As an example of this
commitment to foster complements, at the time of launch, one of Xbox’s marquee game titles, Halo, had originated outside of
the company.
This analysis suggests the broad and nuanced presence of these kinds of OpeneClosed dynamics across industries, and
while a lead innovator may lean towards preferring a Closed strategy based on proprietary technology, customers may force
migration to a more Open strategy to reduce their lock-in, while benefitting from positive network effects, technological
advancement, and engaged complementors.

Enabling competencies

A firm’s ability to capture the gains to its strategic choice ultimately dictates its willingness to sustain participation in open
initiatives, but which competencies inside the firm most effectively enable value capture? Two critical competencies that
contribute to a firm’s inter-temporal ability to pursue Open Strategy are its technology prowess and organizational processes
that support openness.

The technology prowess competency

To further understand the firm-level assets that push companies either toward or away from open innovation initiatives,
we draw lessons from an analytical model developed by Appleyard et al. (2008). As developed in the model, the pivotal
competency that sways the firm’s strategic decision between open and closed innovation is its technology prowess relative to
its potential collaborators. Technology prowess is defined as the firm’s ability to apply its knowledge stock to the innovation
initiative at hand. But because of the “ecosystem effect” explained below, even firms with superior technology prowess might
wish to collaborate to expand the market and lessen the uncertainty of sustained market adoption, because a supporting
ecosystem is more likely to rally around a larger user base.
In terms of a firm’s ability to capture value from its open initiatives, the model considers profit levels in different scenarios:
cooperative, duopolistic, and monopolistic. It assumes there is a technology leader that initiates the innovation project. In
instances where the technology prowess of the leader’s staff overlaps with that of close competitors, the firms will produce a
similar product such that they will split the end-product market. The value to the user who consumes their end products is
determined by the firms’ respective technology prowess and by users’ specific needs, because users will have to incur a cost to
adapt the product if it is not perfectly aligned with their needs.
Depending on a firm’s level of technology prowess, it can be shown that a firm could become a natural monopolist even if a
competitor’s product were free, in the setting where the former firm’s product is so technologically advanced and fits the
user’s needs so precisely. However, for other levels of the focal firm’s technology prowess, where their innovation capabilities
are so similar to those of the other firm, there is a willingness to cooperate in an open initiative. Although still positive, the
profit level for the focal firm falls, as multiple firms share the profits, but a supporting ecosystem emerges because of the
larger, joint market created by the focal firm and its collaborators.
The ecosystem effect of the Appleyard et al. (2008) model captures the willingness of an ecosystem to form in order to
support deployment of the new product. In the software sector, crucial members of a supporting ecosystem are the systems
integrators, like Accenture or IBM, who combine software and hardware into a functional system. If the lead innovator in an
open source software project were to close off the project too quickly, the market might be too small for systems integrators to
be willing service the market. Similarly, if the closed project does not open up enough it might be preempted by an open
alternative that siphons off support from systems integrators. This “mutualism” experienced by competing innovators reflects
a shared need for ecosystem support and provides a countervailing force to abandoning an open strategy.
Applying this intuition to the Google Android example from above, the switch from Open to Closed can be triggered by a
change in Google’s technology prowess. As Google’s technology prowess in software development has grown, not only
through experience working on Android but by hiring leading Android software developers into Google (Dahlander and
Wallin, 2006), Google’s payoffs approach the natural monopoly situation. This result is similar to a learning race where
 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321 317

one strategic alliance partner learns more from the other partner than vice versa and then abandons the relationship (Kale
et al., 2000; Khanna et al., 1998).
The Closed to Open example of Cisco’s OpenDaylight project reflects the reverse dynamic, where a company with a large
lead it in its technology prowess is actively inviting other companies to collaborate, which effectively erodes its relative
prowess over time. In this case, Cisco is willing to share (cede) its technology prowess to avoid being disrupted in a
Christensen (1997) sense, while ensuring a say in the direction of the emerging de facto standard in software-defined
networking (Shapiro and Varian, 1999). The willingness of Cisco to share control of the project and become more inclusive
over time can be verified by monitoring the code contributions to important modules in the project. This monitoring is
possible because of the transparency of the source code.6

The organizational support competency

Another critical competency required to effectively execute Open Strategy is a firm’s ability to provide organizational
support for the open initiative (Herzog, 2011). Utilizing Open Strategy requires substantial attention to how a firm will
support the initiative across all levels of the organization (Chesbrough and Crowther, 2006; Dodgson et al., 2006;
Jarzabkowski et al., 2007; Sieg et al., 2010; Whittington, 2006). Developing internal organizational processes that facilitate
participation is paramount, particularly for firms that are shifting their strategy from Closed to Open.
To understand the necessary changes in organization processes, we interviewed leaders from a cross-section of organi-
zations within the open source software sector. These interviews reflect the evolution of open innovation projects, where
people no longer question whether openly sourced initiatives will survive, but instead are asking how their internal and
external processes need to evolve to fully deploy Open Strategy.
The process question becomes how to set the level of commitment to Open Strategy beyond merely opening up parts of
one’s product portfolio. Mike Woster, COO of the Linux Foundation, proposes four organizational strategies depending on the
firm’s level of commitment to the open initiative: consumer; participant; influencer; and initiator (Woster, 2014). The consumer
model expressly targets cost savings by adopting free and open source software, similar to consuming third party off-the-
shelf software. Participants typically pursue limited interactions with specific OSS communities, such as finding and pro-
posing solutions to software bugs. Companies exerting an influencer strategy can sway the direction of the software project
with software contributions that extend or modify functionality (Dahlander and Wallin, 2006). Finally, initiators are
contributing software and building communities around it so as to make it a market standard, which often leads to the
commoditization of competitors in particular application areas (West and Gallagher, 2006).
This progression requires a conversion of organizational practices. One of the most extensive commitments to the initiator-
end of the strategy spectrum by an established (proprietary) company has been by IBM. Not only has IBM contributed over $2
billion to the development of Linux (IBM, 2013; Wilcox, 2000), but the company reengineered its internal processes to sustain
Open Strategy. According to Dan Frye, the head of Open Systems Development at IBM for over a decade, they had to alter their
internal processes at both an organizational-level and individual developer-level to work with OSS communities (Frye, 2014).
Because of its crisis years in the late 1980s and early 1990s (Mills and Friesen, 1996) and the company’s experimentation
with business models (Chesbrough, 2010), IBM’s senior leadership was responsive to understanding OSS and how to change
its processes to support a leadership role in the emerging ecosystems. While IBM’s OSS business model has increasingly
focused on the services it wraps around OSS, the company initially captured value by selling the hardware underneath, so on
balance, OSS was not seen as a threat (even though IBM had a sizable proprietary software business) but as an opportunity.
This garnered top leadership support and allowed Frye and his team to alter the processes necessary to establish IBM’s role as
an influencer in various OSS ecosystems, with the most noteworthy being the Linux ecosystem.
Another technology leader, Intel, similarly progressed through the participant, influencer, and initiator stages, once open
source initiatives were identified as a strategic imperative in the early 2000s. Like IBM, Intel had a large, in-house, proprietary
software group, and so organizational buy-in was critical. Imad Sousou, General Manager of Intel’s Open Source Technology
Center, observed that senior leadership gave two overarching reasons why it was vital for Intel to aggressively pursue OSS: 1)
A traditional business strategy argument around complements where better, more widely available software would support
mainline businesses (chip and server sales); and 2) If OSS could help accelerate the overall pace of innovation, then Intel
would grow with it (Sousou, 2014). For example, if OSS fueled burgeoning sectors like the Internet of Things (IoT), then Intel
would benefit through the sale of more Intel chips. Furthermore, in the late 1990s, Intel faced instances of being constrained
by proprietary software vendors in the IT infrastructure sector that refused to support Intel systems, and so OSS allowed Intel
to circumvent those barriers. A few early and decisive wins of opening up software to promote the growth of other business
units within Intel induced organizational support of OSS (Sousou, 2014).
With strategic clarity came the need to define internal processes. Similar to IBM, Intel’s OSS vetting committee comprises
legal, business, and technical leaders from each major time zone. Hundreds of projects have gone through their review
process. Such a process allows Intel to analyze the growth potential associated with each OSS project, while mitigating the

6
An earlier example of an erosion of a leader’s technology prowess comes from the opening up of Mozilla. Originally intended to act as a proprietary web
browser, this technology lost its early lead to Microsoft’s Internet Explorer. As noted above, once IE became more pervasive, the Mozilla project was put into
the open domain.
318 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321

risks and potential costs of infringing someone’s intellectual property or unintentionally disclosing a trade secret (Sousou,
2014). Sousou noted that pursuing OSS cannot be a distraction with a limited level of commitment, but rather part of a
sound business strategy, technically feasible, and without any licensing issues. And like IBM, Intel has had to modify its in-
ternal processes to work effectively as a member of OSS projects.

Discussion and extensions

In this paper, we have shown that the choice of how open to be in one’s innovation strategy is not a once-and-for-all
decision. Instead, we have presented evidence that while some projects remain open or remain proprietary over time,
other projects switch modes or even exhibit hybrid modes that blur the binary choice between Open and Closed. Under-
studied in the literature is the transition from Open to Closed, which we call reversion. Evidence of reversion, such as in the
Android operating system, might reflect the customer’s desire for platform stability after a network has grown large enough
to attract a supporting ecosystem, and the lead innovator, Google in this case, has amassed enough technology prowess to be
able to shift to value capture from market expansion. These dynamics have important implications for both the content
branch of Open Strategy as well as the process branch.
This paper proceeds through three layers of analysis. First, we consider the underlying payoffs to openness. The payoffs can
be categorized as consumer surplus, producer surplus, or even both when the innovator is the user. The allocation of gains can
shift over time which can trigger a shift in strategy.
Second, we analyze the user-, firm-, and market-level drivers derived from consumer and producer surplus that may cause
a firm to switch strategies. A synthesis of a subset of these drivers shows how customer preferences for product stability
versus feature variety can interact with the willingness of the lead innovator to relinquish some control over proprietary
assets. As the product market matures, legacy customers who likely prefer stability may prove more influential if fewer new
customers are entering the market, and this dynamic could ultimately lead to reversion.
But a strategic shift can only occur if the third layer of analysis supports it, where the third layer represents organizational
competencies. To be in a position to capture the value created in an open initiative, our analysis finds that competencies
spanning technology prowess and open-supporting organizational processes emerge as vital.
Ironically, a firm’s commitment to an open project might precipitate reversion over time. In the specific setting of
open initiatives that focus on innovation, the growth of associated knowledge assets leads to the deepening of the
technology prowess competency reflecting the accumulation of experience as well as the movement of top talent across
company boundaries. The ability of firms to grow their knowledge assets through these means during an open inno-
vation project can, over time, reduce their need to collaborate with outsiders. Therefore, multiple forces might pre-
cipitate reversion and additional research is needed to assess their strength relative to the countervailing payoffs to
openness.

Conclusions and directions for future research

Open Strategy has evolved to encompass two primary dimensions: a process dimension that examines the effects of
substantially greater participation in the strategy determination process (Whittington et al., 2011); and a content dimension
that considers the sustainability of open innovation approaches (Chesbrough and Appleyard, 2007). More recent proponents
of open innovation have celebrated the value of opening up (Chesbrough, 2003; von Hippel, 2005), but issues of whether and
how this openness can be sustained have not received adequate attention (Chesbrough and Appleyard, 2007).
The possibility of switching strategy of being Closed or Open over time has been neglected by both the process side and the
content side of Open Strategy. From the process branch of Open Strategy, it is likely that at least some adherents of a much
more participatory strategy formulation process at a later time revert to a more restricted process. Understanding when this
reversion occurs will help to clarify the boundary conditions that enable Open Strategy formulation processes to thrive over
time. Within the content branch of Open Strategy, the underlying dynamics that a lead innovator might face with regards to
its technology prowess relative to a collaborator, coupled with customer preferences for stability or variety, might alter the
anticipated payoffs to a particular innovation strategy and cause the pendulum to swing back.7 The possibility of switching
content strategies also calls attention to the internal organizational practices of firms, and the range of payoffs that figure into
the calculus of whether to switch or not. Only future systematic analysis across industries will be able to assess the limits to
Open Strategy and open innovation more generally.
While our primary examples have been within the ICT and software sectors, we strongly suspect that these dynamics are in
play in other parts of the economy, meriting future research. Wherever there are extensive value chains, supported by sur-
rounding ecosystems of complementary suppliers and third parties that require coordinating actions and investments, the
dynamics of the lead innovator’s motivations and the market’s characteristics are likely to apply. For example, in many
standards-setting battles, seemingly small initial differences can tip in the favor of a more open standard over a more
restricted one d or vice versa. The classic Beta vs. VHS example has been analyzed this way by Cusumano et al. (1992).

7
In software, open projects worried about being closed off can thwart such moves. The code can be forked into another version governed by an open
license, as happed to MySQL as it was being acquired by Oracle as part of Oracle’s acquisition of Sun (Pearce, 2013).
 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321 319

Empirical research into standards organizations reveals a wide range of rules and a substantial role for discretion for the lead
innovator as documented by Lemley (2002), including the initiator proposing the standard.
The pursuit of Open Strategy necessitates the construction of business models that will lead to economic viability by
allowing participating firms to not only create but capture value. How value is created and captured can change over the life of
an open initiative, and in some cases, can lead the firm away from openness. Viable Open Strategy business models have been
demonstrated in the realm of open source software (Chesbrough and Appleyard, 2007; Perr et al., 2010; West and Gallagher,
2006), but parallel (or alternative) business models in other industries have received less attention. Software firms that value
time-to-market often opt for open source, but wrap proprietary add-ons (or services) around it. To help formulate a strategy
for portfolio balancing between Open and Closed products, companies like Red Hat are consulting with system integrators
like Accenture. These relationships deepen the software sector’s ability to determine when companies should pursue
(traditional) open source projects and when they should instead apply open source philosophies to internal, proprietary
development projects (so-called inner sourcing). Future studies can survey firms that participate in open initiatives to detail
where in the value chain they are capturing value and how Open and Closed products and services are co-existing in their
portfolios.
Not only have business models that sustain Open Strategy received limited attention outside of software, but few studies
detail the internal competencies a traditional, proprietary-oriented firm must develop to achieve openness. Two prevailing
competencies that we consider in this paper are technology prowess and organizational processes that will support the non-
traditional strategic choice of Open Strategy (Barney, 1991; Chesbrough, 2003). As Chesbrough (2003) establishes, without a
supportive organizational structure, other competencies that are valuable, rare, and costly to imitate will not lead to sustained
competitive advantage. Delving into the organization processes including reward systems, promotion paths, and norms
governing interactions with others in open initiatives could help with the understanding of how firms are able to migrate
along the Open Strategy spectrum.
This research implies testable hypotheses that warrant further research. Future empirical research could specifically
analyze how technology prowess influences participation. As noted, we anticipate that accumulated technology prowess
would be expected to increase the likelihood of withdrawal from an open initiative and even possibly trigger reversion. One
can go about testing this hypothesis through a longitudinal examination of contributions to open projects, like the code
contributions made over time in open source software projects.
More generally, future research could test whether Open Strategy leads to heightened competitive payoffs. As noted, open
projects are anticipated to enjoy time-to-market, cost, and quality advantages, and these advantages would be expected to
translate into improved margins in the short-run and greater market capture and profitability in the long-run. Open projects
that embrace transparency like open source software publish who is contributing to which modules of the project. Such
information aligns the varied interests of developers, users, complementors and third parties within the ecosystem,
improving performance while reducing cost and accelerating time-to-market. The projects that switch their modes over time
face the possibility of misalignment with one or more external parties as a result of the switch. The presence of alignment or
the potential for misalignment may affect the performance of these projects and hence the firms supporting these projects.
Parties who are no longer properly aligned with the new mode may suffer diminished performance, when compared to others
who have aligned themselves. Comparisons of product market performance, profitability, and stock performance between
member firms of an open initiative and matched non-member firms offer metrics by which to judge the advantages of
openness. For example, stock price performance might be assessed following a switch in mode from Open to Closed or Closed
to Open.
Another interesting research question is the apparent lack of concern exhibited by participants in open projects that are
reverting to a more proprietary approach. One might expect volunteer contributors to these projects to have strong, negative
reactions to this reversion, and this anticipated effect could be tested, for example, by analyzing patterns in code contributions
in open source software projects that are starting to close off. Countering the expectation that volunteer contributors might
curtail their contributions, we have found only modest evidence of defection in response to reversion in our interviews with
leaders in the OSS arena.
It is clear that open innovation is an inherently dynamic process. Strategies to support open innovation must similarly
incorporate dynamic elements if they are to reflect the reality of today’s technological world. For both reasons of rigor and
of relevance, we hope others will pursue these dynamics in future work on both the content and the process of Open
Strategy.

Acknowledgments

The authors would like to thank AnnaRose Adams, Devdeep Aikath, Bart Massey, Desire
e Pacheco, Lauren Simon, and Jesse
von Doom for helpful input, and the industry leaders who generously provide their time and insights during interviews. The
editors of this special issue and two anonymous reviewers provided very useful comments on an earlier draft of this research,
but all errors remain the responsibility of the authors.

References

Allen, R.C., 1983. Collective invention. Journal of Economic Behavior and Organization 4 (1), 1e24.
320 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321

Amadeo, R., 2013. Google’s iron grip on Android: controlling open source by any means necessary. Ars Technica, October 20, 2013. http://arstechnica.com/
gadgets/2013/10/googles-iron-grip-on-android-controlling-open-source-by-any-means-necessary/1/. accessed 23.06.14.
Android, 2014. Why Apache Software license? http://source.android.com/source/licenses.html#why-apache-software-license accessed 23.06.14.
Appleyard, M.M., Wang, C.Y., Liddle, J.A., Carruthers, J., 2008. The innovator’s non-dilemma: the case of next-generation lithography. Managerial and
Decision Economics : MDE 29 (5), 407e423.
Baldwin, C., von Hippel, E., 2011. Modeling a paradigm shift: from producer innovation to user and open collaborative innovation. Organization Science 22
(6), 1399e1417.
Baldwin, C., Hienerth, C., von Hippel, E., 2006. How user innovations become commercial products: a theoretical investigation and case study. Research
Policy 35 (9), 1291e1313.
Baldwin, C.Y., Clark, K.B., 2000. Design Rules, Vol. 1: The Power of Modularity. The MIT Press, Cambridge, MA.
Barney, J., 1991. Firm Resources and Sustained Competitive Advantage. Journal of Management 17 (1), 99e120.
Business Wire, 2014. Android and iOS continue to dominate the worldwide smartphone market with Android shipments just shy of 800 million in 2013,
according to IDC. http://www.businesswire.com/news/home/20140212005399/en/Android-iOS-Continue-Dominate-Worldwide-Smartphone-
Market#.U16231erHag accessed 23.06.14.
Chesbrough, H., 2010. Business model innovation: opportunities and barriers. Long Range Planning 43 (2), 354e363.
Chesbrough, H., Crowther, A.K., 2006. Beyond high tech: early adopters of open innovation in other industries. R&D Management 36 (3), 229e236.
Chesbrough, H.W., 2003. Open Innovation: The New Imperative for Creating and Profiting From Technology. Harvard Business School Press, Boston, MA.
Chesbrough, H.W., Appleyard, M.M., 2007. Open innovation and strategy. California Management Review 50 (1), 57e76.
Chesbrough, H.W., Teece, D.J., 1996. When is virtual virtuous. Harvard Business Review 74 (1), 65e73.
Christensen, C.M., 1997. The Innovator’s Dilemma. Harvard Business School Press, Boston, MA.
Cusumano, M.A., Mylonadis, Y., Rosenbloom, R.S., 1992. Strategic maneuvering and mass-market dynamics: the triumph of VHS over beta. Business History
Review 66 (1), 51e94.
Dahlander, L., Wallin, M.W., 2006. A man on the inside: unlocking communities as complementary assets. Research Policy 35 (8), 1243e1259.
David, P.A., 2001. Notes on ‘trading in trade secrets’ and the economics of double-agency. NSF/SIEPR Project on Economic Organization and the Viability of
Open Source Software, Working Paper Series: 2001-02.
Dodgson, M., Gann, D., Salter, A., 2006. The role of technology in the shift towards open innovation: the case of Procter & Gamble. R&D Management 36 (3),
333.
Duffy, J., 2013. Cisco, IBM on the defensive with OpenDaylight. Network World. http://www.networkworld.com/article/2224451/cisco-subnet/ciscoeibm-
on-the-defensive-with-opendaylight.html. accessed 16.05.16.
Dwoskin, E., 2016. Tech firms grapple with how to make open source pay. WSJ.com, March 3, 2016. http://www.wsj.com/article_email/tech-firms-grapple-
with-how-to-make-open-source-pay-1457031731-lMyQjAxMTI2MjAzNDMwMTQwWj. accessed 04.03.16.
Eisenmann, T., Parker, G., Van Alstyne, M., 2011. Platform envelopment. Strategic Management Journal 32 (12), 1270e1285.
Fisher, K., 2006. Sun opens Java under the GPL. Ars Technica, November 13, 2006. http://arstechnica.com/business/2006/11/8205/. accessed 16.03.16.
Franke, N., Shah, S., 2003. How communities support innovative activities: an exploration of assistance and sharing among end-users. Research Policy 32,
157e178.
Frye, D., 2014. Interview regarding Open Source Software Strategy at IBM. Personal communication, October 7, 2014.
Grossman, G.M., Helpman, E., 1994. Endogenous innovation in the theory of growth. The Journal of Economic Perspectives: A Journal of the American
Economic Association 8 (1), 23e44.
Hautz, J., Hutter, K., Füller, J., Matzler, K., Rieger, M., 2010. How to establish an online innovation community? The role of users and their innovative content.
In: Proceedings of the 43rd Hawaii International Conference on System Sciences. IEEE, pp. 1e11.
Herzog, P., 2011. Open and Closed Innovation: Different Cultures for Different Strategies, 2nd ed. Gabler Verlag, Springer Fachmedien, Wiesbaden, Germany.
IBM, 2013. IBM commits $1 billion to fuel Linux and open source innovation on power systems. IBM News room. http://www-03.ibm.com/press/us/en/
pressrelease/41926.wss. accessed 05.11.14.
Jarzabkowski, P., Balogun, J., Seidl, D., 2007. Strategizing: the challenges of a practice perspective. Human Relations; Studies Towards the Integration of the
Social Sciences 60 (1), 5e27.
Kale, P., Singh, H., Perlmutter, H., 2000. Learning and protection of proprietary assets in strategic alliances: building relational capital. Strategic Management
Journal 21 (3), 217e237.
Khanna, T., Gulati, R., Nohria, N., 1998. The dynamics of learning alliances: competition, cooperation and relative scope. Strategic Management Journal 19
(3), 193e210.
Lakhani, K.R., von Hippel, E., 2003. How open source software works: “free” user-to-user assistance. Research Policy 32 (6), 923e943.
Lemley, M.A., 2002. Intellectual property rights and standard-setting organizations. California Law Review 90 (6), 1889e1980.
Lerner, J., Tirole, J., 2002. Some simple economics of open source. The Journal of Industrial Economics L (2), 197e234.
March, J.G., 1991. Exploration and exploitation in organizational learning. Organization Science 2 (1), 71e87.
Mills, D.Q., Friesen, G.B., 1996. Broken Promises: An Unconventional View of What Went Wrong at IBM. Harvard Business School Press, Boston, MA.
Musk, E., 2014. All our patent are belong to you. http://www.teslamotors.com/blog/all-our-patent-are-belong-you accessed 21.06.14.
Nordhaus, W.D., 1969. An economic theory of technological change. The American Economic Review 59 (2), 18e28.
O’Mahony, S., Bechky, B.A., 2008. Boundary organizations: enabling collaboration among unexpected allies. Administrative Science Quarterly 53, 422e459.
Open Handset Alliance, 2007. What types of companies are in the Open Handset Alliance? November 2007. http://www.openhandsetalliance.com/oha_faq.
html. accessed 23.06.14.
Pearce, R., 2013. Dead database walking: MySQL’s creator on why the future belongs to MariaDB. Computerworld, March 28, 2013. http://www.
computerworld.com.au/article/457551/dead_database_walking_mysql_creator_why_future_belongs_mariadb/?pp¼2. accessed: 16.03.16.
Perr, J., Appleyard, M.M., Sullivan, P., 2010. Open for business: emerging business models in open source software. International Journal of Technology
Management 52 (3/4), 432e456.
Peteraf, M.A., 1993. The cornerstones of competitive advantage: a resource-based view. Strategic Management Journal 14 (3), 179e191.
Phipps, S., 2013. Oracle switches Berkeley DB license. InfoWorld, July 5, 2013. http://www.infoworld.com/article/2611450/open-source-software/oracle-
switches-berkeley-db-license.html. accessed 16.03.16.
Pon, B., Seppa €la
€, T., Kenney, M., 2014. Android and the demise of operating system-based power: firm strategy and platform control in the post-PC world.
Telecommunications Policy 38, 979e991.
Porter, M., 1980. Competitive Strategy. Free Press, New York.
Raisch, S., Birkinshaw, J., Probst, G., Tushman, M.L., 2009. Organizational ambidexterity: balancing exploitation and exploration for sustained performance.
Organization Science 20 (4), 685e695.
Raymond, E.S., 2001. The Cathedral and the Bazaar. O’Reilly Media, Inc., Sebastopol, CA.
Romer, P.M., 1990. Endogenous technological change. The Journal of Political Economy 98 (5), S71eS102.
Rothaermel, F.T., Deeds, D.L., 2004. Exploration and exploitation alliances in biotechnology: a system of new product development. Strategic Management
Journal 25 (3), 201e221.
Schweik, C.M., English, R.C., 2012. Internet Success: A Study of Open-Source Software Commons. The MIT Press, Cambridge, MA.
Shah, S.K., Tripsas, M., 2007. The accidental entrepreneur: the emergent and collective process of user entrepreneurship. Strategic Entrepreneurship Journal
1 (1e2), 123e140.
Shapiro, C., Varian, H.R., 1999. Information Rules: A Strategic Guide to the Network Economy. Harvard Business School Press, Boston, MA.
 M.M. Appleyard, H.W. Chesbrough / Long Range Planning 50 (2017) 310e321 321

Sieg, J.H., Wallin, M.W., von Krogh, G., 2010. Managerial challenges in open innovation: a study of innovation intermediation in the chemical industry. R&D
Management 40 (3), 281e291.
Smith, W.K., Binns, A., Tushman, M.L., 2010. Complex business models: managing strategic paradoxes simultaneously. Long Range Planning 43 (2), 448e461.
Sousou, I., 2014. Interview regarding open source software strategy at Intel. Personal communication, November 7, 2014.
Swann, G.M.P., 2014. Common Innovation: How We Create the Wealth of Nations. Edward Elgar Publishing, Cheltenham, UK.
Takahashi, D., 2011. The making of the Xbox: how Microsoft unleashed a video game revolution (part 1). Venturebeat.com, November 14, 2011. http://
venturebeat.com/2011/11/14/making-of-the-xbox-1/view-all/. accessed 31.08.15.
von Hippel, E., 2005. Democratizing Innovation. The MIT Press, Cambridge, MA.
von Hippel, E., von Krogh, G., 2003. Open source software and the “private-collective” innovation model: issues for organization science. Organization
Science 14 (2), 209e223.
Volberda, H.W., Baden-Fuller, C., Van Den Bosch, F.A., 2001. Mastering strategic renewal: mobilising renewal journeys in multi-unit firms. Long Range
Planning 34 (2), 159e178.
Wernerfelt, B., 1984. A resource-based view of the firm. Strategic Management Journal 5 (2), 171e180.
West, J., Gallagher, S., 2006. Challenges of open innovation: the paradox of firm investment in open-source software. R&D Management 36 (3), 319e331.
Whittington, R., 2006. Completing the practice turn in strategy research. Organization Studies 27 (5), 613e634.
Whittington, R., Cailluet, L., Yakis-Douglas, B., 2011. Opening strategy: evolution of a precarious profession. British Journal of Management 22 (3), 531e544.
Wilcox, J., 2000. IBM to spend $1 billion on Linux in 2001. CNET, December 12, 2000. http://news.cnet.com/IBM-to-spend-1-billion-on-Linux-in-2001/2100-
1001_3-249750.html. accessed 13.11.14.
Woster, M., 2014. Interview regarding Open Source Software Strategy and the Linux Foundation. Personal communication, September 30, 2014.

Biographies

Melissa M. Appleyard is an Ames Professor in the Management of Innovation and Technology at Portland State University. She
previously was a research fellow of U.C. Berkeley’s Competitive Semiconductor Manufacturing Center sponsored by the Alfred
P. Sloan Foundation. Her research focuses on how cooperative research and development and knowledge diffusion catalyze
economic growth and business longevity in technology-intensive industries. Her work has been published in leading aca-
demic journals such as California Management Review, Industrial Relations, the Journal of Product Innovation Management,
Managerial and Decision Economics, and the Strategic Management Journal. She holds a BA from UCLA and a PhD from U.C.
Berkeley.

Henry Chesbrough teaches at the Haas School of Business at U.C. Berkeley and ESADE Business School at Ramon Llull Uni-
versity in Barcelona. His research focuses on open innovation. His managerial books include: Open Innovation (HBS Press,
2003); Open Business Models (HBS Press, 2006); Open Services Innovation (Jossey-Bass, 2011). Academic books include: Open
Innovation: Researching a New Paradigm (Oxford University Press, 2006) and New Frontiers in Open Innovation (Oxford Uni-
versity Press, 2014) both with Wim Vanhaverbeke and Joel West. He received an Innovation Luminary award from the Eu-
ropean Commission in 2014, and he has received honorary doctorates from Hasselt University and the University of Vic.
 POINT OF VIEW

3D Printing Disrupts Manufacturing

How Economies of One Create New Rules of Competition
3D printing may represent a disruption to the manufacturing industry as profound as the Industrial Revolution.

Irene J. Petrick and Timothy W. Simpson

Before the Industrial Revolution, goods were produced by machinery—became separate from variable costs—those
local artisans and craftsmen relying primarily on locally expenditures that increased on a per-unit production ba-
available materials and selling primarily to local customers. sis, such as labor and materials. Economies-of-scale pro-
These artisans conceived of and then made products, and duction models meant that high-volume production
they sold these products in their own small shops or out of reduced the contribution of the fixed-cost portion of the
their homes. In this environment, the customer was directly cost equation, thus reducing the per-unit cost. Simply put,
linked to the producer; there was no middleman and no high throughput and efficiency yielded higher profits
supply chain. (Pine 1993).
The Industrial Revolution ushered in an era of innova- Today we are entering an era many believe will be as
tion in production methods, mining methods, and ma- disruptive to the manufacturing sector as the Industrial
chine tools that enabled mass production and allowed the Revolution was—the age of 3D printing and the digital tools
replacement of labor with machines and of traditional en- that support it (Koten 2013). At a EuroMold fair in Novem-
ergy sources such as wind, water, and wood with coal- ber 2012, 3D Systems used one of its 3D printers to print a
powered (and later gas-powered) machines. In the past hammer. The Economist (2012) used this example to com-
200 years, the elements of production have been refined, pare the traditional supply chain design-build-deliver model
but the underlying economics have remained: competitive with the emerging 3D printing model:
advantage goes to the company or companies (organized
Ask a factory today to make you a single hammer to your
into a supply chain) that can produce the highest quality own design and you will be presented with a bill for
part at the lowest cost. Fixed costs—infrastructure and thousands of dollars. The makers would have to produce
a mould, cast the head, machine it to a suitable finish,
turn a wooden handle and then assemble the parts. To do
that for one hammer would be prohibitively expensive.
Irene J. Petrick is a Penn State University professor and managing director
of the TrendScape Innovation Group. She is an internationally recognized If you are producing thousands of hammers, each one of
expert in strategic roadmapping. Her research interests include technology them would be much cheaper, thanks to economies of
forecasting, collaborative innovation, and business ecosystem develop- scale. For a 3D printer, though, economies of scale will
ment. She is actively engaged with a wide variety of organizations in their matter much less. Its software can be endlessly tweaked
innovation and technology strategy activities. She has over 25 years of ex- and it can make just about anything.
perience in technology planning, management, and product development
in both academic and industrial settings. Irene is author or coauthor on According to Richard D’Aveni (2013), “businesses all along
more than 135 publications and presentations. [email protected]
the supply, manufacturing, and retailing chains [will need]
Timothy W. Simpson is a professor of mechanical and industrial engineer-
to rethink their strategies and operations” (34). Indeed,
ing at Penn State University with appointments in Engineering Design and
Information Sciences & Technology. His research and teaching interests in- the rise of 3D printing and additive manufacturing will re-
clude product family and product platform design, mass customization, de- place the competitive dynamics of traditional economies-
sign innovation and entrepreneurship, and additive manufacturing, and he of-scale production with an economies-of-one production
has authored more than 250 technical publications. From 2007 through
model enabled by 3D printing and additive manufactur-
2012, he served as the director of the Learning Factory, which coordinates
over 150 industry-sponsored capstone design projects for over 800 engi- ing, at least for some industries and products. In essence,
neering students in 12 departments each year. During that time, he helped future manufacturers will be governed by two sets of rules:
establish Penn State’s Digital Fabrication Network (DIGI-Net) to increase economies of scale for interchangeable parts produced at
access to digital fabrication tools and 3D printing across campus. He now
serves as co-director of the Center for Innovative Materials Processing
high volumes, and economies of one for highly customiz-
through Direct Digital Deposition (CIMP-3D), a DARPA-funded manufactur- able products that can be built layer by layer. Each model
ing demonstration facility for additive manufacturing. [email protected] brings its own sources of competitive advantage and eco-
DOI: 10.5437/08956308X5606193 nomic factors (Table 1).

Research-Technology Management • November—December 2013 | 1

TABLE 1. Economies of scale versus economies of one
Economies of Scale Economies of One
Source of competitive advantage Low cost, high volume, high variety End-user customization
Supply chain Sequential linear handoffs between distributed Non-linear, localized collaboration with
manufacturers with well-defined roles and responsibilities ill-defined roles and responsibilities
Distribution High volume covers transportation costs Direct interaction between local consumer/
client and producer
Economic model Fixed costs + variable costs Nearly all costs become variable
Design Simplified designs dictated by manufacturing constraints Complex and unique designs afford
customization
Competition Well-defined set of competitors Continuously changing set of competitors

The Traditional Supply Chain and the Competitive This extended supply chain functioned with a linear hand-
Dynamics of Economies of Scale off between suppliers, with complex assembly and delivery
Traditional manufacturing relies on a design-build- often controlled by the original equipment manufacturers
deliver model. In this competitive model, roles and re- (OEMs). OEMs drove the conceptualization of product needs,
sponsibilities are well established between the various often acting as product designers. The path to the customer for
participants. Designers translate customer needs into vi- companies within these supply chains was through the OEM,
able products. Producers own facilities that emphasize ef- which controlled much of the supply chain participants’ activi-
ficiency and low-cost production. In the past four decades, ties, and which often reaped the lion’s share of the profits.
these producers have increasingly relied on a distributed
and extended supply chain, sourcing the lowest-cost pro- The 3D Production Model
viders to build components and subassemblies on a global The terms 3D printing and additive manufacturing are often
scale. The production methods employed by these manu- used interchangeably, as both refer to the layer-by-layer
facturers have relied heavily on subtractive manufactur- creation of physical objects based on digital files that repre-
ing methods, which begin with a solid physical form that sent their design. 3D printing has been used for more than
is ground, cut, drilled, milled, lathed, and otherwise has two decades, primarily for rapid part prototyping and small-
material removed from it to make the shapes needed to run production in a variety of industries (Gibson, Rosen,
build components, subassemblies, and ultimately com- and Stucker 2010). Meanwhile, the term additive manufac-
plete products. turing has come to represent the use of 3D printing to create
In this production model, reducing variation to enable final parts and metallic components, differentiating from
repetitive production of interchangeable parts provides a the more traditional subtractive manufacturing processes.
competitive advantage. In the 1990s, companies built on 3D printing uses computer-generated designs to create
this advantage by pursuing design for manufacturing (DFM) “build paths” that reproduce a digital model through con-
strategies (see for example, Ulrich and Eppinger 1995; solidation of materials with an energy source. The process
Boothroyd, Dewhurst, and Knight 2002), which empha- typically uses a binder, a laser or an electron beam that so-
sized designing parts that could be built cost-effectively us- lidifies material as it is directed along the build path or
ing traditional manufacturing processes. This model, which scanned over a pre-placed layer of material. To date, this
changed the object of design from creative expression to method has been used successfully with polymers, metals,
cost-effective production and assembly, required simplified and ceramics.1 Polymers have the highest proportion of
designs developed according to a series of design rules that functional prototypes and finished parts produced with this
favored reproducible parts optimized for high-volume man- method, often requiring only limited additional finishing.
ufacturing and material-handling methods. Several genera- Metals, on the other hand, are still in their infancy in terms
tions of designers and engineers have been schooled in this of finished part production. Metallic parts produced with
approach; many now view design as a creative process of 3D printing methods frequently require additional heat
circumnavigating the constraints imposed by traditional treatment or other finishing and post-processing steps to
manufacturing processes. achieve specified tolerances. Capabilities continue to im-
Under the design-build-deliver model, the companies prove for all three types of material systems. Similarly, there
that could achieve high-quality products at the lowest cost is a tradeoff between how fast a part can be produced and
were more successful, and as transportation improved and its final quality—the slower the build rate, the better the
coordination between companies was facilitated by digital surface finish, for example. Parts producers still need to ex-
technologies, the low-wage countries began to dominate in periment with 3D printing speeds and feeds depending on
the production phase. The China price was born, and dis- the material system, and there are not well-understood
tributed supply chains became the norm as the labor sav-
ings significantly offset the added costs of shipping and 1
For a detailed description of 3D printing, see Lipson and Kurman (2013)
transportation. or Barnatt (2013).

2 | Research-Technology Management Point of View

standards or design rules to adequately address this chal-
lenge at the present time. This is, however, an area of much
active research. The combination of hardware and
Any 3D printing process begins with a digital solid model,
often created through computer-aided design (CAD) and software returns to manufacturing the
analyzed with computer-aided engineering (CAE) software. ability to produce anything that can be
For complex product geometries and material combina-
tions, CAE is often further facilitated by high-performance imagined.
computing resources. The combination of hardware and
software returns to manufacturing the ability to produce
anything that can be imagined, rather than limiting designs
to production constraints. For instance, a polymer mesh can The Changing Nature of Design
be made on a 3D printer without any assembly; this mesh, In the simplified version of the traditional supply chain, de-
which is flexible and interlocking, is produced by the melt- signers are the precursor to production. The roles and re-
ing and deposition of the polymer filament. Digital tech- sponsibilities between design and production are well
nologies also exist to scan physical objects and reverse established and clearly delineated. In the 3D printing world,
engineer the computer models and designs needed to re- these roles become blurred, and the notion of who is a de-
produce them on a 3D printer. signer is called into question, as anyone can design products
3D printing has been used successfully for single-unit to be printed. Likewise, the traditional coupling between
and very low-volume production in a variety of sectors computer-aided design and computer-aided manufacturing
ranging from aerospace (Economist 2012) to prosthetics (CAD/CAM) is fractured, as numerous players may be
(Shinal 2013), dental implants (Murray 2012), hearing needed to translate from 2D to 3D, as the unique operating
aids (Sharma 2013), sports equipment (Luna 2013), and codes for the various printers may require additional exper-
even art (Rawsthorn 2013) and fashion (Brooke 2013). tise in file preparation. These shifts have implications across
When it comes to metal 3D printing, aerospace appears to the manufacturing process, from the initial imagination of a
be leading the way, seeing opportunities to produce light- product to its translation into a physical thing.
weight components, reduce manufacturing lead-times, Design has already expanded beyond the expert realm to
and improve the “buy-to-fly” ratio of components—the include hobbyists and prosumers (people who both produce
amount of material purchased to produce a part versus the and consume a product) who work with digital design kits
amount of material actually in the part. Meanwhile, the (like those available at GrabCAD.com) and other resources
ability to produce sophisticated internal geometries using to develop their own customized products. Chris Anderson
additive manufacturing has sparked considerable interest has explored the rise of hobbyist manufacturers in Makers
in the turbine and energy industries, and the medical in- (Anderson 2012), and we hear of “Maker Guilds” starting to
dustry sees opportunities to produce customized devices emerge in large multinational companies like GE to support
and implants for individual patients. Even tool-makers in designers and engineers who have long been divorced from
the oil and gas industry see opportunities to create func- the physical production due to global supply chains (Dods
tionally graded parts that provide different material prop- 2013). Innovations in solid-modeling software that allow
erties to prolong tool life and reduce down-time of well sophisticated models to be accessed and manipulated
operation. While some of these opportunities may take through user-friendly interfaces will bring the power once
several years to realize, test flights with 3D-printed airframe available only to experts to a much wider audience. Au-
components are under way, and several FDA-approved ti- toDesk, which now offers a free app to turn images into 3D
tanium hip and knee replacements are now being fabri- objects (see www.123dapp.com), and other software ven-
cated using additive manufacturing processes. It may be a dors are building these sophisticated systems and beginning
while before the automotive industry invests heavily in to integrate hardware, software, and even cloud services
additive manufacturing, as the build rates and speed of into a unified 3D printing–based production chain.
the current technology are too low to support their high- Finally, digital models will not come only from a diverse
volume production needs. array of designers; they will also be created through reverse
engineering using digital scanning devices that model both
The Competitive Dynamics of 3D Printing external and internal features, creating new intellectual
The 3D production ecosystem will have major effects in property challenges that are just starting to be explored
each of the three major stages of the design-build-deliver (Weinberg 2010). 3D printing frees designers from the con-
model. It will change the nature of design, it will increase straints of traditional manufacturing processes; some even
the interactivity between design and production, and it will argue that it flips DFM—design for manufacturing becomes
radically localize manufacturing as consumers interact more manufacturing for design (Beaman 2013).
directly with each other and with manufacturers, some of There are some bumps in the road toward this vision for
which act as printer hubs, offering services to anyone with design, including software and hardware compatibility is-
a design to print. sues. Realizing a design requires production models that

Point of View November—December 2013 | 3

 treatment) to achieve functional tolerances and perfor-
mance targets. Many industries are wrestling with how
This requires collaborative innovation parts built using additive manufacturing will be qualified.
Many 3D printing systems do not have the process moni-
between materials suppliers, product toring and feedback tools needed to control the process in
designers, and product producers at a real time, and the machine-to-machine and part-to-part
variation that currently exists worries end users. To date,
level never before seen. we have sophisticated destructive and nondestructive tests
to assess strength, flexibility, density, and other critical
product characteristics, but the variation between parts
built on the same machine is forcing many to rethink
methods for inspection and testing, particularly when only
specify build paths. While the STL (STereoLithography) file
a single custom part is needed. Likewise, validating com-
format has become standard input, many 3D printers use
plex internal geometries remains an issue. While full den-
unique software to generate instruction sets and machine
sity may be a very important characteristic to mimic a cast
code to operate the specific machine. A 3D print file devel-
part, for example, in many applications, planned porosity
oped for one printer is not necessarily viable for use on a
or complex geometric shapes inside a part are going to be
different printer, although companies like Microsoft are
more difficult to test and qualify. This remains a barrier to
hoping to change that (Boettcher 2013). The 3D printing
achieving finished quality products, particularly in metal-
software, and CAD packages in particular, are becoming the
lic components, but ongoing research with sophisticated
limiting factor in the design-build-deliver process.
computed tomography equipment should soon change
that.
Experimental Design
Not only will the nature of design change, but there will be
a very tight coupling between how a product is conceived, Localized Distribution and Printer Hubs
how it is manufactured, and how it is tested and qualified. A 3D printing environment where consumers interact di-
The traditional handoff between companies along the sup- rectly with producers will bring two critical changes to the tra-
ply chain and between stages of production within a com- ditional distributed supply chain. First, because manufacturing
pany will no longer be advantageous. A process parameter no longer needs to be centralized for high-volume production,
framework unique to the 3D printer will emerge. Process low-cost sourcing of suppliers no longer makes economic
plans, tool paths, speeds, feeds, and build orientation will be sense. This will result in the localization of both production
directed by the designer-specified product features and will and sourcing and further reduce economies of scale. Inventory
ultimately determine the actual features of the 3D-printed and shipping that now happens in warehouses and large-scale
part. containers will be replaced by smaller-scale shipping methods
In addition, material characteristics will become a critical such as the US Postal Service, FedEx, and UPS. It is even fea-
aspect of design and production. In many cases, materials sible that inventory management will be entirely transformed;
are supplied as powders, and the powder characteristics we are already seeing signs of this transformation. UPS and
(particle size, shape, and distribution) influence the result- Stratasys recently announced a partnership to offer 3D print-
ing microstructure based on the selected processing param- ing in UPS stores across the United States (Nanowerk News
eters, which in turn impacts material properties. This 2013), and Staples, the first major retailer in the United States
requires collaborative innovation between materials suppli- to sell 3D printers directly to customers (Cautela 2013), has
ers, product designers, and product producers at a level begun offering 3D printer services in some of its stores (Senese
never before seen, and the interaction will result in a highly 2012). Even Microsoft is getting into the hardware game and
iterative design process where the goal will be to fail fast will start selling MakerBot 3D printers in several of its US
and often to achieve a workable product. stores (De Zeen Magazine 2013).
Here, too, some challenges remain. The net or near-net The largest shift in distribution, however, may be the rise
shapes resulting from additive manufacturing will still re- of printer hubs that directly support hobbyist and prosumer
quire finishing and post-processing (for instance, heat needs. These are already emerging. For example, Shape-
ways, a 3D printing services company that spun out of
Royal Phillips Electronics, allows clients to post or access
designs, modify them, and upload them to the site via the
Internet. Shapeways then feeds these digital files into their
Economies of scale and economies of one 3D printers to produce the desired object. In 2011, Shape-
ways shipped nearly 750,000 parts in materials ranging
will continue to coexist, but they will not
from plastic and stainless steel to silver and ceramics. In this
be used for the same things. business model, the customer pays per part built; equip-
ment goes from a fixed cost to a variable cost, completely
disrupting the economies-of-scale model.

4 | Research-Technology Management Point of View

Conclusion References
Lipson and Kurman (2013) note, “Bursts of innovation Anderson, C. 2012. Makers: The New Industrial Revolution. New
happen when an emerging technology removes a once York: Crown Publishing Group.
prohibitive barrier of cost, distance, or time” (59). Certainly Barnatt, C. 2013. 3D Printing: The Next Industrial Revolution. Ex-
3D printing removes the cost barrier of traditional fixed- plainingTheFuture.com.
Beaman, J J. 2013. 3D printing, additive manufacturing, and
equipment manufacturing and the distance barrier raised
solid freeform fabrication: The technologies and applications
by widely distributed suppliers sourced based on cost. From
of the past, present and future. Presentation given at NSF
a time perspective, 3D printing has the potential to reduce Workshop on Frontiers of Additive Manufacturing Research
the time barrier through a tighter coupling of design and and Education, Arlington, VA, July 11–12. http://nsfam.
production in an experimental fashion. Meanwhile, the ca- mae.ufl.edu/Slides/Beaman.pdf
pability to print 3D metal parts opens the door to innova- Boettcher, S. 2013. 3D printing with Windows. [Blog entry,
tions in numerous industries, such as aerospace, medical, June 26.] The Official Microsoft Blog. http://blogs.technet.
and oil and gas. com/b/microsoft_blog/archive/2013/06/26/3d-printing-
Economies of scale and economies of one will continue to with-windows.aspx
coexist, but they will not be used for the same things. Com- Boothroyd, G., Dewhurst, P. and Knight, W. 2002. Product
panies based on economies of scale will still support com- Design for Manufacture and Assembly. New York: Marcel
Dekker.
modity and high-volume production, but in instances where
Brooke, E. 2013. Why 3D printing will work in fashion.
end-user customization is highly desirable, where production
TechCrunch, July 20. http://techcrunch.com/2013/07/20/why-
is single unit or very small volume, or where the end product 3d-printing-will-work-in-fashion/
requires features that cannot be manufactured by traditional Cautela, M. 2013. Staples first major U.S. retailer to announce
means, 3D printing and additive manufacturing will become availability of 3D printers. Press release, May 3. http://investor.
a viable and competitive option. staples.com/phoenix.zhtml?c=96244&p=RssLanding&cat=
The emerging dynamics of economies of one have five news&id=1814995.
likely outcomes: D’Aveni, R. 2013. 3D printing will change the world. Harvard
Business Review 91(3): 34.
1. There will be few clear boundaries in the design-build- De Zeen Magazine. 2013. Microsoft to sell MakerBot 3D printers
deliver paradigm. in American stores. De Zeen Magazine, August 8. http://www.
2. Design and production will be tightly coupled through dezeen.com/2013/08/08/microsoft-to-sell-makerbot-3d-
experimentation. printers-in-american-stores/
3. Competitive advantage will reside in both designs that Dods, B. 2013 Factoring the impact of additive manufacturing:
are simple to manufacture and assemble and designs that A model for university, industry, & government collabora-
are highly customized and complex; the challenge will tion. Presentation given at NSF Workshop on Frontiers of Addi-
be in arenas where manufacturers are seeking simple de- tive Manufacturing Research and Education, Arlington, VA, July
signs, and customers are seeking customized, complex 11–12. http://nsfam.mae.ufl.edu/Slides/Dods.pdf
Economist. 2012. Print me a jet engine. [Blog entry, November
products.
22.] Schumpeter Business and Management, The Economist.
4. Proximity between supplier, manufacturer, and customer http://www.economist.com/blogs/schumpeter/2012/11/
will matter, and localized production will be not only additive-manufacturing
more feasible but more desirable. Gibson, D., Rosen, W. and Stucker, B. 2010. Chapter 18: Busi-
5. Planning will go from long term to real time. ness opportunities and future directions. Additive Manufac-
turing Technologies: Rapid Prototyping to Direct Digital
In the coming decade, economies of one will make competi-
Manufacturing, 437–446. New York: Springer.
tion increasingly uncertain. Gibson, Rosen, and Stucker Koten, J. 2013. Revolution in the making. The Wall Street Jour-
(2010) coined the term digiproneurship to describe the nal, June 11: R1.
transformation of the digital to the physical product through Lipson, H., and Kurman, M. 2013. Fabricated: The New World of
entrepreneurship. In many sectors, manufacturing digipro- 3D Printing. Indianapolis, IN: John Wiley & Sons.
neurship will allow anyone to return to the garage and Luna, T. 2013. 3-D print tech may lead to custom running
make things that satisfy the needs of one or a very few cus- shoes. Boston Globe, April 15. http://www.bostonglobe.com/
tomers. Entry-level 3D printers are now at the same price business/2013/04/14/custom-fit-for-running-shoes/
point as laser printers were when they became desktop fix- urm1RB4Zbvry9dGpGiabEK/story.html
tures, and prices on high-end machines are dropping. Exist- Murray, P. 2012. New at the dentist: 3D printing “dental
ing companies need to understand the challenges of this crowns while you wait.” SingularityHUB, November 7. http://
singularityhub.com/2012/11/07/new-at-the-dentist-
future and begin to think about changes needed in several
3d-printing-dental-crowns-while-you-wait/
key areas. The very resources and practices that have acted
Nanowerk News. 2013. UPS stores to test in-store 3D printing
as barriers to entry for their competitors will become barri- services. Nanowerk, August 4. http://www.nanowerk.com/
ers to change for themselves. This will be particularly true news2/gadget/newsid=31679.php
for companies that have a large installed base and a hierar- Pine, J. B. 1993. Mass Customization: The New Frontier in Busi-
chical organization staffed with highly seasoned employees ness Competition. Cambridge, MA: Harvard Business School
in clearly defined and differentiated roles. Press.

Point of View November—December 2013 | 5

Rawsthorn, A. 2013. Catching up to 3D printing. New York Shinal, J. 2013. New tech economy: 3D printing’s promise in
Times, July 21. http://www.nytimes.com/2013/07/22/arts/ prosthetics. USA Today, March 17. http://www.usatoday.
design/Catching-Up-to-3D-Printing.html?pagewanted= com/story/tech/2013/03/17/autodesk-phillips-electronics-3d-
all&_r=0 printing/1990703/
Senese, M. 2012. Staples announces in-store 3-D printing ser- Ulrich, K., and Eppinger, S. 1995. Product Design and Develop-
vice. CNN Tech, December 1. http://www.cnn.com/2012/ ment. Boston, MA: McGraw-Hill.
11/30/tech/innovation/staples-3-d-printing Weinberg, M. 2010. It Will Be Awesome if They Don’t Screw It
Sharma, R. 2013. The 3D printing revolution you have not Up: 3D Printing, Intellectual Property, and the Fight Over
heard about. [Blog entry, July 8.] Forbes. http://www. the Next Great Disruptive Technology. White paper, Public
forbes.com/sites/rakeshsharma/2013/07/08/the-3d- Knowledge, November. http://publicknowledge.org/it-will-
printing-revolution-you-have-not-heard-about/ be-awesome-if-they-dont-screw-it-up

6 | Research-Technology Management Point of View

 Number 1 of 1
AUTHOR QUERIES
DATE 14/09/2013
JOB NAME RTM
ARTICLE 0193
QUERIES FOR AUTHORS Petrick and Simpson
PLEASE ANSWER THE AUTHOR QUERIES WHERE THEY APPEAR IN THE TEXT.
THERE ARE NO QUERIES
Hastings Law Journal
Volume 65 | Issue 6 Article 2

8-2014

Intellectual Property Infringements & 3D Printing:

Decentralized Piracy
Ben Depoorter

Follow this and additional works at: https://repository.uchastings.edu/hastings_law_journal

Part of the Law Commons

Recommended Citation
Ben Depoorter, Intellectual Property Infringements & 3D Printing: Decentralized Piracy, 65 Hastings L.J. 1483 (2014).
Available at: https://repository.uchastings.edu/hastings_law_journal/vol65/iss6/2

This Symposium is brought to you for free and open access by the Law Journals at UC Hastings Scholarship Repository. It has been accepted for
inclusion in Hastings Law Journal by an authorized editor of UC Hastings Scholarship Repository.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

Intellectual Property Infringements & 3D

Printing: Decentralized Piracy

Ben Depoorter*

By drastically reducing the role of intermediaries in manufacturing, 3D printing is

likely to set about the next wave of decentralized, non-commercial infringements of
intellectual property rights. Drawing upon the lessons from the entertainment
industry’s litigation campaign against illegal file sharing, this paper describes some of
the common characteristics of decentralized piracy. I show that, like copyright
enforcement on file-sharing networks, intellectual property enforcement of 3D
printing faces economic and social norm complications that make traditional,
litigation based enforcement ineffective and possibly counterproductive.

* Hastings Professor of Law, University of California, Hastings College of the Law; Ghent
University; CASLE and Affiliate Scholar, Stanford Law School, Center for Internet & Society.
Contact: [email protected]. I am grateful to Ben Buchwalter, Emily Goldberg Knox, and the
other editors at the Hastings Law Journal for their input and suggestions.

[1483]
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1484 HASTINGS LAW JOURNAL [Vol. 65:1483

Table of Contents
Introduction .............................................................................................. 1484
I. 3D Printing’s Potential for Intellectual Property
Infringements....................................................................................... 1487
II. 3D Printing and Intellectual Property Uncertainty ................ 1489
III. Decentralized Piracy and Enforcement ..................................... 1493
A. The Economics of Enforcing Decentralized
Infringement........................................................................... 1495
B. Decentralized Enforcement and Social Norm
Complications ......................................................................... 1497
Conclusion: Alternatives ...................................................................... 1502

Introduction
If you learn from a loss you have not lost.
—Austin O’Malley, Keystones of Thought 191 (1914).

3D printing technologies promise to revolutionize manufacturing.1

By uploading digital blueprints to computers, users of 3D printers can
create seamless physical objects in an additive process.2 As 3D
technologies become more widespread, the general public will be able to
use cheap, affordable home 3D printers to design and manufacture most
products available on retail markets.3 A growing group of consumers are
already putting 3D printers to use at home to manufacture household
items, small sculptures, and spare parts.4
It is already clear that 3D printing will radically change the role of
intermediaries in manufacturing.5 The declining costs of fabricating
physical items will reduce the necessity for specialization in

1. See Deven R. Desai & Gerard N. Magliocca, Patents, Meet Napster: 3D Printing and the
Digitization of Things, 102 Geo L.J. (forthcoming 2014) (manuscript at 10), available at
http://georgetownlawjournal.org/files/2013/10/Desai-and-Magliocca-3D-Printing-Draft.pdf (“3D printing
will unleash the power of digitized things on manufacturers.”).
2. See Anna Kaziunas France, Skill Builder: 3D Scanning, Make, Winter 2013, at 92; Sean
Buckley, MakerBot’s Digitizer Will Go on Sale Next Week, Promises 3D Scanning to the Masses,
Engadget (Aug. 14, 2013, 8:05 PM), http://www.engadget.com/2013/08/14/makerbots-digitizer-will-go-
on-sale-next-week.
3. Chris Anderson, Makers: The New Industrial Revolution 90–92 (2012); Charles W.
Finocchiaro, Note, Personal Factory or Catalyst for Piracy? The Hype, Hysteria, and Hard Realities of
Consumer 3-D Printing, 31 Cardozo Arts & Ent. L.J. 473, 473 (2013); Daniel Harris Brean, Asserting
Patents to Combat Infringement via 3D Printing: It’s No “Use”, 23 Fordham Intell. Prop. Media &
Ent. L.J. 771, 771 (2013).
4. Neil Gershenfeld, How to Make Almost Anything: The Digital Fabrication Revolution,
91 Foreign Aff. 43, 43–45 (2012).
5. See 3D Printing: The Printed World, Economist, Feb. 12, 2011, at 77; A Factory on Your
Desk, Economist, Sept. 5, 2009, at 26.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1485

manufacturing and the attainment of economies of scale.6 As a result,

manufacturing is no longer synonymous with assembly lines and
processing plants. Consumers can copy, adapt,7 and design items on their
computers and print them out at home, or email blueprint design files to
companies that operate more advanced 3D printers.8
By reducing the costs of designing, manufacturing, and distributing
goods, 3D printing is predicted to set about an “industrial counter-
revolution.”9 Although personal 3D printers might not be well-suited for
mass scale production, they promise to make households largely self-
sufficient.10 As such, 3D printing shifts some core sectors of retailing
away from traditional large manufacturers to the various intermediaries
that facilitate 3D printing, including (but not limited to) individual
designers of blueprints and industries that create and distribute 3D
printing technologies and peripherals.
Revolutionary, disruptive technologies bring about novel legal
challenges.11 Likewise, 3D printing is likely to raise a host of legal
issues.12 Especially intellectual property (“IP”) holders are likely to be
affected because 3D printing makes the infringement of IP rights cheaper
and more attractive.13 While users of 3D printers may end up designing
original products or acquiring licensed blueprints from designers, users

6. Simon Bradshaw et al., The Intellectual Property Implications of Low-Cost 3D Printing, 7

Scripted 5, 11 (2010).
7. On the customization possibilities of 3D printing, see Neil Gershenfeld, Fab: The Coming
Revolution on Your Desktop—From Personal Computers to Personal Fabrication Loc. 500
(Kindle Edition, 2005).
8. E.g., Steven Kurutz, Bringing 3-D Power to the People, N.Y. Times, Mar. 27, 2014, at D1.
9. See, e.g., 3D Printing: The Printed World, supra note 5; Marshall Burns & James Howison,
Digital Manufacturing—Napster Fabbing: Internet Delivery of Physical Products , 7 Rapid
Prototyping J. 194, 196 (2001).
10. See generally Aaron Council & Michael Petch, 3D Printing: Rise of the Third Industrial
Revolution Electronic Book, 2014).
11. On the impact of technological advances on the development of tort law, see, for example,
James E. Krier & Clayton P. Gillette, The Un-Easy Case for Technological Optimism, 84 Mich. L.
Rev. 405, 417–26 (1985) (discussing political obstabcles to the efficient regulation of new technlogies);
Richard A. Posner, A Theory of Negligence, 1 J. Legal Stud. 29, 52–57 (1972) (discussing railroad
crossings); see also Mark F. Grady, Why Are People Negligent? Technology, Nondurable Precautions,
and the Medical Malpractice Explosion, 82 Nw. U. L. Rev. 293 (1988) (demonstrating the different
effects of durable and non-durable prevention technologies on tortious accidents).
12. These issues include potential claims involving products liability and gun-regulation. See
generally Peter Jensen-Haxel, Note, 3D Printers, Obsolete Firearm Supply Controls, and the Right to
Build Self-Defense Weapons Under Heller, 42 Golden Gate U. L. Rev. 447 (2012) (describing how
3D printing renders current firearm regulations obsolete); Nora Freeman Engstrom, 3-D Printing and
Product Liability: Identifying the Obstacles, 162 U. Pa. L. Rev. Online 35 (2013) (discussing
challenges to product liability in 3D printing markets); Mark Gibbs, The End of Gun Control?,
Forbes.com (July 28, 2012, 4:24 PM), www.forbes.com/sites/markgibbs/2012/07/28/the-end-of-gun-
control; Chris Brandrick, 3D Printer Lets You Print Your Own Prescription, PC World (Apr. 19,
2012, 3:59 PM),
http://www.techhive.com/article/254118/3d_printer_lets_you_print_your_own_prescription.html.
13. See Bradshaw et al., supra note 6, at 12.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1486 HASTINGS LAW JOURNAL [Vol. 65:1483

are also likely to print items protected by IP rights without proper

authorization. For example, users might download unauthorized
blueprints online and print unauthorized reproductions of retail products
that violate the exclusive rights of patent, trade dress, or copyright
holders. Such IP infringement requires only the purchase of raw input
print materials14 and the blueprints, which are often freely available
online.15
Although manufacturing and retail industries have always faced
commercial counterfeit markets, the digital revolution of 3D printing
presents an unfamiliar challenge to these industries: highly decentralized
piracy where consumers obtain counterfeit goods cheaply, without
assistance from commercial counterfeiters. In this regard, 3D printing
presents some of the same challenges as the digitization of music, books,
and movies before it.
This shift from traditional commercial counterfeit markets to
decentralized, non-commercial piracy sets the stage for a potential
explosion of IP infringements. IP enforcement against decentralized
infringements is an uphill battle, as the entertainment industry
experienced in the wake of the widespread use of file-sharing
technologies. Commentators already predict that IP enforcement on 3D
printing is a lost cause based on the limited success of copyright
enforcement against file sharing.16
What makes IP enforcement so difficult when infringement is
decentralized? Is IP enforcement a lost cause in the context of 3D
printing? This Article discusses common characteristics of decentralized
infringement and the unique challenges they present for the enforcement
of IP rights. I argue that decentralized piracy creates unique practical, as
well as normative obstacles that render traditional means of IP
enforcement troublesome. The social costs of enforcing IP rights in these
settings are likely to be high, making alternative regulatory approaches
more appealing.
Part I of this Article examines the IP rights affected by 3D printing.
Part II considers the legal uncertainty involved with 3D printing. Part III
describes common characteristics of decentralized piracy and discusses
the main enforcement challenges. This Article concludes with a few
reflections to address the challenges of 3D printing.

14. Printing Material Suppliers, RepRap.org, http://reprap.org/wiki/Printing_Material_Suppliers

(last visited Aug. 1, 2014).
15. Legitimate blueprint download sites might emerge as well. See Gershenfeld, supra note 7, at
Loc. 590–92.
16. Infra Part III.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1487

I. 3D Printing’s Potential for Intellectual Property

Infringements
Producers and distributors of physical objects have generally been
shielded from the problem of widespread, direct infringement online.17
As opposed to literature, music, film, and other types of copyrighted
content, most physical objects cannot be rendered into a digital format
without taking away their utilitarian features. For instance, the
unauthorized pictorial rendition of a designer clock online presents
potential copyright issues relating to copyright in the photo itself. Such
copyright infringement, however, leaves unaffected (and may even
promote) the demand for the actual clock as a useful and/or aesthetic
physical object.
Enter 3D printing. A 3D scanner can create a computer-aided
design (“CAD”) map of the clock.18 This CAD blueprint can be
distributed on the Internet, and anyone with a 3D printer can print an
exact replica of the clock.19 The creation and distribution of
unauthorized reproductions of the clock by way of 3D technology
potentially infringes on the exclusive rights of holders of utility patents,
design patents, and trademarks.20
First and foremost, 3D printing may involve the infringement of
multiple patent rights. When a user of a 3D printer renders the physical
object that is the subject of a patent, the CAD blueprint might infringe
the applicable patent rights on that article. Using a 3D printer to then
reproduce a patented product infringes on the patent right. Moreover,
any use of the printed object may violate the rights on the patent.21
Second, trademark holders will also be affected by 3D printing. Any
sale of a 3D printed object that includes a trademark might infringe on
the rights of the trademark owner. In most instances, the user of the 3D
printer can avoid trademark infringement by reproducing the product
without the logo or trademark. But even on products that do not have a
logo, trademark right violations might take place whenever a user seeks
to print products that are protected by trade dress. Even if the 3D printer
does not seek to include a trademark or the seller does not actively pass
off the item as being manufactured by the owner of the right, it might be
an infringement to sell any printed products in a shape that resembles

17. Of course, illegal markets for counterfeit goods became more accessible online. Yet, as
opposed to digital goods, physical counterfeit goods are relatively expensive to produce and
infringement is less anonymous and more risky, since such illicit markets require physical locations
and addresses to ship from.
18. See France, supra note 2.
19. Peter Schmitt, Works By Peter Schmitt 2000–2014 9 (2014),
http://web.media.mit.edu/~peter/about/Portfolio-Peter-Schmitt.pdf (last visited Aug. 1, 2014).
20. For an overview, see Desai & Magliocca, supra note 1 (manuscript at 2).
21. Id.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1488 HASTINGS LAW JOURNAL [Vol. 65:1483

established trade dress. Such unauthorized third-party uses of

trademarks constitute an infringement, and sales of those products might
run afoul of the post-sale confusion doctrine even if purchasers are not
confused.22
Finally, 3D printing technology also presents challenges to
industries protected by copyright law. Whenever 3D printers are used to
make counterfeit items, this might infringe the copyright on creative
designs that appear on such items. That is, if the decorative features of a
product can be separated, either physically or conceptually, from the
useful aspects of the article,23 then the holder of the copyright in the
aesthetic design has a potential claim against any unauthorized
reproduction.24 Take the example of an individual who reproduces a coat
hanger. The individual might have downloaded the blueprint of the
particular coat hanger because it is the right size for her wall. If, however,
the coat hanger has decorative elements (such as a colorful drawing,
etched designs, etc.) that can be identified separately from the useful
aspects of the coat hanger, the original manufacturer might be able to
bring a claim for copyright infringement for the reproduction of the
decorative design. In the context of 3D printing, a copyright claimant can
pursue a claim of direct infringement against any individual who
reproduced the creative design on a physical object (owners of 3D
printers). Claimants may also pursue claims for indirect infringement on
the basis of contributory or vicarious copyright liability against
manufactures, writers of CAD designs, and other intermediaries that
facilitated the infringing activity.

22. The trademark law post-sale confusion doctrine holds that infringing trade dress can infringe
upon the rights of the owner of a mark even if purchasers of the infringing product are not confused.
See Connie Davis Powell, We All Know It’s A Knock-Off! Re-Evaluating the Need for the Post-Sale
Confusion Doctrine in Trademark Law, 14 N.C. J.L. & Tech. 1, 35 (2012).
23. The 1976 Copyright Act defines pictorial, graphic, and sculptural works. 17 U.S.C. § 101
(2011) (“‘Pictorial, graphic, and sculptural works’ include two-dimensional and three-dimensional
works of fine, graphic, and applied art, photographs, prints and art reproductions, maps, globes, charts,
diagrams, models, and technical drawings, including architectural plans. Such works shall include
works of artistic craftsmanship insofar as their form but not their mechanical or utilitarian aspects are
concerned; the design of a useful article, as defined in this section, shall be considered a pictorial,
graphic, or sculptural work only if, and only to the extent that, such design incorporates pictorial,
graphic, or sculptural features that can be identified separately from, and are capable of existing
independently of, the utilitarian aspects of the article.”). When a work’s useful and aesthetic features
are so intertwined that they cannot be separated physically, courts must consider whether there is
conceptual separability between the form and function of a work. Carol Barnhart Inc. v. Economy
Cover Corp., 773 F.2d 411 (2d Cir. 1985).
24. For a discussion of copyright issues involving 3D printing, see, for example, Brian Rideout,
Printing the Impossible Triangle: The Copyright Implications of Three-Dimensional Printing, 5 J. Bus.
Entrepreneurship & L. 161, 163–64 (2012); Haritha Dasari, Assessing Copyright Protection and
Infringement Issues Involved With 3D Printing and Scanning, 41 AIPLA Q.J. 279 (2013); Edward Lee,
Digital Originality, 14 Vand. J. Ent. & Tech. L. 919 (2012).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1489

3D printers also might be used to produce unauthorized derivate

works. These secondary derivate work markets, such as t-shirts, toys, and
game boards, are a means for copyright holders to recoup some of their
initial investment in an underlying music album, movie, or TV show.25
One might be tempted to suggest that derivate work markets can even
explain some creative decisions.26 3D printers enable fans to scan,
reproduce, and modify these derivate physical products without
authorization. For instance, a fan of the HBO series Game of Thrones,
one of the most illegally downloaded TV shows of all time,27
independently developed and printed a Game of Thrones iPhone
docking station.28 Such unauthorized reproductions are appealing in the
context of 3D printing because the technology enables fans to create or
otherwise obtain customized, altered versions of derivate works. As
Deven R. Desai and Gerard N. Magliocca illustrate:
Today, if you buy a doll, a Lego set, or a car, the ability to alter, tinker,
or improve your purchase is low. 3D printing, however, opens the door
to personal improvement. You still buy the doll or dollhouse; but once
a child is bored, 3D printing allows you to design and create new heads,
limbs, or furniture. Instead of relying on Lego to decide what a piece
looks like or does, the consumer can make new ones.29
In this process, 3D printing potentially disrupts the secondary markets
that help sustain copyright holders in today’s world of illegal music and
movie file sharing, where illegal digital downloading has already eroded
revenues from record and DVD sales.30

II. 3D Printing and Intellectual Property Uncertainty

Innovation is often rapid and unpredictable. Legal systems take
time to adapt to new technological advancements. Lawmaking is a
complex process that involves planning, procedures, and the

25. Take the example of Game of Thrones. Reporting on a recent infringement issue, HBO Vice
President of Corporate Affairs Jeff Cusson commented that “with a show like Game of Thrones, the
amount of product licensed around . . . is exorbitant.” Nathan Hurst, HBO Blocks 3-D Printed Game
of Thrones iPhone Dock, Wired (Feb. 13, 2013, 1:57 PM), http://www.wired.com/design/2013/02/got-
hbo-cease-and-desist.
26. For example, a new generation of toy buyers could help explain the Star Wars and Lego
movies, etc.
27. Ernesto, ‘Game of Thrones’ Most Pirated TV-Show of 2013, TorrentFreak (Dec. 25, 2013),
http://torrentfreak.com/game-of-thrones-most-pirated-tv-show-of-2013-131225 (“Game of Thrones has
the honor of becoming the most downloaded TV-show for the second year in a row.”); Kory Grow,
‘Game of Thrones,’ ‘Breaking Bad’ Among the Most Pirated Shows of 2013, Rolling Stone (Dec. 26,
2013, 2:50 PM), http://www.rollingstone.com/movies/news/game-of-thrones-breaking-bad-among-the-
most-pirated-shows-of-2013-20131226.
28. See Hurst, supra note 25.
29. See Desai & Magliocca, supra note 1 (manuscript at 12).
30. See, e.g., Stan J. Liebowitz, Testing File-Sharing's Impact by Examining Record Sales in
Cities, 54 Mgmt. Sci., 852 (2008) (presenting empirical evidence that file sharing is responsible for
historical decline in record sales).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1490 HASTINGS LAW JOURNAL [Vol. 65:1483

participation of various stakeholders. The dynamic and unpredictable

nature of technological innovation makes it difficult for lawmakers to
predict or anticipate forthcoming inventions. As a result, courts and
legislators have a difficult time responding proactively to avoid delays
between the time people begin to use a technology and its legal
classification (including the allocation of liability). Additionally, many
areas of law apply open-ended standards. While standards reduce the
cost of errors and enable copyright decisionmakers to be more flexible,
these open-ended standards increase the number of difficult questions
that courts must confront.31
The initial ambiguity as to the potential social and economic
implications of a novel technology also contributes to the gap between
the use of a technology and its legal classification by courts and
legislators. It must first become apparent that the use of novel
technology entails substantial opportunity costs to producers—that is,
that there are “gains to be internalized.”32 Not until the opportunity costs
of unregulated use of 3D printers become fully clear will manufacturers
push to obtain protection through litigation and legislation. Until then,
manufactures and users of 3D printers will continue to operate in a
vacuum of considerable legal uncertainty.33 Consider, for example, the
difficulty of perfectly predicting ex ante how courts will apply the law to
new circumstances ex post.34

31. On the distinction between rules and standards, see, for example, Hans-Bernd Schäfer, Rules
Versus Standards in Rich and Poor Countries: Precise Legal Norms as Substitutes for Human Capital
in Low-Income Countries, 14 Sup. Ct. Econ. Rev. 113, 113 (2006) (“Legal norms can be precise rules,
which are blueprints for action and allow for mechanical decisions by judges and civil servants.
Alternatively, they can be vague, mission-oriented standards, which delegate decisions from the maker
of the law to the judiciary and the administration. Rules economize on the costs of adjudication and
administration. Standards economize on the costs of norm specification.”); Louis Kaplow, Rules
Versus Standards: An Economic Analysis, 42 Duke L.J. 557 (1992); Pierre Schlag, Rules and
Standards, 33 UCLA L. Rev. 379 (1985).
32. On the evolution of IP rights, see Ben Depoorter, The Several Lives of Mickey Mouse: The
Expanding Boundaries of Intellectual Property Law, 9 Va. J.L. & Tech. 1, 34–41 (2004) [hereinafter,
Depoorter, Several Lives]; Ben Depoorter, Technology and Uncertainty: The Shaping Effect on
Copyright Law, 157 U. Pa. L. Rev. 1830, 1841–42 (2009) [hereinafter, Depoorter, Technology and
Uncertainty] (“The initial ambiguity of the socioeconomic implications of a new technology can be
illustrated, for example, by peer-to-peer music exchanges. The music industry discovered that huge
profits could be made by delivering music in a compressed format (MP3) only after such exchanges
were already relatively common.”). See Peter S. Menell, Envisioning Copyright Law’s Digital Future,
46 N.Y.L. Sch. L. Rev. 63, 99 (2002) (“Even with the introduction and rapid popularity of digitally-
encoded compact disks (CDs) and the proliferation of microcomputers beginning in the early 1980s,
the record industry did not appreciate the dramatic changes that would be brought about by the
emerging digital technologies.”).
33. One current definition of legal uncertainty is that it describes a situation where an act is “said
by informed attorneys to have an expected official outcome at or near the 0.5 level of predictability.”
Anthony D’Amato, Legal Uncertainty, 71 Calif. L. Rev. 1, 2 (1983).
34. Note that there is a distinction between risk and uncertainty. Individuals are subject to risk if
(1) an event may or may not happen in the future, and (2) the chance that the event will happen is
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1491

Although uncertainty pervades all areas of the law,35 technological

breakthroughs, by their nature, make it more difficult to apply judicial
precedent by analogy.36 This is certainly the case for 3D printing. As
commentators have noted, 3D printers, like file sharing before them,
present a series of novel legal questions that require courts to stretch
existing doctrines to a radically new technology. IP rights holders will
face hurdles in cabining new uses of 3D printers into existing IP
doctrines.37 For patent right holders, 3D printing presents novel legal
issues pertaining to infringement. When pursuing claims against
individual users of 3D printers, a patent right holder will need to make
the case that the creation and distribution of CAD files is the legal
equivalent of a use or sale of the underlying patented invention.38 When
suing manufacturers of the printers or providers of the CAD/Computer
Aided Manufacturing (“CAM”) files based on of patent law’s indirect
infringement doctrine, patent holders will need to meet the difficult
burden of proving that the alleged infringer possessed actual knowledge
of a specific, infringed patent.39 As a result, although “consumer use of
3D printers may create multiple instances of patent infringement,
policing and protecting patent rights in inventions copied on 3D printers
may present significant challenges for patent holders.”40
Similarly, trademark owners face a number of obstacles that hinder
effective enforcement of their rights. First, trademark owners will find it
difficult to enforce their rights as against 3D printouts of trademarked
goods when the actual logo or trademark is removed from the printed
product, while owners of product configurations will always need to

known. By contrast, an event is uncertain if (1) it may or may not happen in the future, and (2) we do
not know the chances that it will happen. See generally Frank H. Knight, Risk, Uncertainty and
Profit (Harper & Row 1965) (1921).
35. On legal uncertainty generally, see D’Amato, supra note 33 (describing a trend toward
greater uncertainty); Isaac Ehrlich & Richard A. Posner, An Economic Analysis of Legal Rulemaking,
3 J. Legal Stud. 257 (1974) (examining the optimal level of precision for rules and standards); Werner
Z. Hirsch, Reducing Law’s Uncertainty and Complexity, 21 UCLA L. Rev. 1233 (1974) (discussing
considerations involved in, and obstacles to, reducing uncertainty); Jason Scott Johnston, Uncertainty,
Chaos, and the Torts Process: An Economic Analysis of Legal Form , 76 Cornell L. Rev. 341 (1991)
(examining how rules and balancing approaches evolve out of litigation).
36. Even what looks straightforward now in hindsight is often anything but obvious at the time a
new technology emerges. File sharing, for instance, challenged our understanding of piracy: many
direct infringements lack a commercial purpose, and there are no conventional intermediaries.
37. Infra Part III.
38. For an in-depth discussion of the doctrinal issues involving 3D printing, see Brean, supra note 3.
39. See Finocchiaro, supra note 3 (noting that the substantial non-infringing uses of 3D printing
bring the technology closer to the Sony precedent than Grokster).
40. Bryan J. Vogel, IP: 3D Printing and Potential Patent Infringement, Inside Counsel (Oct. 29, 2013),
http://www.insidecounsel.com/2013/10/29/ip-3d-printing-and-potential-patent-infringement.html.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1492 HASTINGS LAW JOURNAL [Vol. 65:1483

demonstrate that their trade dress has acquired secondary meaning.41

Moreover, trademark plaintiffs will need to demonstrate that the alleged
infringer has “used the mark in commerce,” as opposed to having
engaged in a merely personal use.42 On the copyright front, most
instances of copyright infringement involving 3D printing will force
courts to concentrate on an area of law that is mired in doctrinal
confusion and uncertainty: the conceptual separation of useful and
creative aspects of a product.43 In order to determine whether separate
copyrightable features are present, courts apply opposing artistic theories
of interpretation, often mixing various incompatible theories together
within one decision.44
It is difficult to predict ex ante how courts or legislators will
categorize some of the potential uses of 3D printers. The resulting
ambiguity provides ample opportunity for owners of 3D printers to
engage in self-serving interpretations and to convince themselves that no
infringement has been committed.45 As research in the field of cognitive
psychology demonstrates, individuals are inclined to construct facts in
ways that align with their own preconceived beliefs.46 Just as was the case

41. While simple trade dress can be “inherently” distinctive, valid product-design trade dress
must acquire a distinct association with a specific manufacturer in the public’s mind. Wal-Mart Stores,
Inc. v. Samara Bros., 529 U.S. 205, 215 (2000).
42. 15 U.S.C. § 114 (2011). For a discussion of trademark issues relating to 3D printing, see
Michael Weinberg, Public Knowledge, It Will Be Awesome If They Don’t Screw It Up: 3D
Printing, Intellectual Property, and the Fight over the next Great Disruptive Technology 1, 3
(2010); Desai & Magliocca, supra note 1 (manuscript at 34).
43. For a discussion of the doctrinal confusion among courts, see, for example, Ben Depoorter &
Robert Kirk Walker, The Dangerous Undertaking: How Courts Should Approach Aesthetic
Judgments in Copyright Law, 108 Nw. U. L. Rev. (forthcoming 2014) [hereinafter Depoorter &
Walker, Dangerous Undertaking]; Alfred C. Yen, Copyright Opinions and Aesthetic Theory, 71 S.
Cal. L. Rev. 247 (1998); Christine Haight Farley, Judging Art, 79 Tul. L. Rev. 805 (2005).
44. For instance, the Second Circuit held that the useful aspects of decorative belt buckles could
be sufficiently separated from their ornamental, aesthetic aspects; the Second Circuit applied all three
major, but internally incompatible, theories of artistic interpretation. See Depoorter & Walker,
Dangerous Undertaking, supra note 43 (“In finding that the buckles in question ‘rise to the level of
creative art,’ the court referred to the intention of the author, but also noted that the buckles were well
received in art and fashion circles. Institutional interpretations surfaced as well when the court
rejected the notion that the utilitarian nature of fashion items excludes such articles from copyright
protection: ‘body ornamentation has been an art form since the earliest days, as anyone who has seen
the Tutankhamen or Scythian gold exhibits at the Metropolitan Museum will readily attest.’” (quoting
Kieselstein-Cord v. Accessories by Pearl, Inc., 632 F.2d 989, 990 (2d Cir. 1980)).
45. See generally Depoorter, Technology and Uncertainty, supra note 32 (exploring the enabling
effect of legal uncertainty and delay on norm formation processes generally and in the context of peer-
to-peer file sharing).
46. See Linda Babcock & George Loewenstein, Explaining Bargaining Impasse: The Role of
Self-Serving Biases, 11 J. of Econ. Perspectives 109, 113 (1997) (authors assigned participants in a
study to either the plaintiff or defendant in a hypothetical automotive accident tort case with a
maximum potential damages payment of $100,000. The plaintiff’s prediction of the likely judicial
award was on average $14,527 higher than the defendant’s. The plaintiff’s average nomination of a
“fair” figure was $17,709 higher than the defendant’s). See generally Donald Braman and Dan M.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1493

with file sharing,47 by the time that courts and legislators have ruled on
the legal effects of many uses of 3D printers, many users will no longer
be neutral bystanders. Users of 3D printers will likely form beliefs and
attitudes that support liberal uses of 3D printers and will reject legal
reform to the contrary.48 Users of 3D printers might experience loss
aversion when what they consider to be legitimate is suddenly found to
be illegal. In this process, the perception of having something “taken
away” might add to the resistance we can expect when IP rights will be
enforced on products of 3D printing.49
In the meantime, while courts begin to examine a fit between
existing IP laws and 3D printing, the legal ambiguities will foster 3D
printing activities without much consideration for IP laws. Moreover, the
decentralized nature of infringement in a 3D setting adds another
complication to the enforcement of IP rights.

III. Decentralized Piracy and Enforcement

Broadband networks, digital music files, and peer-to-peer file-
sharing applications widely expanded access to unlicensed copies of
copyrighted material. Organizations such as the Recording Industry
Association of America (“RIAA”), the Motion Picture Association of
America (“MPAA”), and the Entertainment Software Association
(“ESA”) argue that file sharing is responsible for significant declines in
the sale of music, DVDs, and videogames.50 Over the past decade the
entertainment industry worked hard to combat online infringement of

Kahan, More Statistics, Less Persuasion: A Cultural Theory of Gun-Risk Perceptions, 151 U. Pa. L.
Rev. 1291 (2003) (describing culturally biased interpretations of data).
47. See Depoorter, Several Lives, supra note 32 (describing cyclical process of technology, norm
adaptation and judicial determination).
48. On the internalization of social norms, see Robert Cooter, Three Effects of Social Norms on
Law: Expression, Deterrence, and Internalization, 79 Or. L. Rev. 1 (2000); Robert Cooter, Do Good
Laws Make Good Citizens? An Economic Analysis of Internalized Norms, 86 Va. L. Rev. 1577 (2000).
49. Experimental research demonstrates that individuals value certain resources or legal
entitlements that they possess more than they would value the exact same thing had they never
possessed it at all. See, e.g., Daniel Kahneman, et al., Experimental Tests of the Endowment Effect
and the Coase Theorem, 98 J. of Pol. Econ. 1325 (1990); George Lowenstein & Samuel Issacharoff,
Source Dependence in the Valuation of Objects, 7 J. of Behav. Decision Making 157 (1994).
50. For empirical evidence, see, for example, Stan J. Liebowitz, File Sharing: Creative
Destruction or Just Plain Destruction?, 49 J.L. & Econ. 1, 14–17 (2006) (presenting evidence that file
sharing reduced recording industry revenues); Rafael Rob & Joel Waldfogel, Piracy on the High C’s:
Music Downloading, Sales Displacement, and Social Welfare in a Sample of College Students , 49 J.L.
& Econ. 29, 30 (2006) (downloading reduced purchases by individuals in their sample by about ten
percent during 2003). For an overview, see Stan J. Liebowitz, Economists Examine File Sharing and
Music Sales, in Industrial Organization and the Digital Economy 145 (Gerhard Illing & Martin
Peitz eds., 2006). But see Felix Oberholzer-Gee & Koleman Strumpf, The Effect of File Sharing on
Record Sales: An Empirical Analysis, 115 J. Pol. Econ. 1, 3 (2007) (finding no negative impact of file
sharing on CD sales). On video game piracy, see Ben Depoorter, What Happened to Video Game
Piracy?, Comm. of the ACM (Ass’n for Computing Mach., New York, N.Y.), May 2014, at 33.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1494 HASTINGS LAW JOURNAL [Vol. 65:1483

copyrighted content on file-sharing networks and BitTorrent sites. The

music industry has waged a litigation campaign on various fronts,
targeting individual end users of file-sharing technologies,51 as well as
software developers and various other intermediaries, such as Internet
service providers and website hosts.52 These enforcement efforts have
failed to create a satisfactory reduction in file sharing for the industry. By
most accounts, file sharing and massive online copyright infringement
continue unabated.53 In 2012, the industry decided to abandon
deterrence strategies in favor of more cooperative models,54 while
continuing to pursue favorable legislation.55
As a practical matter, 3D printing technologies enable
decentralized, mainstream piracy. 3D printing fundamentally alters the
production function of piracy because it enables consumers to obtain
counterfeit goods cheaply, without assistance from commercial

51. For a discussion of the RIAA litigation campaign against individual file sharers, see Ben
Depoorter et al., Copyright Backlash, 84 S. Cal. L. Rev. 1251 (2011) [hereinafter Depoorter et al.,
Copyright Backlash]. For an overview of RIAA litigation actions, see, for example, Recording
Industry Association of America Case Activity from Lexis/Nexis Courtlink, Lexis Nexis,
http://www.lexisnexis.com/trial/nalm100181clinkriaa.asp (last visited Aug. 1, 2014); infra note 60.
52. See, e.g., In re Aimster Copyright Litig., 334 F.3d 643, 646 (7th Cir. 2003); A&M Records,
Inc., v. Napster, Inc., 239 F.3d 1004, 1013 (9th Cir. 2001) (“Plaintiffs claim Napster users are engaged
in the wholesale reproduction and distribution of copyrighted works, all constituting direct
infringement.”); Metro-Goldwyn-Mayer Studios, Inc. v. Grokster, Ltd., 259 F. Supp. 2d 1029 (C.D.
Cal. 2003) (plaintiff claiming contributory and vicarious copyright infringement on behalf of producers
of file-sharing application), aff’d, 380 F.3d 1154 (9th Cir. 2004), vacated, 545 U.S. 913 (2005). For
Internet service providers, see, for example, In re Verizon Internet Servs., Inc. 240 F. Supp. 2d 24, 26
(D.D.C. 2003) (copyright holders seeking enforcement of subpoenas demanding the identities of
copyright infringers), rev’d, 351 F.3d 1229 (D.C. Cir. 2003).
53. Press Release: BitTorrent and µTorrent Software Surpass 150 Million User Milestone;
Announce New Consumer Electronics Partnerships, BitTorrent (Jan. 9, 2012),
http://www.bittorrent.com/intl/es/company/about/ces_2012_150m_users (“BitTorrent Mainline and
µTorrent software clients have grown to over 150 million monthly active users worldwide.”); Ernesto,
BitTorrent Accounts for 35%of All Upload Traffic, VPNs are Booming, TorrentFreak (May 18, 2013),
http://torrentfreak.com/bittorrent-accounts-for-35-of-all-upload-traffic-vpns-are-booming-130518 (“New
data published by the Canadian broadband management company Sandvine reveals that BitTorrent
can be credited for one third of all North American upload traffic during peak hours. BitTorrent usage
also remains strong in Europe, Latin America and Asia Pacific. The report further confirms that SSL
traffic has more than doubled in a year, partly due to an increase in VPN use.”).
54. On the U.S. Copyright Alert system, see, for example, Abigail Phillips, The Content Industry
and ISPs Announce a “Common Framework for Copyright Alerts”: What Does it Mean for Users?,
Electronic Frontier Found. (July 7, 2011), https://www.eff.org/deeplinks/2011/07/content-industry-
and-isps-announce-common; Chris Morran, COMCASTIC 3 Report: Comcast Sends Out Around
1,800 Copyright Alert Notices Each Day, Consumerist (Mar. 6, 2014),
http://consumerist.com/2014/02/07/report-comcast-sends-out-around-1800-copyright-alert-notices-
each-day.
55. Past attempts include, for example, the Piracy Deterrence and Education Act of 2004, H.R.
4077, 108th Cong. (2004) (bolstering copyright enforcement on the Internet); Author, Consumer, and
Computer Owner Protection and Security Act of 2003, H.R. 2752, 108th Cong. (2003) (expanding
domestic and international copyright enforcement); Piracy Deterrence and Education Act of 2003,
H.R. 2517, 108th Cong. (2003) (expanding the authority of the government to seize pirated works).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1495

counterfeiters. In this regard, 3D printing shares common ground with

the digitization of music, books, and movies before it. The technology
facilitates massive direct, non-commercial infringement of IP rights by
individual consumers. In other words, 3D printing brings the “Napster”
revolution of decentralized infringement to the doorstep of
manufacturers and retailers of physical articles.56 As a result,
commentators predict that 3D printing infringement will devalue IP
rights and that “even the best efforts to stop this surge in infringement
will fall short.”57 The assumption is that 3D printing creates similar
infringement dynamics that have made the enforcement of copyright law
for file sharing so difficult. The next Subpart of this Article examines this
assumption and analyzes the common characteristics of decentralized
piracy. Drawing upon the lessons from the entertainment industry’s
litigation campaign against file sharing, it is clear that IP litigation is not
likely to provide an effective countermeasure to 3D printing copyright
infringement.

A. The Economics of Enforcing Decentralized Infringement

3D printing drastically reduces the role of intermediaries in
manufacturing. At the same time, it also reduces the potential role of
intermediaries in the production of counterfeit goods and other piracy
related activities. Individuals that possess personal 3D printers can print
infringing materials inside their home with hardly any outside assistance.
Whoever owns a 3D printer and a 3D scanner can simply scan and copy
any existing product in their possession. Worse, 3D printer users can
download blueprint CAD files of existing products online and print
physical copies at home. Although printing larger items may require the
involvement of intermediaries operating commercial 3D printers,
consumers can manufacture regular-size counterfeit items. As with peer-
to-peer file sharing and music copyright before it, counterfeit piracy
becomes a mainstream, non-commercial activity in a world of 3D
printing.
The novel, decentralized nature of 3D piracy has a profound effect
on the enforcement of IP rights as they relate to 3D printing. To the
extent that third parties, such as designers of blueprints or distributors of
CAD files, are involved with facilitating 3D printing IP infringements,
these third parties may not be held liable. As was clarified in the context

56. See Davis Doherty, Downloading Infringement: Patent Law as a Roadblock to the 3D
Printing Revolution, 26 Harv. J.L. & Tech. 353, 355 (2012) (“3D printing also enables widespread
patent infringement in the form of digital downloads in much the same manner that the advent of
digital music enabled widespread copyright infringement.”); Brean, supra note 3.
57. Desai & Magliocca, supra note 1 (manuscript at 4). But see Finocchiaro, supra note 3 (arguing
that physical and technological limitations make 3D printing less likely to threaten IP rights than peer-
to-peer and file-sharing technologies before it).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1496 HASTINGS LAW JOURNAL [Vol. 65:1483

of decentralized file-sharing technologies, contributory or vicarious

liability requires a certain level of control or involvement in enabling
infringing uses of the technology that might not be present in this
context.58 Moreover, even if creators and distributors of blueprint
designs can be held liable, enforcement will be difficult as a practical
matter. The creation and distribution of blueprint designs do not
necessitate substantial investments. Rather, these activities may be
driven by intrinsic and social motivations, reputation, and informal
relationships. In such online, user-driven, and peer-production
environments, the absence of a central, hierarchical residual claimant
complicates case enforcement because it is more difficult to target
offenders whose sole presence is online and whose purpose is not
commercial.
The reduced involvement of intermediaries drastically increases the
economic costs of enforcement overall. If no intermediary can be held
liable, monitoring and enforcement cannot be delegated onto these third-
party intermediaries. As a result, instead of pursuing claims against a few
intermediaries, rights holders must file claims against individual
infringers.59
When facing decentralized enforcement, IP rights holders thus face
a daunting challenge. First, since most infringement occurs inside private
homes, there is a greater perception of safety and anonymity with
unauthorized 3D printing than when purchasing illegal goods in markets
or online using a credit card. Like music and movie downloading on
peer-to-peer networks, most infringement will be difficult to detect.
Second, since infringement is easy and relatively cheap, affordable
3D printers are likely to generate a great number of infringements. Most
users of 3D printers will realize that the probability of being caught is
remote, given the vast amount of infringing uses that occur at any given
moment.

58. For instance, in the context of indirect liability of providers, courts rejected the application of
Napster to decentralized file-sharing services because liability for contributory infringement implies
“actual knowledge of infringement at a time when [file-sharing services] can use that knowledge to
stop the particular infringement.” Metro-Goldwyn-Mayer Studios, Inc. v. Grokster, Ltd.,
259 F. Supp. 2d 1029, 1037 (C.D. Cal. 2003), aff’d, Metro-Goldwyn-Mayer Studios, Inc. v. Grokster,
Ltd., 380 F.3d 1154 (9th Cir. 2004). Additionally, 3D print technologies are capable of substantial non-
infringing uses. See Sony Corp. of America v. Universal City Studios, Inc., 464 U.S. 417 (1984)
(protecting time shifting of television recordings as substantially non-infringing use).
59. On the economics of gate-keeper liability in the context of copyright infringement, see
Douglas Lichtman & William Landes, Indirect Liability for Copyright Infringement: An Economic
Perspective, 16 Harv. J.L. & Tech. 395, 396 (2003) (“The argument in favor of [indirect] liability is that
third parties are often in a good position to discourage copyright infringement either by monitoring
direct infringers or by redesigning their technologies to make infringement more difficult.”); William
Landes & Douglas Lichtman, Indirect Liability for Copyright Infringement: Napster and Beyond,
17 J. Econ. Persp. 113 (2003). See generally Reinier Kraakman, Third-Party Liability, in Palgrave
Dictionary of Economics and the Law 583 (Peter Newman ed., 1998).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1497

Third, as the amount of infringement increases, the probability of

getting caught reduces further for every potential infringer. Because the
rights holders’ resources to pursue IP violations are limited, as the
number of infringers increases, each individual infringer’s chance of
being caught decreases. This, in turn, will likely lower inhibitions against
producing counterfeit items on 3D printers even further.
Unless IP rights holders can easily obtain the identities of infringers
and file class-action lawsuits, something that has so far eluded copyright
owners in the context of file sharing,60 IP rights holders will not have the
resources to pursue most infringement. Efforts by rights holders to
reduce the costs of enforcement will likely create new legal questions
involving a wide array of issues relating to privacy, procedural issues such
as class certifications, subpoena requests, and the involvement of
Internet service providers.61 It will take time for these questions to work
their way through the courts or Congress, as was the case with peer-to-
peer litigation concerning many other revolutionary technologies.
Meanwhile, the legal uncertainty and delay sets up an environment
conducive to infringing behavior.
In order to have a deterrent effect and create a perception of
effective enforcement, IP rights holders might resort to aggressive tactics:
strike hard and set salient examples that highlight the dangers of
infringing patent, trademark, or copyrights by way of unauthorized 3D
printing. In doing so, however, overly aggressive enforcement might
undermine public support for IP rights.62

B. Decentralized Enforcement and Social Norm Complications

By resorting to aggressive tactics, IP rights holders face the risk of
undermining public support for the very rights that they are seeking to
protect. The punitive litigation campaign of the RIAA, which included
mass settlement demand letters and large statutory damage awards,
illustrates the public relations downside to heavy-handed enforcement in
a decentralized infringement environment. First, due to the large number
of infringers, any instance of individual enforcement appears random.
Targeting a limited number of infringers in an effort to set an example
risks creating the perception that a few individuals are arbitrarily being
singled out in a negative litigation lottery. This is one of the lessons from
the peer-to-peer litigation campaign. When the music industry attempted

60. On the different stages of the copyright enforcement campaign and the various setbacks, see
RIAA v. The People: Five Years Later, Electronic Frontier Found. (Sept. 30, 2008),
https://www.eff.org/wp/riaa-v-people-five-years-later.
61. Id.
62. Infra Part III-B.
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1498 HASTINGS LAW JOURNAL [Vol. 65:1483

to set examples,63 the resulting statutory damage awards were widely

condemned as arbitrary and excessive, setting in motion proposals to
reform copyright law.64
Alternatively, IP holders might ramp up the scale of enforcement.
For instance, like copyright holders today, rights holders might try to
stem the distribution of CAD files, and target distributors by way of
“algorithmic enforcement,” where computer programs (“bots”) scour the
Internet looking for content that bears the markings of an
infringement.65 By using bots rather than human spotters, a broader
range of infringing activities can be detected at far less cost than is
required for manual enforcement. But economizing enforcement efforts
has drawbacks. Invariably, such automated enforcement efforts are more
prone to errors66 and inadvertent public relation mistakes.67 Moreover,
mechanical enforcement operations (like the RIAA’s mass settlement
campaign) are more likely to be perceived negatively, as is the case with
formal litigation business models associated with trolling.68
Overall, as rights holders ramp up enforcement, the negative
perception is likely to increase, which could be detrimental to the
interests of IP rights holders. If the public perceives enforcement to be

63. Following 17 U.S.C. § 504 (2006), a prevailing plaintiff may recover statutory damages
between $750 and $30,000 per copyrighted work. In the case of willful infringement by the defendant,
damages of up to $150,000 per work may be recovered. Statutory damages were awarded in two peer-
to-peer cases. Capitol Records, Inc. v. Thomas-Rasset, 692 F.3d 899, 902 (8th Cir. 2012) (upholding
damages of $222,000 for the infringement of twenty-four songs by a single mother); Sony BMG Music
Entm’t v. Tenenbaum, 721 F. Supp. 2d 85, 116–17 (D. Mass. 2010), aff’d in part, rev’d in part, 660 F.3d
487, 515 (1st Cir. 2011) (upholding the constitutionality of statutory damages for copyright violations
and remanding for reconsideration of the remittitur motion), on remand, 103 U.S.P.Q.2d 1902
(D. Mass. 2012) (reinstating initial jury award of statutory damages, imposing $22,500 per song shared
by Boston University graduate student on peer-to-peer network).
64. See generally Pamela Samuelson & Tara Wheatland, Statutory Damages in Copyright Law:
A Remedy in Need of Reform, 51 Wm. & Mary L. Rev. 439 (2009) (discussing copyright damages
reform).
65. See Geeta Dayal, The Algorithmic Copyright Cops: Streaming Videos Robotic Overlords,
Wired (Sept. 6, 2012, 6:00 AM), http://www.wired.com/threatlevel/2012/09/streaming-videos-robotic-
overlords-algorithmic-copyright-cops/all (describing automated copyright systems that search for
copyrighted material in real time).
66. See generally Ben Depoorter & Robert Kirk Walker, Copyright False Positives, 89 Notre
Dame L. Rev. 319 (2013) (describing sources and consequences of copyright enforcement mistakes ).
67. For instance, media outlets reported that the RIAA filed a claim against a twelve-year-old
New York girl whose mother lived in a low-income area of New York City, and had accused an eighty-
three-year-old woman who had died over a month earlier. See John Borland, RIAA Settles with 12-
Year-Old Girl, CNET News (Sept. 9, 2003, 4:05 PM), http://news.cnet.com/2100-1027-5073717.html;
Andrew Orlowski, RIAA Sues the Dead, The Register (Feb. 5, 2005), http://www.theregister.co.uk/
2005/02/05/riaa_sues_the_dead.
68. James DeBriyn, Shedding Light on Copyright Trolls: An Analysis of Mass Copyright
Litigation in the Age of Statutory Damages, 19 UCLA Ent. L. Rev. 79, 79 (2012) (“To supplement
profits from copyrighted works, copyright holders have devised a mass-litigation model to monetize,
rather than deter, infringement . . . utiliz[ing] the threat of outlandish damage awards to force alleged
infringers into quick settlements.”).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1499

excessive, this might reinforce or strengthen a belief that the legal regime
is not legitimate or that a legal rule is unjust.69 This is likely what
happened in the wake of the highly publicized litigation campaign of the
record industry. The public perceived that the industry was targeting
college students in a settlement extortion campaign because the
proposed $3000 settlement was always preferable to the individual’s
financial exposure in litigation.70 Media reports that a single mother and
a college student were each charged with six-figure damage awards for
copyright infringements involving a handful of songs71 likely undermined
the public support for the enforcement of copyright law and might have
contributed to a public belief that the campaign against file sharing was
excessive.72
How much should such negative public attitudes be of concern to
content rights holders? Can rights holders afford to ignore public
backlash? What if increasing deterrence is the only way to discourage
infringing behavior?73
First, social science scholarship warns against purely coercive
approaches to enforcement. For instance, experiments show that
individuals tend to obey rules when they believe that it is the right thing
to do.74 For that reason, it may matter a great deal whether the law or its

69. See generally Francesco Parisi & Georg Von Wangenheim, Legislation and Countervailing
Effects from Social Norms, in Evolution and Design of Institutions 25 (Christian Schubert & Georg
Von Wangenheim eds., 2006), available at http://www.law.gmu.edu/assets/files/publications/
working_papers/04-31.pdf (providing formal mathematical model of norm backlash effects).
70. Grant Gross, Congress Scrutinizes RIAA Tactics, IDG News (Sept. 17, 2003),
http://www.pcworld.com/article/112535/article.html; Katie Dean, Senator Wants Answers From
RIAA, Wired (Aug. 1, 2003), http://archive.wired.com/politics/law/news/2003/08/59862; J. Cam
Barker, Note, Grossly Excessive Penalties in the Battle Against Illegal File-Sharing: The Troubling
Effects of Aggregating Minimum Statutory Damages for Copyright Infringement, 83 Tex. L. Rev. 525,
526 (2004); Daniel Reynolds, Note, The RIAA Litigation War on File Sharing and Alternatives More
Compatible with Public Morality, 9 Minn. J.L. Sci. & Tech. 977, 978–87 (2008). See generally Pamela
Samuelson & Ben Sheffner, Debate, Unconstitutionally Excessive Statutory Damage Awards in
Copyright Cases, 158 U. Pa. L. Rev. 53 (2009) (discussing the constitutionality of copyright statutory
damage awards).
71. See supra text accompanying note 63.
72. Depoorter et al., Copyright Backlash, supra note 51.
73. See generally George J. Stigler, The Optimum Enforcement of Laws, 78 J. Pol. Econ. 526
(1970) (theory of rational law enforcement); Richard A. Posner, Economic Analysis of Law (5th ed.
1998) (explaining fundamental concepts of the economic approach to law).
74. For instance, in the context of tax compliance, extensive literature suggests that social
motivations (ethical concerns, social norms, perceptions of fairness, etc.) can be stronger determinants
of taxpaying behavior than material considerations. See Michael Wenzel, Motivation or
Rationalization? Causal Relations Between Ethics, Norms and Tax Compliance, 26 J. Econ. Psychol.
491, 492 (2005); see also John S. Carroll, Compliance with the Law: A Decision-Making Approach to
Taxpaying, 11 Law & Hum. Behav. 319, 319–35 (1987) (applying decisionmaking models to tax law);
Simon James, et al., Developing a Tax Compliance Strategy for Revenue Services, 55 Bull. for Int’l
Fiscal Documentation 158–64 (2001). In fact, a number of empirical studies find that norms and
beliefs are a stronger determinant of compliance than deterrence. For an overview, see Leandra
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1500 HASTINGS LAW JOURNAL [Vol. 65:1483

enforcement is perceived as being “just” or “fair.” Conversely,

individuals might decide not to obey a legal command if the rule is
considered “unjust.” In this process, normative intuitions about morality
might cause individuals to set aside the risks associated with the illegal
behavior.
Second, when behavior is driven by normative viewpoints,
unbalanced enforcement efforts might reinforce and strengthen the
underlying opposition against the applicable laws.75 This effect is
particularly strong when law enforcement conflicts with social norms or
personal beliefs of what should be allowed. In the context of file sharing,
the strongest opponents of enforcement were the most frequent file
sharers. Frequent infringers, of course, had the most to lose from
stringent enforcement, so their personal beliefs might simply reflect a
self-serving bias. Moreover, research on cognitive dissonance suggests
that individuals often adjust their attitudes and beliefs76 when they
experience a conflict in their perceptions of reality.77 The next step in this
process is to generalize these personal views to others. Indeed,
individuals often believe that others are more like themselves than they
actually are in reality. As a result, predictions about others’ beliefs or
behaviors based on casual observation are very likely to err in the
direction of one’s own personal beliefs.78
Legal scholarship describes how vigorous legal condemnations of
norms may end up strengthening the very antisocial norms that they are
meant to combat.79 In the context of criminal law, for instance, William J.
Stuntz has described situations in which prosecutions can work against
the very norms on which they rest, causing “popular norms . . . to move

Lederman, The Interplay Between Norms and Enforcement in Tax Compliance, 64 Ohio St. L.J. 1453
(2003).
75. See generally Dan M. Kahan, Social Meaning and the Economic Analysis of Crime,
27 J. Legal Stud. 609 (1998); Timur Kuran & Cass R. Sunstein, Availability Cascades and Risk
Regulation, 51 Stan. L. Rev. 683 (1999); Tracey L. Meares & Dan M. Kahan, Law and (Norms of)
Order in the Inner City, 32 Law & Soc’y Rev. 805 (1998).
76. Joshua D. Rosenberg, The Psychology of Taxes: Why They Drive Us Crazy, and How We
Can Make Them Sane, 16 Va. Tax Rev. 155, 200 (1996) (describing how people amend their cognitive
frame to reduce conflict by incorporating new perceptions of reality).
77. See, e.g., Leo Festinger, A Theory of Cognitive Dissonance (1957); Jon Elster, Sour
Grapes: Studies in the Subversion of Rationality (1983). The classic example of this is expressed in
the fable where a fox sees some high-hanging grapes and wishes to eat them. When the fox is unable to
reach the grapes, he surmises that the grapes are probably not worth eating, as they must not be ripe
or that they are probably sour. Aesop, The Fox and the Grapes (ca. 620–564 BCE).
78. The false consensus effect is described in, for example, Brian Mullen, et al., The False
Consensus Effect: A Meta-Analysis of 115 Hypothesis Tests, 21 J. of Experimental Soc. Psychol. 262
(1985); Lee Ross, et al., The “False Consensus Effect”: An Egocentric Bias In Social Perception And
Attribution Processes, 13 J. of Experimental Soc. Psychol. 279 (1977).
79. For a theoretical model, see Parisi & Wangenheim, supra note 69 (describing a cycle of
opinion formation in which public acts of disobedience and protest undermine the legitimacy of
legislation, which leads to further opposition).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1501

in the opposite direction from the law.”80 Specifically, enforcement may

reinforce or strengthen a belief that the legal regime is not legitimate or
that a legal rule is unjust. This is especially true if the public perceives the
legal sanction to be excessive in relation to the punished behavior.
Several studies document that enforcement measures can backfire and
cause an overall increase in illicit behavior.81 These dynamics help
explain the public backlash in response to the music industry’s litigation
campaign. By increasing enforcement without regard to the public’s
perception, the record industry might have increased, rather than
decreased, the rate and frequency of infringing activities, particularly
among college students.82
3D printing will likely exhibit similar dynamics. Users of 3D printing
obviously benefit from the use of their printer and might adopt a liberal
viewpoint on what can be printed out at home without much regard for
IP rights. Much like users of peer-to-peer networks, owners of 3D
printers might externalize these personal beliefs and come to consider
their own beliefs as social norms. Such liberal, non-commercial 3D
printing norms might be particularly strong since, like file sharing before
it, users might (1) download and print materials that they would not have
bought anyway; (2) believe that their unauthorized print-outs might
inspire others—say a visitor to their home—to buy the actual product in
stores; (3) have altered the design of the items to their own preferences,
creating an attribution effect that causes them to feel entitled to print the
altered version without a license. Given these circumstances, it is
reasonable to assume that pro-3D printing norms will be quite robust
and that enforcement measures might induce a counterproductive
backlash effect.
Punitive deterrence measures may well undermine the interests of
IP rights holders in the forthcoming confrontation with counterfeit 3D
printing. First, enforcement measures that are perceived as excessive
might cause potential infringers to disregard the law on inconsequential
grounds. If the deterrent or coercive aspects of enforcement create a
belief that the underlying legal rules are unjust, individuals might decide
to disobey the law—despite the added costs associated with breaking the
law. Second, evading unjust or immoral enforcement measures might be

80. William J. Stuntz, Self-Defeating Crimes, 86 Va. L. Rev. 1871, 1872 (2000) (suggesting that
misguided enforcement priorities can inadvertently shift public support away the underlying laws).
81. See, e.g., John S. Carroll, A Psychological Approach to Deterrence: The Evaluation of Crime
Opportunities, 36 J. of Personality & Soc. Psychol. 1512 (1978); Harold G. Grasmick & Donald
Green, Legal Punishment, Social Disapproval and Internalization as Inhibitors of Illegal Behavior, 71
J. of Crim. Law & Criminology 325 (1980); Kent W. Smith, Integrating Three Perspectives on
Noncompliance: A Sequential Decision Model, 17 Crim. Just. & Behav. 350 (1990).
82. See Depoorter et al., Copyright Backlash, supra note 51 (providing experimental evidence of
normative backlash effect in the context of file sharing and copyright enforcement).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1502 HASTINGS LAW JOURNAL [Vol. 65:1483

gratifying to some individuals, perhaps sufficiently so that it may

outweigh the risks involved with engaging in infringing behavior. Finally,
in the long run, normatively excessive deterrence may undermine the
political support for the underlying protected rights. As public attitudes
change, political actors are more likely to consider revoking or
moderating the legal framework and available remedies. When
confronting 3D printing infringements, manufacturers would be wise to
heed these lessons from the record industry’s experience with file
sharing.

Conclusion: Alternatives
The enforcement of IP rights on 3D printing presents a vexing
dilemma for holders of patents, trademarks, and copyrights: when
noncompliance and infringement are widespread, effective deterrence
cannot be attained without raising enforcement to levels that threaten to
undermine support for IP rights.
If property rights are costly to enforce (litigation expenses, social
friction) and enforcement is not very effective, the legitimacy of the legal
rights are undermined. In the classic treatment of the emergence of
property rights, Harold Demsetz sets out the conditions that foster the
creation of property rights.83 Several scholars have since elaborated on
the effects of enforcement costs on the efficiency and effectiveness of
private property rights.84 As the classic story of the enclosure movement
illustrates, reductions in private enforcement costs (such as the invention
of barbed wire, for instance) are traditionally associated with an
expansion of private property rights. In this framework, recent
technological changes might suggest a move away from private property
rights to a society favoring the idea of common property.85 It is worth
considering whether 3D printing is one such technology. Like peer-to-
peer file sharing and copyright enforcement before it, 3D printing
technologies will make it increasingly difficult to enforce IP rights on
physical items. Because infringers are decentralized, enforcement is
ineffective and costly. Moreover, efforts to bolster enforcement through

83. Harold Demsetz, Toward a Theory of Property Rights, 57 Am. Econ. Rev. 347 (1967).
Demsetz attributes the relative absence of private property rights on the Southwestern plains to the
high costs of containing wide range, migratory animals. For Native Americans of the Labrador
Peninsula, fencing forest animals was relatively less expensive. Variance in the degree of private
property rights protection can be explained in relation to the costs involved in the “fencing” of those
assets.
84. See Robert C. Ellickson, Property in Land, 102 Yale L.J. 1315, 1315–44 (1993); Barry C.
Field, The Evolution of Property Rights, 42 Kyklos 319, (1989); Terry L. Anderson & P.J. Hill, The
Evolution of Property Rights: A Study of the American West, 18 J. Law & Econ. 163, 164–68 (1975).
85. Saul Levmore, Two Stories About the Evolution of Property Rights, 31 J. Legal Stud. 421,
429–33 (2002) (describing how technologies may affect a shift from commons to private property rights
and back to common property).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

August 2014] DECENTRALIZED PIRACY 1503

technological protections are likely to set off an arms race between

manufacturers and hackers, involving costly, non-productive uses of
resources. These difficulties are illustrated in the long-standing litigation
involving peer-to-peer technologies and copyright law.86
These insights increase the appeal of alternative measures that
might protect the incentives and investments of product designers and
manufactures in a world of 3D printing. All options have their
downsides, of course. Regulating the manufacturing and technical design
of 3D printers to prevent infringing uses may interfere with innovation
and inadvertently tax legitimate uses of 3D printers.87 Imposing IP levies
on raw materials or 3D printers indiscriminately affects all users of 3D
printers, making even legitimate uses of 3D printers more costly.
The costs of enforcing IP rights must be considered fully when
regulating and setting the boundaries of rights on new technologies. 3D
technology is no different. It is important to note that in certain
industries, 3D printing not only reduces the costs of IP infringements, but
also the costs of designing and producing new items of manufacture.
Given that IP rights are created to provide incentives and enable
property right holders to recoup their investments, when those
investment costs decrease, so does the need for IP rights.
When 3D printing reduces the costs of production as well as
distribution of products, the enforcement of IP rights on 3D printing is
likely to create social costs without generating offsetting benefits. This
reason alone justifies taking a hard look at IP doctrine with an eye
towards adapting it to the social and economic changes affected by the
revolution of 3D printing.

86. See Depoorter, Technology and Uncertainty, supra note 32.

87. See Desai & Magliocca, supra note 1 (manuscript at 2).
H - Depoorter_13 (E. Goldberg).doc (Do Not Delete) 8/17/2014 4:52 PM

1504 HASTINGS LAW JOURNAL [Vol. 65:1483

***
 WINTER 2013 V O L . 5 4 N O. 2

Jeroen P.J. de Jong and Erik de Bruijn

Innovation Lessons
From 3-D Printing

REPRINT NUMBER 54212

 I N N O VAT I O N

Producing a copy
of oneself, as Ultimaker
blogger Martijn Elserman
did, is just one potential
application for 3-D printing.

Innovation Lessons
From 3-D Printing
When people in online user communities start collaboratively developing
open-source innovations that have the potential to change an industry,
how should the existing companies in the industry respond? THE LEADING
QUESTION
BY JEROEN P.J. DE JONG AND ERIK DE BRUIJN
Is open
source
innovation a
THESE DAYS, 3-D PRINTING is big news.1 The use of 3-D printing and other related tech-
threat or an
opportunity?
nologies is seen as having potentially transformative implications. “Just as the Web democratized
innovation in bits, a new class of ‘rapid prototyping’ technologies, from 3-D printers to laser cutters, FINDINGS
is democratizing innovation in atoms,” Wired magazine’s longtime editor-in-chief, Chris Anderson, Collaborative user
innovation is most
stated in his new book Makers: The New Industrial Revolution.2 “A new digital revolution is coming, likely to happen in
three kinds of envi-
this time in fabrication,” MIT professor Neil Gershenfeld wrote in a recent issue of Foreign Affairs.3 ronments.
But in addition to 3-D printing’s technological implications, recent evolutions in 3-D printing Existing companies
have five possible
offer important management lessons for executives about the changing face of technological in- responses.
novation — and what that means for businesses. In this article, we examine the rapid emergence of Proactive compa-
a movement called open-source 3-D printing and how it fits into a general trend toward open- nies can take
advantage of user-
source innovation by collaborative online communities. We then discuss how existing companies improved designs.

COURTESY OF ULTIMAKER WINTER 2013 MIT SLOAN MANAGEMENT REVIEW 43

I N N O VAT I O N

can respond to open-source innovation if it occurs create 3-D objects. In 2011, total industry revenues
in their industry — and whether such collaborative for industrial and professional purposes had grown
innovation projects represent a threat or an oppor- to more than $1.7 billion, including both products
tunity for existing businesses. and services. The industry’s compound annual
growth rate has been 26.4% over its 24-year history,
Trends in 3-D Printing and double-digit growth rates are expected to con-
Also known as “additive manufacturing” or “rapid pro- tinue until at least 2019.5
totyping,” 3-D printing is the printing of solid, physical In terms of unit sales, Stratasys Ltd., which now
3-D objects. Unlike machining processes, which are has headquarters in both Eden Prairie, Minnesota
subtractive in nature, 3-D printing systems join to- and Rehovat, Israel, is the world market leader in
gether raw materials to form an object. Drawing on a 3-D printing; Stratasys recently finalized a merger
computer-aided design (CAD) file, the design for an with Objet Ltd., an important producer at the higher
object is first divided into paper-thin, cross-sectional end of the 3-D printing market. Another significant
slices, which are then each ‘printed’ out of liquid, pow- 3-D printing supplier is 3D Systems Corp., based in
der, plastic or metal materials in sequence until the Rock Hill, South Carolina, which recently acquired a
entire object is created. The use of 3-D printing makes it company known as Z Corp.6 Another major supplier
possible to build physical models, prototypes, patterns, is also EOS, based in Munich, Germany, a producer
tooling components or production parts. Design and of higher-end 3-D printing systems.
manufacturing organizations use it for product parts in While early systems were mainly sold to large, mul-
the consumer, industrial, medical and military markets. tinational customers, 3-D printing manufacturers
The longer-term implications of 3-D printing tech- more recently started to focus on the lower end of the
nologies are believed to be large.4 Direct advantages market, offering increasingly cheaper machines to
include enabling designers to operate more efficiently make 3-D printing a viable option for small busi-
and conveniently. They can quickly prototype their nesses, self-employed engineers and designers,
designs in order to test their viability or demonstrate schools and individual consumers. Indeed, 3-D print-
them. In addition, 3-D printing is increasingly used to ing is expected to eventually become a mass market.7
manufacture products or parts in small batches that In the past two years, 3-D printers for home use
would be too costly for a traditional production line. have emerged on the scene. Not very long ago, no such
Moreover, 3-D printing enables entirely new prod- product existed, but thousands of 3-D printers for
ucts to be developed. With layer-wise 3-D printing home use are now being made annually by new startup
processes, many limitations of existing production companies. Five years ago, users started to collabora-
processes are removed, allowing a wider range of de- tively develop home printer designs and to share their
sign options. Widespread use of 3-D printing will open-source designs on the Web. This attracted new
also likely have implications for production, logistics users, some of whom also made innovative contribu-
and retail, since there will be less need to centrally tions, and an open-source 3-D printing community
fabricate products and distribute them if individuals soon developed. More recently, users started to found
can download and locally print a product’s design. their own businesses, which are now commercializ-
However, 3-D printing is not new. Industrial ing open-source-based 3-D printers.
3-D printing manufacturers have been offering
their products for more than 20 years now. Their How the 3-D Printing
machines were initially sold to larger R&D-based Market is Changing
organizations that require high-quality objects and Early 3-D printing systems were expensive — typi-
are able to afford a premium price. Currently, more cally priced at $250,000 and more — and were
than thirty 3-D printing companies around the designed for a limited market. Stratasys, for example,
globe offer a range of industrial 3-D printing sys- initially sold its products to major corporate cus-
tems drawing on various technologies. More tomers such as General Motors and Pratt & Whitney
expensive systems produce fine-grained metal and for use in those companies’ internal R&D processes.
polymer parts, while simpler systems use plastics to As their technologies evolved, 3-D printing compa-

44 MIT SLOAN MANAGEMENT REVIEW WINTER 2013 SLOANREVIEW.MIT.EDU

nies started to focus on the lower end of the market, Bowyer’s original designs. It did not take long for
and applications of 3-D printing for medical uses some members of the RepRap community to start
such as hearing aids and dental implants became selling commercial versions of the RepRap. By 2012,
more common. More recently, 3-D printing compa- user-founded companies were shipping thousands
nies started to have success with systems priced at of machines annually, at prices of $2,500 or less.
$10,000 to $30,000, making 3-D printing viable for These are mostly offered as kits that buyers need to
schools and medium-sized enterprises. Stratasys put together themselves. Such printers are more
also collaborates with traditional printing giant standardized, thus offering more reliability but at the
Hewlett-Packard Co., which is now selling Stratasys’s expense of some flexibility, and it still takes an effort
lower-end products as HP-branded machines. New to get a RepRap operational. In the past year, how-
companies are also forming to offer 3-D printing as a ever, user-founded companies started to offer fully
service. Shapeways Inc., for example, a spinoff of assembled home-use 3-D printing systems. Exam-
Dutch multinational Royal Philips Electronics, ples of user-founded companies include Bits From
prints consumer designs using excess industrial 3-D Bytes, based in Clevedon, United Kingdom; Maker-
printing capacity. Meanwhile, 3D Systems recently bot Industries, based in Brooklyn, New York; and
introduced its Cube, advertised as a home-use 3D Ultimaking Ltd., based in Geldermalsen, the Nether-
printer, starting at $1,299. lands. (Ultimaking plans to change its name to
The RepRap (replicating rapid prototyper) is an Ultimaker soon.)
open-source home-use 3D printer whose popular- All three companies were founded by active Makerbot, cofounded by
Bre Pettis, produces The
ity has grown quickly in the past three years. Like RepRap community members who stepped into Replicator 2, a low-cost
commercial version of the
Stratasys’s machines, the RepRap’s design revolves the lower-end market segment of 3-D printing, original open-source
around a heated nozzle from which a fine filament serving individual designers, artists, inventors and home-use 3-D printers.

of molten plastic is extruded. The nozzle is then students. These open-source printers represent an
moved in X,Y and Z dimensions by computer- incipient challenge to existing 3-D printer compa-
driven motors to successively create each layer of nies at the low end of the market. In response, 3D
the object to be made. Since many of the RepRap’s Systems, for example, acquired Bits From Bytes and
parts are made from plastic, the machine can Botmill, another user-founded business. 3D Sys-
largely replicate itself, with a kit that anyone can tems also recently introducing a fully assembled
assemble given time, materials and a minor invest- Cube 3-D printer to serve individual end users.
ment, typically of about $400.
The RepRap was created in 2005 by Adrian Bow- Innovation by User Communities
yer, a lecturer in mechanical engineering at The rapid growth of open-source 3-D printing is a
University of Bath in the United Kingdom. Bowyer typical example of the broader and emerging phe-
envisioned a machine that would be owned and used nomenon of open collaborative innovation. An
by people to make things at home. Bowyer shared his open collaborative innovation project involves
design for free under the GNU general public contributors who share the work of generating a
license. In its first two years, the project did not re- design and also reveal the outputs from their indi-
ceive much attention, but after early members of the vidual and collective design efforts openly for
RepRap community managed to self-replicate the anyone to use. Contributors to such open-source
machine, adoption began to take off in the summer projects typically contribute for personal need, en-
of 2007. By mid-2010, the size of the RepRap com- joyment and/or reputational gains, to help others
munity was estimated to be more than 3,800 or to develop their skills.
individuals and was estimated to be doubling every While open-source communities are probably
six months.8 More recently, in early 2012, an esti- best known for software development, they are by no
mate of the total number of RepRap machines, means restricted to software or even information
including derivative commercial kits, was 29,745.9 products; as the RepRap community demonstrates,
Thousands of enthusiasts built copies of the Rep- such communities are also viable for developing
Rap for themselves and collaboratively improved physical products. Moreover, innovation by user

COURTESY OF MAKERBOT INDUSTRIES WINTER 2013 MIT SLOAN MANAGEMENT REVIEW 45

I N N O VAT I O N

communities may be expected to increasingly com- experimentation by early users. Accordingly, a com-
pete with and, in some cases, displace corporate mon pattern of industry emergence is that users first
innovation in many parts of the economy. This shift innovate on their own, somewhat later in communi-
is driven by new technologies — the transition to in- ties, while commercial production is seen only later,
creasingly digitized and modularized design and when demand is more certain.12
production practices, coupled with the availability In the earliest days of 3-D printing, users were
of very-low-cost, Internet-based communication — influential sources of innovation, too. While some
and a general trend toward better-educated citizens incumbent 3-D manufacturers are academic spi-
capable of engaging in innovation activities.10 noffs, Stratasys was founded by Scott Crump after
Key questions for innovation-oriented compa- he tried to make a toy frog for his young daughter
nies then become: using a glue gun loaded with a mixture of polyeth-
1. Under what circumstances are companies most ylene and candle wax. His idea was to create the
likely to see innovation by user communities? shape layer by layer. This triggered him to automate
2. Does user innovation pose a threat or opportu- the process and invent the technology that became
nity for our company? the heart of Stratasys’s products.13
3. How can we respond — and how can generally
useful innovations be identified from user com- Type 2: Existing Industries, Where Some Poten-
munities? tial Users Are Not Yet Served After a new industry
Anything that can be cost-justified for develop- emerges, user communities can still be influential.
ment by a user collaborative and that does not The second type of industry in which this is the case is
involve significant economies of scale in replication those in which some potential users are not yet
and diffusion is theoretically a candidate for open served. This situation applies to products that are tar-
collaborative innovation and diffusion. Such col- geted initially toward high-end users able to
laborative innovation is more likely to happen in afford the high prices necessary to overcome initial
three types of environments: nascent industries, investment costs. For example, early cell phones were
industries where some potential users are not yet too expensive for a mass market and were mainly
served and industries where some users are not used by mobile professionals such as doctors; mass-
served adequately. market products were introduced and sold only later.
Many high-technology industries, after their incep-
Type 1: Nascent Industries A lot of innovation ac- tion, can be characterized by a sequence in which
tivity by users takes place in emerging new industries, only big customers are initially served — think of
when commercial markets do not yet exist or are still governments and multinational corporations.
too small and uncertain to attract established com- Medium-sized and small organizations follow later,
panies. User communities may innovate and use and individual end consumers are served last, after
new products prior to commercial production. The the technology has sufficiently matured.14
Wright brothers, for example, developed aircraft for In such instances, user communities may emerge
personal need rather than anticipated commercial building their own versions of high-end products
benefits. They were representatives of a worldwide before a consumer version of the technology is
User-founded 3-D printing community of aviation pioneers seeking after the widely available commercially. In particular, indi-
businesses regularly
incorporate community “holy grail” of controlled flight, while commercial vidual end users may collaborate to develop
innovations in their newest
releases.
aviation companies emerged only later.11 Similar do-it-yourself personal products that are still
patterns have been witnessed in new sports such as insufficiently affordable for an individual. Such
kitesurfing, snowboarding and whitewater kayaking. communities are most likely to form in the case of
Initially, users may justify their investments based on hobbyist products — when use provides enjoyment
expected personal benefits, while commercial man- as opposed to providing pure economic benefit,
ufacturers must be concerned about a larger market and with significant lifestyle benefits. Beyond engag-
potential upon which their profits depend. This ing in communities, hobbyist users are also more
uncertainty is reduced after a period of use and likely to found businesses to commercialize their

46 MIT SLOAN MANAGEMENT REVIEW WINTER 2013 COURTESY OF MAKERBOT INDUSTRIES

inventions. Typically, they launch undertakings that contain innovative user communities and busi-
might be viewed as costs, but which they consider nesses founded by former consumers.17
fun and in line with their personal needs.15
The case of 3-D printing is an example of this Opportunity or Threat?
type of industry. Commercial 3-D printers have Innovative user communities do not necessarily
been available for commercial purposes for 20 pose a threat to existing companies already doing
years, but the systems available for many years were business in an industry. In nascent industries, user
expensive and mainly focused on high-end corpo- communities, in effect, reveal features commercial
rate customers. More recently, cheaper systems products are still lacking, and proactive companies
have made 3-D printing viable for basically all with relevant capabilities and/or similar products
industrial users, but individual consumers had to may take advantage of this and step in. In existing
create their own machines because low-end industries, users may also develop complementary
machines were still comparatively unaffordable. In innovations that increase the value of an existing
effect, the RepRap community and its commercial manufacturer’s current products. Both potential
spinoffs have jumped into the gap at the bottom of roles of innovating user communities are discussed
the market. next.18

Type 3: Existing Industries, Where Some Users Competing Innovations User communities may
Are Not Adequately Served Another environment develop direct alternatives to the products of exist-
in which open-source innovation is likely is when ing companies that are cheaper or more suitable to
commercial products have limited variety and are, in their personal needs. Such innovation alternatives
effect, one-size-fits-all. Manufacturers tend to follow are more likely (but not necessarily or exclusively)
product development strategies to meet the needs of found in industries in which some potential users
homogenous market segments. They are motivated are not yet served. As real-life examples such as
by perceived opportunities to serve sufficiently large Linux demonstrate, the existence of user-devel-
numbers of users to justify their innovation invest- oped competitive innovations does not imply that
ments, but this potentially leaves some users existing companies in the industry will be wiped
dissatisfied with the commercial products on the out. Potential users of the product developed by the
market. Especially when demand is heterogeneous, user community may face high adoption costs, so
many users may not get precisely what they want, and that a product ends up being available either via a
some of them will be motivated to modify the prod- peer-to-peer channel (in other words, through a
uct or spend time and money to develop a home-built user community freely revealing a design) or the
version that exactly satisfies their needs.16 A well- traditional marketplace (in other words, through
known historical example of this phenomenon was companies with commercial offerings).
the Model T Ford, which was produced and sold in The current situation in 3-D printing is an exam-
only one type. Users engaged in massive modification ple of this dynamic. It takes time, skills and effort to
activities to make this car a better match with their build a RepRap, and the self-assembled printer typi-
personal needs. In fact, Model Ts were modified to cally breaks down from time to time. Printer kits
form both snowmobiles and tractors, and these sold by user-founded businesses are better in this re-
modifications were sold as do-it-yourself kits at the spect but still require skills to build and maintain, so
time. some users prefer a traditional commercial product
Note that manufacturers may have good reasons made by an established business.
not to invest in tailor-made products. User seg- The presence of competitive user-developed
ments may be too small for commercial viability, innovations has two types of implications for exist-
or companies may just overlook an opportunity if ing companies in that industry.19 On the one hand,
it’s not easy to spot. Nevertheless, industries char- such innovations decrease users’ willingness to pay
acterized by heterogeneous customer preferences for a commercial product, so price discipline is im-
and many peripheral segments are more likely to posed on existing companies. Not surprisingly, we

SLOANREVIEW.MIT.EDU WINTER 2013 MIT SLOAN MANAGEMENT REVIEW 47

I N N O VAT I O N

currently see that existing 3-D printing manufactur- community, for example, delivers content for Thingi-
ers are offering their products at increasingly lower verse, a community website that allows anyone to post
prices. On the other hand, existing companies may his or her CAD designs, which other users can then
benefit from user-developed competing innovations download, modify and print. Likewise, users have de-
that are potentially commercially attractive. These veloped easy-to-use CAD modeling software, which
innovations identify what is currently missing in has been further refined by user-founded businesses.
commercial products, and companies may just While early RepRap enthusiasts needed to master pro-
adopt these designs and further improve them to fessional CAD software, today they can generate input
offer better products, realize cost savings and so on. data by medical scanners, entertainment software and
Thus, innovating users are a source of free ideas and simple drawing and sketching programs. These com-
prototypes that can enhance a company’s offerings. plementary tools have increased the general value of
lower-end 3-D printers to consumers — regardless of
Complementary Innovations User communities whether those 3-D printers are open-source or based
can also create products that are essential or useful on a manufacturer’s proprietary design.
complements to existing companies’ offerings. In such
cases, user communities may well increase existing Five Ways Existing Companies
companies’ profit potential. Complementary inno- Can Respond
vations are more often found in industries where To respond to innovative user communities, existing
some users are not adequately served. For example, companies have a range of options, and they may be
toolkits were sold to modify a Model T Ford into a applied in parallel. Five strategic responses include:
tractor, and this enhanced Ford’s revenues. (1) monitor, (2) attack, (3) adopt, (4) acquire, and
Complementary innovations by user communities (5) facilitate.20
are found in 3-D printing, too. The 3-D printing user
Monitor Whatever the industry, it never hurts
ABOUT THE RESEARCH
Drawing on the authors’ combination of substantial experience studying user innova-
1 an existing company to keep track of what user
communities are doing — so that one can react or
tion and direct access to the 3-D open-source printing community, we conducted a appropriate what is observed. Close monitoring
Web survey of members of the RepRap open-source 3-D printing community. This
enables existing companies to make more accurate
survey was announced and distributed in collaboration with the RepRap core devel-
opment team. Over three weeks, 892 individuals around the world had checked the sales forecasts and optimize their supply strategies.
survey’s introductory page, and 384 of those completed the questionnaire. In 3-D printing, for example, lower-end manufac-
The survey entailed three sections. First, we asked for respondents’ general char- turers keep track of how the RepRap and derivative
acteristics, including experience with open-source 3-D printing, education, technical
commercial machines are evolving, and most of
skills, personal contacts with other members, time spent on 3-D printing and the allo-
cation of that time. Next, we asked if respondents had created an innovation with
them have employees who build these machines,
respect to RepRap’s hardware or software. If yes, a detailed innovation description either as part of their jobs or in their leisure time.
was obtained, which we used to screen out cases that were in fact without functional If user communities develop potentially competi-
novelty. After the screening process, we ended up with 202 respondents who had in- tive products, existing manufacturers should be
novated, and those respondents reported on a total of 246 innovations. (Some
concerned with understanding users’ adoption costs
community members had reported more than one innovation.)
Finally, we followed up with specific questions about these innovations, including and the size and growth of the do-it-yourself commu-
whether the innovation was developed in collaboration with other members, how nity, which need to be accounted for in their own
much had been invested, if the respondent had freely revealed his or her innovation and pricing strategies. Moreover, user communities’
if s/he considered the innovation easy to adopt. (Some examples required advanced
innovation activities may inform existing manufac-
skills in mechanical engineering, software programming, design or tooling.) A key ques-
tion in this section was if the respondent knew any other members who had adopted turers about their own products’ weaknesses, which
the innovation — something that was true for 28% of the validated innovations. they can then address in subsequent product develop-
Our analysis then focused on identifying key determinants of innovation and diffu- ment and product improvement.
sion. Simple tests of significant differences were done to explore what makes To monitor user innovation effectively, companies
RepRap members innovate and what factors affect whether their innovations are
need to be able to identify users who are innovating.
adopted by others. Finally, we estimated a probit selection model to simultaneously
estimate the determinants of innovation and diffusion — a more robust analysis to What’s more, because prior research has found that
find out which characteristics make innovations valuable to others. the majority of consumer innovations are not ad-

48 MIT SLOAN MANAGEMENT REVIEW WINTER 2013 SLOANREVIEW.MIT.EDU

 WHO INNOVATES IN OPEN-SOURCE 3-D PRINTING?
Our research found that members of the RepRap open-source 3-D printing community are more likely to innovate the more they are experi-
enced with building 3-D printers, maintain personal contact with other community members and focus their 3-D printing time on developing
improvements. In other words, experienced “village elders” who are well-connected to other members are more likely to innovate than the
average RepRap community member.

INNOVATES?
NO YES
CHARACTERISTICS OF COMMUNITY MEMBERS (sample of 182) (sample of 202) SIGNIFICANCE*
Experience in open-source 3-D printing, in years 0.5 1.2 ++
Advanced degree (master’s or Ph.D. degree) 29% 25% o
Technical skills (in mechanical systems, CAD, rapid prototyping systems and
2.2 2.4 +
tooling; 1 = rookie; 4 = expert)
Personal contact with … other members
1.8 2.7 ++
(number of members met face-to-face in past 3 months)
Time spent on 3-D printing, in hours per week 8.4 12.2 +
Division of time:
•Build/fix 3-D printers 54% 41% +
•Use 3-D printers/print objects 16% 17% o
•Develop improvements/ innovating 9% 20% ++
•Help other community members 7% 10% o
•Learn/improve personal skills 14% 12% o
100% 100%

* o indicates no significant relationship with innovation

+ indicates significant relationship with innovation at 5% level
++ indicates significant relationship at 5% level when other variables are controlled for

opted by others,21 companies need rules of thumb to other words, experienced “village elders” who are
identify which user innovations have the most poten- well-connected to other members are more likely to
tial to be seen as useful by others and spread. innovate than the average RepRap community mem-
In the RepRap community, many innovations are ber.22 They are more familiar with the technology and
developed, but only a few prove generally useful and are probably also more expert than newbies, and they
therefore picked up by many other community mem- may enjoy better access to information that can be
bers. To explore the determinants of innovation and its recombined to create an innovation.
diffusion in this community, we organized a Web sur- But not all user innovations are alike. Some inno-
vey. (See “About the Research,” p. 48.) We identified vations are adopted by other community members,
rules of thumb regarding what kinds of community and those are likely to be of most interest to existing
members are likely to innovate and which of their in- companies, because they have already demonstrated
novations are most likely to be adopted by others. some general appeal. Our survey found that, in the
We found that RepRap community members are RepRap community, innovations are more likely to be
more likely to innovate the more they are experienced adopted by other community members if they are
with building 3-D printers, maintain face-to-face freely revealed by the innovators, are easy to adopt
contact with other community members and focus and are concerned with software rather than hard-
their 3-D printing time on developing improvements. ware. (See “When Do Innovations in Open-Source
Innovators also report that they have better technical 3-D Printing Spread?” p. 51.) Innovations are also
skills; they spend more time on 3-D printers overall more likely to spread if they are developed by mem-
but are less likely than other community members to bers who are well connected within the community
spend time on building or fixing their machine. (See and who allocate more of their 3-D printing time to
“Who Innovates in Open-Source 3-D Printing?”) In printing objects and/or using their printers.

SLOANREVIEW.MIT.EDU WINTER 2013 MIT SLOAN MANAGEMENT REVIEW 49

I N N O VAT I O N

Those users who spend much time trying to Acquire Companies may absorb relevant
innovate are, not surprisingly, more likely to inno-
vate, but interestingly, their innovations do not
4 knowledge and skills from user communities
via individuals in the community. Key community
have a better chance of being adopted by others. members may be recruited to join a company and/or
Simultaneously, those spending their time using accept to collaborate with a company. Existing com-
the machine are not more likely to innovate, but if panies may also acquire startups founded by
they do, their innovations are more likely to be community members in order to get a foothold in an
picked up by other users. emerging market. Such initiatives make particular
sense when user communities develop innovations
Attack Another strategy existing companies that could compete with a company’s products. In
2 may consider is attacking innovating user com-
munities, which might seem viable in cases where
3-D printing, some existing system manufacturers
have engaged in quite proactive acquisition behav-
users infringe upon a company’s patents. In some in- iors. For example, 3D Systems bought the
stances, such a response makes sense — for example, user-founded venture Bits From Bytes in October
if users modify a company’s product in an undesired 2010. As Bits From Bytes offers low-cost, extrusion-
way, such as to create a weapon. In general, however, based systems based on the RepRap design, this gave
bringing individual users to court is not likely to be 3D Systems a foothold in the hobbyist and educa-
effective. While direct competitors can be sued, fight- tional marketplace. In addition, 3D Systems has been
ing thousands of individuals around the globe is active in acquiring service providers.25
usually impractical, not beneficial in terms of reve-
nues and potentially harmful to the company’s Facilitate Existing companies may also seek to
reputation among customers. Attacking users may
well escalate into a conflict in which users appeal to
5 influence the direction and nature of the efforts
of user communities. This option may make partic-
higher principles and inflict substantial damage.23 In ular sense when communities develop innovations
3-D printing, system manufacturers have so far re- that are complementary to, rather than competitive
frained from bringing individual user community with, a company’s existing products. The more indi-
members to court. This might change if user-founded viduals are able to contribute and use these
companies become increasingly competitive, or set a complements, the more customers there will be for
new industry standard, but existing manufacturers the company’s own commercial products. Accord-
should still be aware of potential negative reactions.24 ingly, companies may want to stimulate the
emergence of such user communities and facilitate
Adopt In some instances, companies will want their innovation and sharing activities. This can be
3 to adopt or copy the technologies, methods or
improvements developed by user communities.
done by methods such as online toolkits that facili-
tate user innovation and product modification,
These can generally be obtained for free, as many contests and awards, or sponsoring websites or key
innovators in user communities do not protect community members’ contributions.
their contributions with patents. In 3-D printing, When companies try to influence user communi-
incorporating community innovations is an al- ties, they first need to understand why community
most routine practice. For example, Delta Micro members participate in the community. LEGO’s expe-
Factory Corp., based in Beijing, China, introduced rience with its user communities, for example, suggests
an extrusion-based portable printer called UP! at that companies should not just focus on getting a job
the lower end of the market which incorporates done, but rather recognize that users’ needs are differ-
some of the innovations developed by the RepRap ent from those of employees, and intrinsic rewards
community. Likewise, user-founded businesses like (recognition, social bonding, fun) generally may be
Makerbot and Ultimaking still intensively collabo- more helpful than money in motivating users. More-
rate with the communities from which they over, community members often do not like intellectual
emerged, regularly incorporating community in- property protection. The potential tension is that a
novations in their newest releases. company often would prefer that users freely provide

50 MIT SLOAN MANAGEMENT REVIEW WINTER 2013 SLOANREVIEW.MIT.EDU

 WHEN DO INNOVATIONS IN OPEN-SOURCE 3-D PRINTING SPREAD?
Our survey found that, in the RepRap 3-D printing community, innovations were more likely to be adopted by other community members
if they were freely revealed by the innovators, were easy to adopt and were software innovations rather than hardware innovations.
Innovations were also more likely to spread if they were developed by members who are well-connected within the community and who
allocated more of their 3-D printing time to printing objects and/or using their printers.

DIFFUSION OF INNOVATION?
NO YES
CHARACTERISTICS OF THE INNOVATOR (sample of 177) (sample of 69) SIGNIFICANCE*
Experience in open-source 3-D printing, in years 1.1 2.0 +
Advanced degree (master’s or Ph.D. degree) 23% 26% o
Technical skills (in mechanical systems, CAD, rapid prototyping systems and
2.4 2.5 o
tooling; 1 = rookie; 4 = expert)
Personal contact with … other members
2.0 6.0 ++
(number of members met face-to-face in past 3 months)
Time spent on 3-D printing, in hours per week 11.5 15.7 +
Division of time:
•Build/fix 3-D printers 44% 28% +
•Use 3-D printers/print objects 14% 24% ++
•Develop improvements/ innovating 20% 24% o
•Help other community members 8% 14% +
•Learn/improve personal skills 14% 10% o
100% 100%
CHARACTERISTICS OF THE INNOVATION
Concerned with hardware (versus software) 69% 58% ++
Developed in collaboration with other members 27% 45% +
Time investment, in days 28.3 32.4 o
Money investment, in dollars 112.0 101.6 o
Was freely revealed (for example, via blog or wiki) 44% 83% ++
Is easy to adopt (according to innovator) 20% 33% ++

* o indicates no significant relationship with innovation diffusion

+ indicates significant relationship with innovation diffusion at 5% level
++ indicates significant relationship with innovation diffusion at 5% level when other variables are controlled for

their ideas, prototypes and/or solutions for feedback industries in the future. This is due to several exoge-
and further improvement by the company, but on the nous trends, including increased understanding of
other hand, the company would like to keep knowledge modular design practices, decreasing design and col-
within its own doors in order to protect it.26 laboration costs thanks to cheaper and more capable
computerized design tools, decreasing communica-
What’s Next? tion costs thanks to the Internet, and increasingly
The emergence of open-source 3D printers illus- better educated populations of citizens across the
trates how quickly high-tech markets can be changed globe. These same advances also diminish the costs of
by innovative, collaborative users and the companies innovation diffused via peer-to-peer networks. In-
they form. And, as the case of 3-D printing shows, deed, these factors are all drivers behind the
emerging user communities offer lots of opportuni- emergence of open-source 3D printing, but they will
ties to existing producers — as well as competition. also affect other industries in the future. Smart com-
What’s more, we believe that innovation by user panies should start rethinking their innovation
communities will be increasingly seen in more management practices accordingly.

SLOANREVIEW.MIT.EDU WINTER 2013 MIT SLOAN MANAGEMENT REVIEW 51

I N N O VAT I O N

Jeroen P.J. de Jong is an associate professor of stra- 12. E. von Hippel, S. Ogawa and J.P.J. de Jong, “The Age
tegic management and entrepreneurship at of the Consumer-Innovator,” MIT Sloan Management Re-
Rotterdam School of Management, Erasmus Univer- view 53, no. 1 (fall 2011): 27-35.
sity in Rotterdam, the Netherlands. Erik de Bruijn is a
13. See www.fundinguniverse.com/company-histories/
cofounder of Ultimaking Ltd., a startup in Gelder-
stratasys-inc-history.
malsen, the Netherlands that manufactures
Ultimaker 3-D printers. He is a member of the RepRap 14. A. Afuah, “Innovation Management: Strategies, Im-
core development team. Comment on this article at plementation, and Profits,” (Oxford: Oxford University
http://sloanreview.mit.edu/x/54212, or contact the Press, 2003).
authors at [email protected].
15. S.K. Shah and M. Tripsas, “The Accidental Entrepre-
neur: The Emergent and Collective Process of User
ACKNOWLEDGMENTS Entrepreneurship,” Strategic Entrepreneurship Journal 1,
no.1 (November 2007): 123-140.
The authors thank Eric von Hippel of the MIT Sloan School
of Management for useful advice and suggestions while 16. E. von Hippel, “Democratizing Innovation” (Cam-
developing their research. The survey was financially sup- bridge, Massachusetts: MIT Press, 2005).
ported by the SME and Entrepreneurship research 17. For an extensive discussion, see Shah and Tripsas,
program, which is conducted by Panteia/EIM in Zoeter- “The Accidental Entrepreneur.”
meer, the Netherlands.
18. Industries can be characterized as user-contested
and/or user-complemented markets. The nature and mi-
REFERENCES croeconomic effects of innovating users in both types of
markets are discussed by C. Raasch and E. von Hippel,
1. In newspapers and magazines, 3-D printing has been “Modeling Interactions Between User and Producer
described enthusiastically as everything from “the tech- Innovation: User-contested and User-complemented
nology that could re-shape the world” to “one that is Markets,” working paper (June 2012). See http://ssrn.
about to transform every single aspect of our lives.” See, com/abstract=2079763.
for example, S. Richmond, “3D Printing: The Technology
That Could Re-Shape the World,” The Telegraph, July 28, 19. For a detailed discussion, see Raasch and von Hippel,
2011; and H. Lipson, “3D Printing: The Technology That “Modeling Interactions.”
Changes Everything,” New Scientist, July 30, 2011, p. 20. 20. These strategies were partly identified by S. Flowers,
2. C. Anderson, “Makers: The New Industrial Revolution” “Harnessing the Hackers: The Emergence and Exploita-
(New York: Crown Business, 2012): 14. tion of Outlaw Innovation,” Research Policy 37, no. 2
(March 2008): 177-193.
3. N. Gershenfeld, “How to Make Almost Anything: The
Digital Fabrication Revolution,” Foreign Affairs 91, no. 6 21. See von Hippel, Ogawa and de Jong, “The Age of the
(November-December 2012): 43-57. Consumer-Innovator.”
4. T.T. Wohlers, “Wohlers Report 2012: Additive Manu- 22. A similar finding was reported by L. Dahlander and L.
facturing and 3D Printing State of the Industry, Annual Frederiksen, “The Core and Cosmopolitans: A Relational
Worldwide Progress Report” (Fort Collins, Colorado: View of Innovation in User Communities, ” Organization
Wohlers Associates, 2012). This annual report elaborates Science 23, no. 4 (July-August 2012): 988-1007.
substantially on the 3-D printing industry, technologies
23. E. Mollick, “Tapping into the Underground,” MIT
and trends.
Sloan Management Review 46, no.4 (Summer 2005):
5. Ibid., p. 125 and 131. 21-24.
6. Ibid., p. 133. See also “Stratasys and Objet Complete 24. 3D Systems recently sued Formlabs, a new venture
Merger,” Stratasys press release, December 3, 2012, offering a home use system printing high-resolution
http://investors.stratasys.com/releasedetail. metal parts. See www.3ders.org/articles/20121121-3d-
cfm?ReleaseID=724378. systems-files-patent-infringement-lawsuit-against-
7. Ibid., p. 65 and 256. formlabs-and-kickstarter.html. Although Formlabs did
not originate from the RepRap community, the case is
8. E. de Bruijn, “On the Viability of the Open Source De-
also marked by negative responses from home users:
velopment Model for the Design of Physical Objects:
see http://lucept.com/2012/11/24/3d-systems-suing-
Lessons Learned From the RepRap Project” (master’s
kickstarter/, accessed on November 28, 2012.
thesis, Tilburg University, 2010).
25. See Wohlers, “Wohlers Report 2012,” p. 118.
9. Wohlers, “Wohlers Report 2012,” p. 137.
10. C. Baldwin and E. von Hippel, “Modeling a Paradigm 26. Y.M. Antorini, A.M. Muñiz and T. Askildsen, “Collabo-
Shift: From Producer Innovation to User and Open Collab- rating With Customer Communities: Lessons From the
orative Innovation,” Organization Science 22, no. 6 Lego Group,” MIT Sloan Management Review 53, no. 3
(November-December 2011): 1399-1417. (spring 2012): 73-79.

11. P.B. Meyer, “Open Technology and the Early Airplane

Industry” (presented at the Economic History Association Reprint 54212.
2012 Annual Meeting, Vancouver, Canada, Sept. 21-23, Copyright © Massachusetts Institute of Technology, 2013.
2012). All rights reserved.

52 MIT SLOAN MANAGEMENT REVIEW WINTER 2013 SLOANREVIEW.MIT.EDU

PDFs ■ Reprints ■ Permission to Copy ■ Back Issues
Articles published in MIT Sloan Management Review are
copyrighted by the Massachusetts Institute of Technology
unless otherwise specified at the end of an article.
MIT Sloan Management Review articles, permissions,
and back issues can be purchased on our Web site:
sloanreview.mit.edu or you may order through our
Business Service Center (9 a.m.-5 p.m. ET) at the
phone numbers listed below. Paper reprints are
available in quantities of 250 or more.
To reproduce or transmit one or more MIT Sloan
Management Review articles by electronic or
mechanical means (including photocopying or archiving
in any information storage or retrieval system) requires
written permission.
To request permission, use our Web site:
sloanreview.mit.edu),
or
E-mail: [email protected]
Call (US and International):617-253-7170
Fax: 617-258-9739
Posting of full-text SMR articles on publicly accessible
Internet sites is prohibited. To obtain permission to post
articles on secure and/or password-protected intranet sites,
e-mail your request to [email protected].
Customer Service
MIT Sloan Management Review
238 Main Street E48-570
Cambridge, MA 02142
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Industrial and Corporate Change, 2017, 1–19
doi: 10.1093/icc/dtx006
Original article

The importance of ergonomic design in product

innovation. Lessons from the development of the
portable computer
Paul Windrum1, Koen Frenken2,* and Lawrence Green3
1
Nottingham University Business School, Nottingham, UK. e-mail: [email protected],
2
Innovation Studies, Copernicus, Institute of Sustainable Development, Utrecht University, Utrecht,
Netherlands. e-mail: [email protected] and 3Birmingham Institute of Art and Design, Birmingham City
University, Birmingham, England. e-mail: [email protected]
*Main author for correspondence.

Abstract
The article addresses the role of ergonomic design in product innovation. Designers meet users’
needs by developing solutions to complex trade-offs—reverse salients—between a product’s charac-
teristics. The fundamental ergonomic design challenge in portable computers concerns the reverse
salient between two ergonomic factors: screen size and weight. It is easier to view information on
larger screens, but portability is negatively affected by the weight of larger batteries required to power
larger screens. This ergonomic reverse salient shaped the innovation trajectory of the portable com-
puter, from the selection of the clamshell portable over alterative design configurations, to the search
for more efficient batteries and new types of screens. Based on hedonic price analysis on data of ergo-
nomic and technological characteristics, we show that (i) screen size and weight are key components
in hedonic price functions, (ii) the interaction between screen size and weight is distinct from inter-
actions between other, technological, characteristics that affect computing power, and (iii) positive
prices are paid for the product solutions to the ergonomic reverse salient.
JEL classification: O32, L63, C12

1. Introduction
This article contributes to a growing body of research on design and innovation by addressing the role of design ergo-
nomics in product development. Prior studies have highlighted the contributions of design and aesthetics to product
development (Bloch, 1995; Postrel, 2003, Eisenman, 2013), the designer as technology interpreter and practical
translator (Lawson, 2006), and the integration of design, engineering, and marketing functions in the new product
development process (Moenaert and Souder, 1990; Perks et al., 2005). In addition, some recent contributions have
focused on design as a driver for innovation (Verganti, 2009), “design thinking” as a means of structuring strategic
product development (Brown, 2008), and the role of design in articulating creativity and innovation (Cox, 2005).
While these contributions foreground key aspects of the design–innovation relationship, their focus falls squarely on
issues of technology, aesthetics, and the management of the product development process.

C The Author 2017. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved.
V
2 P. Windrum et al.

In contrast to the fields listed above, the role of design ergonomics as a critical input to product innovation has re-
mained an under-researched topic. Ergonomics is concerned with the ways in which a physical artifact interacts with
the human body, and with the environment in which the artifact/human is expected to move and operate. It involves
“design for effective use,” which explicitly takes account of the user’s physical and psychological capabilities and
limitations (Boff, 2006; Salvendy, 2012). To assess the fit between the user and the artifact, i.e., the latter’s “human
compatibility” (Karwowski, 2005), the designer must analyze the physical attributes of the typical user, the activity
being performed, and the demands placed on the user by the product during the activity. Of particular importance,
here are the size, shape, weight, and configuration of the product, and how appropriate these are for the task.
Ergonomics is central to the effective design and application of a wide range of products, for example, medical de-
vices that aid hearing or mobility, office equipment that minimizes repetitive strain, or kitchen utensils that provide
safety and comfort in extended use. In portable devices, such as portable computers, designers face the challenge of
addressing a “reverse salient” (Hughes, 1983, 1987) that exists between two ergonomic features: here, screen size
and overall weight. Larger screens are ergonomically beneficial because viewing is easier for the user. However,
larger screens require bulkier and heavier batteries, and these adversely affect the portability of this electronic device.
Hence, the ergonomic penalty of larger screens is the increased weight of a device that the user is likely to need to
transport, and to place in their lap when in use.
The problem of weight in portable computers is an ergonomic reverse salient that impeded the overall rate of pro-
gress of the whole product, since critical components such as screen size, could not be permitted to increase total unit
weight beyond reasonable parameters. The ways in which designers have sought to address this ergonomic reverse sa-
lient have shaped significantly the innovation trajectory of the portable computer. Impacts include the development
of the clamshell design configuration, new types of screen technologies, the development of ergonomic standards for
human–screen interaction, and the search for more efficient battery types.
In addition to meeting a set of ergonomic requirements through the product’s design, designers are required to de-
velop a set of measurable indicators that clearly convey information with respect to the product’s ergonomic per-
formance to the consumer. These indicators are strategically important to firms as a means of differentiating the
quality of their product offerings vis-a-vis rival products. In the Lancaster tradition of product innovation (Lancaster,
1966, 1971), these measurable features are known as “product characteristics.” The information that is reported by
firms in their product specifications is an important input to the reviews conducted by specialist consumer magazines,
and is used by consumers in their purchasing decisions. Seminal research by Alba and Hutchinson (1987, 2000) high-
lights the importance of information on ergonomic and technical performance in consumers’ decision-making.
Knowledgeable consumers place greater weight on product attribute information than on advertising exposure or dir-
ect interactions with salespersons.
The role of designers is to meet the expressed and latent needs of users through product design, given prevailing
and anticipated production capabilities, and the costs of realizing these product characteristics. Consumers are the ul-
timate arbiters of whether designers develop effective solutions to reverse salients. Hence, we use information col-
lected on the ergonomic and technological product characteristics1 of laptop computers, available to consumers
when making their purchasing decisions, to empirically test two research hypotheses. The first hypothesis examines
whether a positive price is paid for designers’ solutions to the screen-weight ergonomic reverse salient. As noted
above, larger-sized screens are easier to read but carry the penalty of larger and heavier batteries required to run
them: this is a penalty that impacts negatively on device portability. The second hypothesis examines whether a posi-
tive price is paid for solutions to the technology reverse salient associated with computing power. These hypotheses
are tested by estimating a set of hedonic price models. Our findings indicate that positive prices are paid for products
that address the ergonomic reverse salient as well as the technological reverse salient. This highlights the need for a
deeper understanding, and analysis, of the contributions of ergonomic design to product innovation. As Stoneman
(2010) has argued, studies which omit these contributions—focusing solely on improvements in technologically
driven performance—significantly under-report innovation.

1 Note that both ergonomic and technological characteristics are examples of “service characteristics” in the sense of
Saviotti of Metcalfe (1984), that is, characteristics that are explicitly valued by users.
The importance of ergonomic design 3

2. Reverse salients and product design

The “reverse salient” concept entered innovation and technological development discourses in the early part of the
1980s, most notably via the contributions of Hughes (1983, 1987). It derives in its current application from the study
of technologies and complex products as “systems,” i.e., those approaches that view technological products as inter-
dependent systems and subsystems of components (Henderson and Clark, 1990; Murmann and Frenken, 2006). In
its most simple form, the notion of reverse salience is applied to reference those components in a complex and coevo-
lutionary nexus in which development is retarded. As a consequence of their limitations, such components are likely
to impede the overall rate of progress of a product or system as a whole.
The concept of reverse salience relates closely to that of “bottlenecks” or “technological imbalances” (Rosenberg,
1969; Dedehayir, 2009) in the coevolution of interlinked elements within a product or system. Where optimal pro-
gress in performance requires that all interdependent components or subsystems develop with orchestrated continu-
ity, the failure to maintain pace of one component—the appearance of a reverse salient—will imply disruption to the
collective system’s “advancing performance frontier” (Dedehayir, 2009: 576). Clearly, where possible, the emergence
of reverse salients is to be avoided: however, where the latter are encountered, interventions are required to ensure
rapid correction (Hughes, 1987). Here, we see the reverse salient as a “focusing device” (Rosenberg, 1969), that is, a
problem around which system actors (technologists, engineers, designers, managers, marketers, etc.) will agglomerate
in the effort to derive appropriate solutions and thus reestablish developmental equilibrium.
The role of designers in tackling reverse salients is central: designers address reverse salients by developing prod-
uct designs that configure the user in specific ways, and different designers may come up with very different design
solutions for their intended consumers (Woolgar, 1991, 1994). The three core areas of competence in which de-
signers contribute to the product development process—ergonomic, aesthetic, and technological—are founded on
two transversal capabilities (Miles and Green, 2008). First, an ability to recognize and respond to expressed and la-
tent needs of potential users. Second, an ability to derive solutions to the complex problems that emerge frequently in
the process of envisioning and creating new industrial and consumer products. Indeed, problem-solving capability
lies at the core of product design endeavor (Suh, 2001; Lawson, 2006), and experienced designers are arguably well-
equipped to manage emergent difficulties in the coevolving nexus of technological, aesthetic, and ergonomic factors
that characterize the development of complex contemporary products.
While several models of the design-led problem-solving process appear in the design literature (Cross, 2001),
most approaches are premised on a sequential (feedback looped) flow that commences with problem framing (or def-
inition), and proceeds in various steps through research and exploration, idea generation, experimentation with alter-
native solutions, idea synthesis and selection, and on to prototyping and implementation. Frequently characterized as
a process that commences with “divergent” and concludes with “convergent” thinking (i.e., one that moves from the
identification of many solutions to the selection of an optimal fix), the resolution of reverse salients—whether these
arise within or between ergonomic, aesthetic, or technological factors in new product development—is an activity
with which the design profession is well-acquainted, and one that is embedded in training and reinforced by practice
(Schon, 1983; Hill, 1998; Cross, 2001).
Two important reverse salients are evident in the developmental trajectory of portable computing. One concerns
“processing power” and is common to both portable and desktop computers. Computing power is a complex phe-
nomenon that governs both computer speed and software stability. The reverse salient that arises in relation to com-
puting power centers on the balance required in the development of microprocessors and disk drives (Baldwin and
Clark, 2000). Computing power depends on interactions between the random access memory (RAM) of a micropro-
cessor and disk drive storage. A computer program requires contiguous working memory. In practice, this is physic-
ally fragmented on RAM and may overflow on to disk storage. Memory is managed by “virtual memory,” which
frees up RAM by identifying areas that have not been used recently and copies them on to the hard disk. The area of
the hard disk that stores the RAM image is called a page file. A balanced design requires developments in RAM that
are matched by developments in disk drive capacity. The advantage of hard disk memory is that it is cheap (compared
to RAM). However, the read/write speed of a hard drive is much slower than RAM and is not as effective in accessing
fragments of data. A design which is overly dependent on virtual memory suffers in terms of performance. In the
worst case, “thrashing” occurs, and the computer grinds to a halt as the operating system constantly swaps informa-
tion between RAM and hard disk memory.
4 P. Windrum et al.

The second reverse salient is ergonomic in nature, and concerns a fundamental trade-off between usability and
portability. Larger screens make it easier for users to view information and to work with data entry and data output.
However, the operation of such screens in typical use-time scenarios requires larger, heavier batteries. The increase in
total weight renders the product less portable, as it is more onerous to carry and less comfortable when placed on
one’s lap. As we shall see in the next section of the article, the ergonomic screen size–weight reverse salient has been
a key driver of innovation in portable computers.
In contrast to the relationship between prices and computing power, portability and the reverse salient between
screen size and weight have been downplayed or sometimes ignored in previous studies. This is even the case in the
few examples of studies of portable computer pricing (Nelson et al., 1994; Berndt et al., 1995; Baker, 1997; Berndt
and Rappaport, 2001; Chwelos, 2003). To the extent that these studies have examined portability as a characteristic,
it has been operationalized typically solely in terms of weight or volume.

3. Screen size–weight reverse salient in portable computers

Compared to contemporary personal computers (PCs), early portables provided significantly reduced processing
power: a key advantage, however, was their mobility. For the first time, salespeople could sit with clients to discuss,
display, and configure product options, and then produce instant quotes using powerful spreadsheet software. This
gave portable users an edge over competitors who needed to refer information back to local offices to have quotes
drawn up and posted out. Salespeople were also able to complete standardized electronic orders remotely, and collect
or log other information that could be used to update company databases on their return to the office. For senior ex-
ecutives, portables enabled remote working and work while travelling. Thus, it became possible to develop presenta-
tions and budget sheets on the move, and to refresh and update information and content between meetings
(Gatignon and Robertson, 1989). For both sales and executive users, larger screen sizes were highly important, as
these permitted the presentation of material to small groups around a table.
The first commercially successful portable computer was a “portable box” design, the Osborne I, released in
April 1981.2 Portable box computers are often referred to as a “luggables” due to their relatively large size—about
the size of a small suitcase—and weight (e.g., the Osborne I weighed almost 24 lbs). The unit opened on one side to
reveal a small, 5” monochrome cathode ray tube (CRT) display and a fold-down keyboard. CRTs were, at that time,
a well-established screen type, having had a long history of use and incremental development in televisions. The big
disadvantage of CRTs was weight, even for modestly sized CRT units. Given the physical size and weight of the port-
able box design, it was intended that operators should sit at a desk, thus limiting the use of portable boxes to an of-
fice or workplace environment. The sheer mass of boxes also limited general mobility for many users.
In the rival “clamshell” design, the user was configured differently. The clamshell is a more compact and lighter
weight design comprising a large flat screen set into a unit that is intended to be balanced on the user’s lap leaving
both hands free to type, hence the term “laptop computer” (Safire, 1988).
The “clamshell” concept was initially created and developed by Bill Moggridge, a leading British industrial designer, in
association with GRiD. The design is a “form factor”—it comprises two sections that fold via a hinge. The components
are kept inside the clamshell, and the latter is opened up when in use. The design was patented (US Patents D280,511 and
4,571,456) for the GRiD Compass portable computer, which was launched in April 1982. The GRiD Compass sported a
large, flat panel (monochrome) electroluminescent display screen. Processing hardware (Intel processor, RAM, and data
storage memory) and the battery were housed in a rectangular magnesium case, designed to ensure high levels of compo-
nent protection and an efficient heat dissipation mechanism. The Compass weighed just 11 lbs (Wilson, 2006).
The ergonomic attractiveness of the clamshell vis-a-vis the portable box design was a key selling point for the
early adopters or “lead users” (von Hippel, 1986) purchasing portables during the early 1980s. As noted, early port-
ables were expensive business machines that were targeted at field salespeople and senior executives. When launched,
the Osborne I had a price tag of US$1795.00, and the GRiD Compass retailed at more than US$8000.00.
The clamshell quickly became the dominant industry design. Still, the ergonomic reverse salient between screen
size and weight persisted as a key innovation driver. Rival product designers engaged in the development of machines
with larger, higher quality flat screens, and in experimentation with new battery types.

2 The first portable computer predates the release of the first IBM PC (5150), which was launched in August 1981 in the
United States.
The importance of ergonomic design 5

Portable designers explored the possibilities of larger screens using liquid crystal displays (LCDs).3 The Toshiba
T1100 (released in April 1985) was the first clamshell to use a backlit LCD. These screens are particularly suited to
the clamshell design: they provide better resolution and luminosity than electroluminescent counterparts, and their
lightness and thinness are particularly suited to use in the clamshell lid. Further, the low electrical power consump-
tion of LCDs places less demand on batteries. Indeed, it was the commercial success of the clamshell portable that
bootstrapped the development of LCDs during the 1990s (Lien et al., 2001).
Improved visibility also required a scientific understanding of screen visualization and the development of a set of
standards to underpin the work of specialist ergonomic designers. Human–screen interaction standards were de-
veloped during the late 1980s and early 1990s and were quickly adopted by portable computer firms. These cover
the recommended reading distance of a display (Boff and Lincoln, 1999), the useful field of view (Ware, 2004), lumi-
nescence (Shneiderman, 1992), font size and font type (Sanders and McCormick, 1993; Mayhew, 1999), and color
contrast (Ware, 2004). With these standards in place, the remaining variable governing user’s ease of reading is total
screen area (height  width).
To address the issue of progressive increase in weight, and to safely power larger LCD screens (overstressing a bat-
tery can result in catastrophic meltdown) designers experimented with new, more powerful nickel metal hydride
(NiMH) and lithium-ion (Li-ion) battery types. In the late 1980s, designers switched from nickel–cadmium batteries
to NiMH batteries. NiMH has a 30%–40% higher capacity over nickel–cadmium, is less prone to battery memory
loss, offers simple storage and transportation, and is more environmentally friendly (Linden and Reddy, 2001). In
the early 1990s, NiMH was in turn replaced by Li-ion batteries that have a longer service life and a higher electro-
chemical potential: even today these cells possess the largest density for weight of all currently available options (van
Schalwijk and Scrosati, 2002). As with screen displays, the scale and economic significance of the portable computer
sector was such that it induced key innovations in the related battery sector.

4. Statistical methods
We have chosen to test our hypotheses using hedonic regression analysis on published data of laptop prices and prod-
uct characteristics during a particular historical period: that of 1993–1996. The reasons for this are as follows. One
of the most important problems facing those estimating product features, regardless of statistical method, is misspeci-
fication due to omitted variables. During the period 1993–1996, portables were stand-alone business machines that
contained relatively few well-defined hardware features, compared to subsequent years. After this period there was a
proliferation of hardware features. If one were to estimate characteristics prices today, for example, omitted variable
problems due to multiple hardware features would be far greater.4
There are sources of omitted variable bias that would adversely affect a study of current laptop machines but
which are avoided by examining this historical period. First, the chosen period predates the commercialization and
widespread use of the Internet and the worldwide Web. It was also an era before software plug-ins and apps.
Another potential source of omitted variable bias is software–hardware bundling. Rival hardware manufacturers
may include alternative types of software within their offer prices (Triplett, 2006). In the period 1993–1996, there
was a high degree of standardization around a limited number of business software packages—certainly by compari-
son with today. The package software market at this time was dominated by Lotus Symphony and Excel (spread-
sheets), WordPerfect and Word (word processing packages), and PowerPoint. We note that all of the laptops listed in
our data set used Microsoft’s Windows 3.0 operating system.

3 An organic liquid is the active ingredient in an LCD panel; argon or neon gas in a gas plasma screen; a metal film in an
electroluminescent screen.
4 An alternative approach to identifying preferences is discrete choice analysis (also known as conjoint analysis). Here
customers are asked to state their willingness to pay for multiple product characteristics. This has a number of well-
known limitations. These include limited levels of characteristics which respondents are asked to consider (in the limit
these are binary options), and the information and computational demands placed on respondents in consistently scor-
ing or ranking more than a few characteristics. Problems of omitted variable bias arise. Over the past decade, the focus
has been on developing computer-based techniques that guide respondents through a limited subset of product charac-
teristics. This does not resolve the issue of omitted variable bias, per se, and there is, as yet, no consensus on these
subset approaches (Hauser and Rao, 2004; Hainmueller et al., 2014).
6 P. Windrum et al.

A further advantage in using this period is that (the limited) prior research on laptop computers by Baker (1997)
and Chwelos (2003) also consider this period. It provides a useful basis of comparison. Also, these papers previously
addressed issues, such as the relationship between product characteristic variables (e.g., megahertz) and benchmark
computing system performance. Chwelos (2003) found, during the era that we are considering, that the price index
differs trivially between benchmark performance measures and a set of product characteristics. Triplett (2005) ob-
serves that one reason for this result is that Chwelos’ product characteristic specification was unusually rich, includ-
ing microprocessor clock speed, cache memory (RAM), and hard disk capacity. The same result may not hold in
simpler hedonic price regression models that include fewer product characteristics.
Another advantage in studying the 1993–1996 period is that there was a clear set of lead users for this product.
Businesses purchased these machines for use by salespeople and senior managers. The ergonomic reverse salient was
an important consideration for these particular users. Larger screens were valued by salespeople in the field because
they were able to demonstrate to clients alternative options and plans. Larger screens were also useful for mobile se-
nior managers when delivering presentations to clients and other business leaders. Minimizing weight was important
to both groups given the requirement for ease of portability while on the road.
Finally, a large number of competing US, European, and Asian manufacturers were producing and selling prod-
ucts internationally during this era. This provides a large number of product observations on a relatively small num-
ber of key ergonomic and non-ergonomic product characteristics.
Ideally, one would like to have data on sales of each individual portable as well as data on prices and product fea-
tures. In reality, this is rarely, if ever, available to the analyst (Bhaskarabhatla and Klepper, 2014). We have collected
historical data from contemporary US Census Bureau’s Current Industrial Report series, “Computers and Office and
Accounting Machines” (annually): domestic shipments, imports, and exports. Using these data, we report in Figure 1
total sales of all portables sold in the United States during this period. What these data show is that the market for
portable computers only started to develop in the mid-1990s, which corresponds to our period of analysis. This
strengthen our belief that the period chosen is the relevant period during which fundamental design issues, possibly
associated with reverse salient, were being addressed and solved.
We apply hedonic regression methods to this data set to estimate whether positive prices are paid for product so-
lutions to the ergonomic and technological reverse salients. The hedonic regression method recognizes that heteroge-
neous goods can be described by their attributes or “characteristics.” This conceptualization follows a long tradition
of work in marketing, decision science, and economics (Court, 1939; Stone, 1956; Griliches, 1957, 1971; Lancaster,
1966, 1971; Green and Wind, 1973; Rosen, 1974).
The hedonic price model posits that a product comprises a set of inherent attributes, or “characteristics” that are
attractive to consumers. Hedonic functions are envelopes that involve both supply and demand factors (Rosen,
1974). Estimated coefficients are estimates of the prices of individual product characteristics, otherwise known as
shadow prices, which depend on both users’ valuations and producers’ costs (Triplett, 2006: 200).
It is important to note that this is an equilibrium model. The prices offered by firms on the market reflect the
underlying marginal costs of producing a set of K characteristics. Ceteris paribus, marginal costs are higher for a firm

Figure 1. Sales data on desktop and portable computers (1978–2006).

The importance of ergonomic design 7

offering a higher quantity of a particular characteristic. In equilibrium, the marginal cost of producing a characteris-
tic with a particular quantity is equal to the marginal benefit which consumers’ receive (Epple, 1987).
Prices (p) of laptops can, therefore, be expressed as a set of ergonomic (E) and technological (T) characteristics:

p ¼ f ðE; TÞ (1)

Rosen (1974) showed that the hedonic regressions identify equilibria intersections between the production possibility
frontiers of producers with varying production technologies and the indifference curves of consumers with varying
tastes. The hedonic price function is derived by taking the first partial derivative of (1). The partial derivative pro-
vides a set of “implicit shadow prices,” or “characteristic prices.” For an existing set of production possibility curves,
the implicit shadow price for a characteristic is the price paid for a marginal improvement in the quantity of one char-
acteristic, holding all other characteristics constant (Griliches, 1971; Pakes, 2003).
The hedonic function is estimated by regression analysis. We consider a differentiated product market in which
i ¼ 1; . . . ; I laptops are sold in t ¼ 1; . . . ; T periods. The consumer demand price pti of laptop i in period t is a
function of a fixed number (K) characteristics, over which our data provide information on differences in the lev-
els, or quantities, of these characteristics ztik . Using data on these variables for the period t, . . ., T, we estimate:

X
K
pti ¼ bt0 þ btk ztik þ eti (2)
k¼1

where eti is a random error term (independent and identically distributed).

The estimated coefficients b are the shadow prices for each of the K product characteristics, ceteris paribus. In our
estimated hedonic model, we include ergonomic characteristics in addition to the contribution of technological
characteristics.
Saviotti and Metcalfe (1984) extended the hedonic framework to consider the relationship technology and the ser-
vice characteristics that are valued by consumers (Saviotti, 1985). Firms compete by offering particular combinations
of service characteristics they believe will be more attractive to consumers than those of their rivals. These combin-
ations of “service characteristics” are related to a set of “technical characteristics,” which are directly related to the
underpinning technologies on which the products are based.
Ordinary Least Squares (OLS), based on the mean of all variables, may not be the most appropriate approach to
capture trade-offs between particular sets of product characteristics. For this reason, we also estimate a set of quan-
tile models, and conduct further robustness analysis using principle component analysis to establish the strength of
the interrelationship between the ergonomic variables, and those technological variables that govern processing
power.
There are two methods for estimating hedonic regression models: the time dummy variable model (TVDM) and
the adjacent period model. We will use the TVDM model which involves pooling observations for a number of years
and including a set of period dummies. The advantage of pooling is that larger number of observations provides
greater degrees of freedom. Pooled models are reliable when short periods are considered, and the dimensions of the
characteristics space are fixed, i.e., completely new characteristics are not introduced during the period under consid-
eration (Requena-Silvente and Walker, 2006). As discussed above, our data set meets both criteria.

5. Hypotheses
Generally speaking, if firms’ designers are effectively tackling a reverse salient in their products, then we expect the
interaction term between the characteristics associated with the reverse salient to be statistically significant and that
the estimated coefficient of the interaction term to be positive. If this were not the case and, alternatively, the esti-
mated coefficient is negative, then it would indicate ineffective design solutions with designers failing to successfully
address the reverse salient within their products. Specifically, in the case of the ergonomic reverse salient, a positive
interaction effect between screen size and weight indicates that users value more weight if the increased weight is ef-
fectively exploited to provide a larger screen size, that is, to overcome the reverse salient.

Hypothesis 1. A positive characteristics price is paid for the interaction between screen size and weight in laptop products
8 P. Windrum et al.

By similar reasoning, we expect there to be a positive coefficient reflecting effective solutions to the technological re-
verse salient between microprocessor clock speed, (RAM), and hard disk capacity. An increase in the value of each of
these characteristic will be higher if accompanied by a balanced improvement in the other two characteristics. Thus,
the two-way interactions as well as the three-way interaction effects are expected to be positive.

Hypothesis 2. A positive characteristics price is paid for the interaction between processing power, RAM and hard disk capacity
in laptop products

6. Data and model specification

Our data set is collected from information published in the UK consumer magazine WhatPC? This is a well-known,
reputable, and publicly available source for secondary data. As a data source, it offers a number of advantages. First,
the data are consistent and complete. Second, the use of an independent, publicly available source enables other re-
searchers to access the same information to replicate results. WhatPC? was a consumer magazine that produced an
annual “Buyers Guide” listing makes, models, recommended retail prices, and features. In total, 746 models are
listed in the Buyers Guides between 1993 and 1996, produced by 83 independent, competing manufacturers.
The dependent variable list_price (1993) is created to account for inflation. Listed model prices are deflated using
the official UK deflator, with 1993 as the base period. The data set contains eight independent ergonomic and
technological characteristic variables. The ergonomic characteristics are screen_area (length  width of screen) meas-
ured in square centimeter; weight (the total weight of each laptop) measured in kilograms; and height (the height of
the base unit) in centimeter. We expect the demand price for height to be curvilinear. Higher base units allowed
larger disk drive units to be installed, but increased base unit height makes a portable bulky and more difficult to
carry, and requires more space or storage. Therefore we include height and height2 in the estimated regressions.
Following Chwelos (2003), we use a rich set of characteristics that together affect computing power. These are
clock_speed (microprocessor speed) measured in megahertz; memory (cache speed or RAM) measured in kilobytes;
and harddisk (hard disk capacity) in megabytes. We expect the demand price for memory to be curvilinear. Some
firms at this time offered, for an additional upgrade price, with double the RAM. We therefore include memory and
memory2 in the estimated regressions.
We also have information on graphics cards. At this time some products in the data set came with lower-quality
color graphics adaptor (CGA) cards, while others offered higher-quality video graphics adaptor (VGA) cards. This
dichotomous variable vga takes a value of 1 if a portable is loaded with VGA graphics card or a value of 0 if it has a
CGA card.
Consumers were also offered a choice between monochrome displays, which were easier and cheaper to produce,
and color displays. color is a dichotomous variable which takes a value of 1 if a portable has a color screen or a value
of 0 if it has a monochrome screen. One would expect consumers to pay higher prices for higher-quality graphics
cards and for color displays. Note that the variables vga and color are independent of screen size.
Our data set includes two control variables: year and firm names. In hedonic price regressions, time and firm vari-
ables are commonly used to control for omitted variables. Time dummies are proxies for omitted market effects.
Since Chow (1967), empirical studies of computers generally include year dummies to control for the Moore’s law
doubling of processing capacity (on circuit boards of a given size and weight) every 18 months (Moore, 1965). As dis-
cussed, this will also pick up the effect of miniaturization in disk drives in the period 1993–1996. 1993 is taken as
the base year, so estimated coefficients for the dummies year94, year95, and year96 are differentials relative to this
base year.
Firm name dummies control for unobserved quality and hardware product features that are additional to our core
set of ergonomic and technological characteristics. These firm name dummies may additionally pick up brand equity
among manufacturers that are able to charge above-average prices for products with the same quality of characteris-
tics as their rivals (see previous studies by Keller, 1993; Ragaswami et al., 1993; Park and Srinivasan, 1994; Berndt
and Rappaport, 2001; and Windrum, 2005). There are a total of 83 firm dummies.5 Peacock is randomly selected as
the base firm.

5 The firm dummies are Acer, AJP, Akhter, Ambra, Amstrad, Apricot, Aria, Aries, AST, Atomstyl, Beltron, Carrera, Centerpr,
CIC, Colossus, Comcen, Compaq, CompuAdd, Compusys, Copam, DCS, DEC, Dell, Delta, Dimension, Dolch, Dual, Elonex,
The importance of ergonomic design 9

We estimate the hedonic model using OLS regression,

pi ¼ b0 þ b1 screen area þ b2 weight þ b3 screen area  weight

þ b4 height þ b5 height2 þ b6 clockspeed þ b7 memory þ b8 memory2
þ b9 harddisk þ b10 clockspeed  memory  harddisk (3)
þ b11 clockspeed  memory þ b12 memory  harddisk
þ b13 clockspeed  harddisk þ b14 colour þ b15 VGA þ controls þ i

If firms are tackling the ergonomic reverse salient effectively (H1), then we expect the interaction term b3 to be statis-
tically significant and that the estimated price for these solutions is positive.
Similarly, if firms are effectively tackling the technological reverse salient that determines processing power (H2),
then we expect the three-way interaction term b10 and the two-way interaction terms b11 , b12 , and b13 , to be statistic-
ally significant and positive.

6.1 Testing for omitted variables

An important concern for any estimated model is misspecification due to omitted variables. There is not a single test
for omitted variables. We shall follow current best practice and perform a number of tests on the saved residuals of
our estimated models. A well-specified model has a distribution of residuals that is normal (Gaussian). Alternatively,
a distribution of residuals that is nonnormal (non-Gaussian) indicates model misspecification. We will inspect the dis-
tribution visually the kernel density of the estimated residuals using standardized normal probability and quintile–
normal plots.
A second test is the Shapiro–Wilk W test. This is a non-graphical test for normality of the residuals, and is appro-
priate for sample sizes between 50 and 2000. A median value of W ¼ 1 indicates the saved residual samples are nor-
mally distributed.
The third test we shall employ is the Ramsey RESET test statistic. This is a test for functional misspecification of
the independent variables included in a model. It tests whether higher-order terms of these variables are significant. It
cannot pick up the influence of other (omitted) variables.

6.2 Robustness
We conduct two types of robustness check. First, quantile methods are applied to the data. In effect, we rerun the
three estimated models for the median priced portable at the 50th percentile of the price distribution. Quantile regres-
sion is a semi-parametric method. The conditional quantile has a linear form but does not impose a set of assump-
tions regarding the conditional distribution, and minimizes the weighted absolute deviations to estimate conditional
quantile (percentile) functions (Koenker and Bassett, 1978; Koenker and Hallock, 2001). For the median (50th per-
centile), symmetric weights are used. By contrast, classical OLS regression minimizes the sums of squared residuals to
estimate models for conditional mean functions.
The issue of heteroskedasticity in standard errors is dealt with using Gould’s bootstrapping procedure (Gould,
1992; Gould, 1979). Standard errors are obtained via 1000 replications of a panel bootstrap. This is drawn using a
fixed initial seed that is 1001, with each individual bootstrapped sample containing the same number of observations
as the original sample. The software used in all our estimations is Stata 12.6
A second robustness check is to apply principal components analysis (PCA) to examine the underlying structure
of interdependencies between variables. The expectation is that strong correlations between the ergonomic character-
istics of screen size and total weight on the one hand, and on the other, product characteristics which together govern
computing power. PCA is an established procedure for identifying the structure of linear relationships among interre-
lated variables. The procedure dates back to Ahamad (1967, 1968), and has been previously been applied in research

Ergo, Escom, Evesham, Gateway, Goldstar, Haval, HiGrade, HP, IBM, ICL, IPC, KT, Leo, Librex, Locland, Maple, Mesh,
Mitac, MJN, Munn, NCR, NEC, Obodex, Olivetti, Olympia, Omega, Opti, Opus, Pacific, Panasonic, Paragon, Peacock,
Redstone, Reeves, Rock, Sanyo, Samsung, Sharp, Sherry, Suntec, Siemens, TA, Tandon, Tandy, TI, Toshiba, Trigem,
Triumph, Tulip, Twinhead, Veridata, Viglen, Vortec, Wyse, and Zenith.
6 http://www.stata.com/stata12/.
10 P. Windrum et al.

Table 1. List of variables

Variable Description

list_price (1993) Listed model prices. Deflated using the official UK deflator, with 1993 as the base period. Dependent variable
screen_area Length  width of laptop screen. Measured in square centimeter. Independent ergonomic variable
weight Total weight of laptop. Measured in kilograms. Independent ergonomic variable
height Height of the base unit. Measured in centimeter. Independent ergonomic variable
clock_speed Microprocessor speed. Measured in megahertz. Independent technological variable
memory Cache speed (or RAM). Measured in kilobytes. Independent technological variable
harddisk Hard disk capacity. Measured in megabytes. Independent technological variable
color Dummy variable ¼ 1 if a laptop has a color screen. Variable ¼ 0 if it has a monochrome screen. Independent
technological variable
vga Dummy variable ¼ 1 if a laptop is loaded with VGA graphics card. Variable ¼ 0 if it has a CGA card.
Independent technological variable
firm Firm dummies: Acer, AJP, Akhter, Ambra, Amstrad, Apricot, Aria, Aries, AST, Atomstyl, Beltron, Carrera,
Centerpr, CIC, Colossus, Comcen, Compaq, CompuAdd, Compusys, Copam, DCS, DEC, Dell, Delta,
Dimension, Dolch, Dual, Elonex, Ergo, Escom, Evesham, Gateway, Goldstar, Haval, HiGrade, HP, IBM,
ICL, IPC, KT, Leo, Librex, Locland, Maple, Mesh, Mitac, MJN, Munn, NCR, NEC, Obodex, Olivetti,
Olympia, Omega, Opti, Opus, Pacific, Panasonic, Paragon, Peacock, Redstone, Reeves, Rock, Sanyo,
Samsung, Sharp, Sherry, Suntec, Siemens, TA, Tandon, Tandy, TI, Toshiba, Trigem, Triumph, Tulip,
Twinhead, Veridata, Viglen, Vortec, Wyse, and Zenith. Peacock is the base firm. Control variable
year Year dummies for 1994, 1995, and 1996. 1993 is the base year. Control variable

on the product characteristics of aeroplanes and helicopters (Saviotti, 1996), cameras (Windrum, 2005), and tanks
(Castaldi et al., 2009).
A set of distinct “components” (each comprising a set of interrelated variables) is estimated using the varimax ro-
tation method with Kaiser normalization. Compared to other clustering techniques, such as factor analysis, PCA
does not make strong prior assumptions regarding the extent and the structure of interdependencies among the ori-
ginal set of variables (Stevens, 1992). A further advantage is that one has a clear understanding of the number of re-
strictions that are used to calculate the principal components. PCA assesses the number of composite variables
required to achieve a sound representation of the original set of variables. Kaiser and Jolliffe criteria retain compo-
nents that have, respectively, eigenvalues greater than 1 or 0.7.

7. Results
7.1 Descriptive results
Table 1 provides the overview of the variables and their definitions, and Table 2 provides the estimated partial correl-
ation coefficients for list price (1993) and the eight product characteristics, together with descriptive data on the me-
dian, mean average, standard deviation, and minimum and maximum values.
The mean average price of £1779.68 (£5924.98 in current prices) is a reminder of just how expensive portable
computers were during the mid-1990s. As discussed, these were business machines, almost exclusively business ex-
ecutives and field sales staff. The cheapest listed model is £595.00 (£1980.90 in current prices), while the most expen-
sive is £6300.00 (£20,974.30 in current prices).7
The mean screen area (length  height) is 452 cm2 (which is approximately the area of a 10-inch  7-inch screen).
This is notable, as it just exceeds the minimum ergonomic size standards for a display intended to be viewed between
30 and 60 cm (see above).
The mean weight of laptops in our data set is 3 kg (6.5 lbs), the lightest model being 1 kg (2.2 lbs), and the heav-
iest 9 kg (20 lbs), highlighting the significant weight of some laptops in the data set.
The partial correlations reported in Table 1 indicate that strong correlations exist between the ergo-
nomic variables. There are positive partial correlations between screen size and weight in Columns 1 and 2,

7 Calculations use the UK consumer price index deflator.

The importance of ergonomic design 11

Table 2. Medians, means, standard deviations, minimum, maximum, and partial correlation coefficients

Variable Median Mean Standard Minimum Maximum 1 2 3 4 5 6 7 8 9

deviation

1. list_price 1544.27 1779.68 878.87 595.00 6300.00 1

(1993)
2. screen_area 626.90 652.88 135.11 84.56 1489.32 0.17*** 1
3. weight 2.90 3.08 1.12 1.00 9.00 0.08* 0.76*** 1
4. height 49.50 53.25 27.07 4.80 355.60 0.13*** 0.46*** 0.75*** 1
5. clock_speed 33.00 43.16 24.14 8.00 133.00 0.03 0.03 0.01 0.09*** 1
6. memory 4096.00 4334.11 2015.62 1024.00 20480.00 0.27*** 0.09* 0.03 0.06* 0.21*** 1
7. harddisk 120.00 192.35 161.82 1.00 1000.00 0.05 0.05 0.03E-1 0.09** 0.55*** 0.62*** 1
8. color 0 0.34 0 1 0.27*** 0.03 0.05E-1 0.08E-1 0.12*** 0.11*** 0.51*** 1
9. vga 1 0.99 0 1 0.04 0.03E-1 0.06* 0.03 0.05 0.10*** 0.08*** 0.03 1

Note. N¼744; ***P<0.01; **P<0.05; *P<0.10.

10
8
Weight (Kg)
6
4
2
0

0 500 1000 1500

Screen size (cm)

Fitted values

Figure 2. Scatter plot of weight against screen size, with fitted prediction line.

which are statistically significant at the 1% level. Portable computers with larger screen size tend to be
heavier in weight due to the larger and more powerful batteries required to support the screen. The scatter
plot of Figure 2 indicates a positive correlation between the weight of the portables in the data set and their
screen size.
In Columns 5 and 6, we see strong partial correlations between the technology variables which together determine
computer processing power. These estimates provide support for H1 and H2 of a distinct ergonomic reverse salient
and a distinct reverse salient in computing power.

7.2 Estimated OLS models

Table 3 presents information on three estimated (OLS) hedonic price models. BoxCox tests of functional form indi-
cate that the log of list price—Log_list_price (1993)—is the correct specification for these models. The log-linear
Model 1 does not contain interactions between ergonomic and computing power variables, or firm dummies. Model
2 includes and tests interactions between the ergonomic variables screen_area and weight (screen_area*weight), and
between the computing power variables clock_speed, memory, and harddisk (clock_speed*memory*harddisk). Since
the latter interaction comprises three variables, a fully specified model also includes pairwise interactions between
clock_speed*memory and memory* harddisk. Model 3 adds the set of firm dummies.
Models 2 and 3 support H1 that positive shadow prices are paid for designs that tackle the ergonomic reverse sali-
ent by addressing the interaction between weight and screen area. The estimated coefficient for screen_area*weight
12 P. Windrum et al.

Table 3. Estimated OLS model for consumers’ willingness to pay for product characteristics

Variables Dependent variable: Log_list_price (1993) Base year: 1993

Model 1 Model 2 Model 3

Coefficient Robust S.E. Coefficient Robust S.E. Coefficient Robust S.E. Standardized
coefficient

screen_area 0.00063*** (0.00016) 0.00003 (0.00023) 0.00005 (0.00022) 0.01477

weight 0.02057 (0.02642) 0.12375*** (0.04031) 0.08556** (0.04201) 0.21836
height 0.00729*** (0.00198) 0.00542*** (0.00187) 0.00545*** (0.00194) 0.33699
height2 0.00001** (0.61e-5) 0.00001* (0.536e-5) 0.00001* (0.565e-6) 0.16567
clock_speed 0.00413*** (0.00081) 0.00645 (0.00215) 0.00325* (0.00194) 0.17913
memory 0.00014*** (0.00002) 0.00011*** (0.00003) 0.00009*** (0.00003) 0.43119
memory2 5.46e-9*** (0.960e-9) 6.32e-9*** (0. 791e-9) 0.727e-8*** (0.854e-9) 0.45230
harddisk 0.00058*** (0.00013) 0.00026 (0.00071) 0.00013 (0.00025) 0.04744
color 0.31049*** (0.02906) 0.31287*** (0.02803) 0.28605*** (0.02645) 0.30902
vga 0.21079* (0.11498) 0.26025*** (0.09890) 0.28560*** (0.10874) 0.06293
screen_area*weight 0.00013*** (0.00004) 0.00013*** (0.00005) 0.42150
clock_speed*memory*harddisk 0.204e-8*** (0.076e-9) 0.201e-8*** (0.656e-9) 0.58919
clock_speed*memory 0.783e-6** (0.468e-6) 0.894e-6** (0.436e-6) 0.43746
memory* harddisk 0.204e-7*** (0.080e-7) 0.177e-7*** (0.658e-7) 0.59055
clock_speed*harddisk 0.205e-4*** (6.81e-6) 0.205e-4*** (6.81e-6) 0.70707
Control variables:
year94 0.14305*** (0.03617) 0.13340*** (0.03281) 0.13118*** (0.03350) 0.12440
year95 0.34256*** (0.03703) 0.34448*** (0.03388) 0.33458*** (0.03852) 0.33475
year96 0.76013*** (0.05066) 0.74644*** (0.04836) 0.72700*** (0.05350) 0.73217
firm dummies Yes
Constant 5.94*** (0.13) 6.34*** (0.18) 6.36*** (0.20)
AIC 342.30 227.71 215.99
BIC 402.25 349.74 347.63
N 744 744 744
F 64.18 49.32 39.24
Adjusted R2 0.63 0.71 0.78
Residual sum of squares 66.6 64.1 50.9
Ramsey RESET test F(3, 727) ¼ 2.15 F(3, 722) ¼2.21 F(3, 711) ¼2.55
P >F ¼ 0.09 P>F ¼ 0.02 P>F ¼0.05
Shapiro–Wilk test W 0.99511 0.99485 0.99579
(P¼0.018) (P¼0.0131) (P¼0.041)

Note. ***P<0.01; **P<0.05; *P<0.10.

is positive and statistically significant at the 1% level in both models. The estimated standardized coefficient indicates
the implicit price for an incremental improvement in this ergonomic interaction.
The inclusion of this interaction variable has a clear impact on the estimated coefficients for the individual
variables of screen_area and weight in Models 2 and 3 (without firm dummies and with firm dummies, respect-
ively). The coefficient for screen_area is statistically insignificant in Models 2 and 3, while in Model 1 (which
does not include the interaction variable) the estimated coefficient is significant at the 1% level. Also, the size of
the estimated coefficient is notably smaller in Models 2 and 3. The lower adjusted R2 of 0.63 for Model 1, com-
pared to 0.71 and 0.78 for Models 2 and 3, respectively, indicates that the model without this interaction is
misspecified.
By contrast, the estimated coefficients for weight in Models 2 and 3 are statistically significant (at the 5% level),
while in Model 1 the coefficient was not significant at P < 0.10. These findings indicate that simpler models, which
The importance of ergonomic design 13

Kernel density estimate

1.5
1
Density
.5
0

-1 -.5 0 .5 1
Residuals

Kernel density estimate

Normal density
kernel = epanechnikov, bandwidth = 0.0605

Figure 3. Kernel density estimate of saved residuals for Model 3.

omit this interaction, are misspecified and are misleading with regard to the underlying relationship between prices,
screen size, and weight.
In Models 2 and 3, the estimated coefficient of clock_speed*memory*harddisk is positive and statistically signifi-
cant at the 1% level. In addition, the two-way interaction effects for clock_speed*memory, memory* harddisk, and
clock_speed*harddisk are also positive and significant, further confirming the reverse salient hypothesis regarding
speed, RAM, and hard disk capacity. This supports H2 that positive shadow prices are paid for the interaction be-
tween these variables, which governs computing power.
Finally, we note that the coefficients for the control variables—year dummies and firm name dummies—are sig-
nificant in Models 2 and 3, respectively, and have the expected positive sign. Laptops with color screens are signifi-
cantly more expensive than those with monochrome screens, while the same holds true for laptops with VGA
graphics card instead of a CGA card.

7.3 Testing for omitted variables

As discussed, there is not a single test for omitted variables and so, following current best practice, we perform a
number of tests on the saved residuals of the estimated models to establish whether these are normally distributed.
Due to space constraints we report here tests on the saved residuals of Model 3, as this model includes the
hypothesized interactions between ergonomic variables and between computing power variables.
Figure 3 is a kernel density graph of the estimated residuals of Model 3. A normal distribution is superimposed on
the kernel density graph. The graph indicates that the residuals are normally distributed.
Figure 4 presents standardized normal probability (pnorm) plot and a quintile–normal (qnorm) plot of the saved
Model 3 residuals. The standardized normal probability plot is more sensitive to deviances near the mean of the dis-
tribution. The standardized normal probability plot for these residuals is ruler flat.
Quintile–normal plots quintiles of residuals vs. quintiles of a normal distribution, and is more sensitive to devi-
ances from normality in the tails of the distribution. Figure 5 indicates three data points as outliers (bottom left-hand
corner). Otherwise, the tails are close to normal.
The second test for model misspecification we apply is the Shapiro–Wilk W test. This is a non-graphical test for
normality, with a median value of W ¼ 1 indicating the saved residual samples are normally distributed. Table 3 re-
ports the Shapiro–Wilk W statistic for each of our estimated models. The critical P-values are indicated along with
the estimated W. The estimated W ¼ 0.99579 (P ¼ 0.041) for the saved residuals of Model 3. We cannot reject H0
(at P ¼ 0.05 level) that these residuals are normally distributed.
The third test we apply is Ramsey RESET test functional misspecification of the independent variables included in
a model. For Model 3, the estimated F statistic ¼ 2.55 (P ¼ 0.05) indicating that further powers of these independent
variables do not jointly add further explanatory power to this model.
14 P. Windrum et al.

1.00
0.75
Normal F[(r-m)/s]

0.50
0.25
0.00

0.00 0.25 0.50 0.75 1.00

Empirical P[i] = i/(N+1)

Figure 4. Standardized normal probability plot of Model 3 residuals.

1
.5
Residuals
0
-.5
-1

-1 -.5 0 .5 1
Inverse Normal

Figure 5. Quintile–normal plot of Model 3 residuals.

7.4 Robustness
The first of our robustness tests is to apply quantile estimation to this set of models. The findings for the median port-
able (50th percentile) in the price distribution are reported in Table 4 below. As with the estimated OLS models, the
inclusion of the interaction variable screen_area*weight is statistically significant in Model 5 (without firm dummies)
and Model 6 (with firm dummies). When the interaction term is included, the coefficient for screen_area is not statis-
tically significant in these models. By contrast, the coefficient is significant, in Model 4, when the interaction terms
are omitted. These findings indicate that consumers pay a shadow price for designs that tackle the ergonomic reverse
salient, and that models which omit this are misspecified.
We next turn to the principle components analysis (PCA) of the set of product characteristic variables in Table 5.
The PCA on these data identifies three distinct components that are orthogonal to one another. The first estimated
component is the set of product characteristics that comprise the computing power reverse salient: clock_speed,
memory, and harddisk. This component accounts for 36% of the variance across the independent variables. The
highest value in this component is harddisk (0.910), followed by clock_speed (0.878) and memory (0.831).
The second estimated component comprises the interrelated ergonomic product characteristics screen_area and
weight, and base unit height. This accounts for 27% of variance across all variables.
This further supports the proposition that strong interactions exist between the characteristics screen size and
weight that together comprise the ergonomic reverse salient, and are distinct to other product laptop characteristics.
The importance of ergonomic design 15

Table 4. Estimated quantile models for consumers’ willingness to pay for product characteristics

Variables Model 4 Model 5 Model 6

50th percentile 50th percentile 50th percentile

Price: £1542.45 Price: £1542.45 Price: £1542.45

Coefficient Robust S.E. Coefficient Robust S.E. Coefficient Robust S.E.

screen_area 0.00064*** (0.00020) 0.00004 (0.00032) 0.00001 (0.00030)

weight 0.00799 (0.03865) 0.11973* (0.06390) 0.09653 (0.06406)
height 0.00771** (0.00373) 0.00384 (0.00305) 0.00400 (0.00373)
height2 0.00001 (0.00001) 6.66e-6 (0.00001) 0.699e-5 (0.00001)
clock_speed 0.00330*** (0.00084) 0.00010 (0. 00004) 0.00532 (0.00337)
memory 0.00015*** (0.00002) 0.14e-3*** (0.23e-4) 0.15e-3*** (0.33e-4)
memory2 5.80e-9 (2.05e-9) 0.757e-8 (0.544e-8) 0.635e-8 (0.411e-8)
Harddisk 0.00066*** (0.00022) 0.00199* 0.00108 0.00184* (0.00089)
color 0.30495*** (0.03840) 0.31421*** (0.03435) 0.30873*** (0.03322)
vga 0.14064 (0.17674) 0.35329** (0.17081) 0.27477* (0.15420)
screen_area*weight 0.00016*** (0.00005) 0.00015*** (0.00006)
clock_speed*memory*harddisk 0.903e-9*** (0.227e-9) 0.157e-8*** (0.194e-9)
clock_speed*memory 0.109e-5 (0.864e-6) 0.605e-6 (0.748e-6)
memory* harddisk 0.405e-6*** (0.194e-6) 0.962e-6*** (0.163e-6)
clock_speed*harddisk 0.000023*** (0.00001) 0.00002** (0.00001)
Control variables:
year94 0.19464*** (0.04467) 0.20217*** (0.04341) 0.17171*** (0.03977)
year95 0.39004*** (0.05284) 0.41718*** (0.06311) 0.41182*** (0.04842)
year96 0.85007*** (0.06454) 0.86977*** (0.09025) 0.81846*** (0.06513)
firm dummies Yes
Constant 5.88*** (0.20) 6.39*** (0.29) 6.26*** (0.26)
N 744 744 744
Pseudo R2 0.62 0.68 0.70
Minimum sum deviations 88.7 80.9 77.8

The third estimated component comprises the characteristic variables color (0.698) and vga (0.991). These two
variables facilitate the rendition of high-quality color images. This estimated component for 12% of variance across
all variables.
The estimated correlation matrix on which the PCA is constructed is within the critical 1% level. The estimated
Kaiser–Meyer–Olkin statistic of sampling adequacy is 0.638, well above the critical 0.5 level.

8. Discussion and conclusions

Our research findings highlight the need for a deeper understanding, and analysis, of the contributions of ergonomic
design to product innovation. Studies that omit these contributions, i.e., focus solely on technologically driven func-
tional performance, significantly under-report innovation. We have shown that the design trajectory in portable com-
puters was strongly shaped by the ergonomic trade-off that exists between screen quality and total weight. Over the
course of time, portable computer designers have sought to improve usability by developing products with larger
screens, while simultaneously addressing the problem of increased weight as this negatively affects portability.
Our empirical analysis has applied hedonic price methods for studying trade-offs in product characteristics: how-
ever, we have extended the analysis to include the key ergonomic variables “screen size” and “weight” in addition to
key technology variables. These results indicate that designers have separately addressed the ergonomic reverse sali-
ent and the technological issue of computer power when engaging in product innovation. Importantly, the findings
indicate that consumers positively value the solutions to the ergonomic reverse salient that rival firms offer up to the
market.
16 P. Windrum et al.

Table 5. Principal components for independent product characteristics

Variables Retained principal components

Computing power reverse salient Ergonomic reverse salient Color

screen_area 0.086 0.796 0.052

weight 0.035 0.967 0.025
height 0.006 0.819 0.039
clock_speed 0.878 0.066 0.066
memory 0.831 0.079 0.115
harddisk 0.910 0.002 0.029
color 0.028 0.016 0.698
vga 0.090 0.029 0.991
Number of observations 746 746 746
Eigenvalues:
Total 2.909 2.173 0.969
% of Variance 36.357 27.164 12.110
Cumulative % of variance 36.357 63.521 75.630

Note. Kaiser–Meyer–Olkin measure of sampling adequacy 0.638.

Bartlett’s test of sphericity: approximate v2 2904.811.
DF: 28.
Significance: 0.000.
Rotation method: varimax with Kaiser normalization.
Rotation converged in four iterations.

The current research has a number of limitations. First, ours is a detailed study of an ergonomic reverse salient in
one particular product class: studies of ergonomic reverse salients in other product classes are required to establish
the generalizability of our findings. Second, our empirical frame has sought to minimize well-known problems associ-
ated with omitted variables in cross-sectional studies. Great care was taken to select a period for data collection: in
the period 1993–1996, the laptop computer was a stand-alone business product (i.e., pre-Internet), using a limited
set of business software that was highly standardized. The set of product characteristics found in the portables at this
time was also limited, certainly in comparison with later periods. Firm dummies capture brand effects and additional,
idiosyncratic product features offered by firms.
It is hoped that the research presented in this article will stimulate further research on ergonomic design, and the
effective management of ergonomic, aesthetic, and technological inputs to innovation. There is a clear need for fur-
ther research into ergonomics, and the role played by ergonomic reverse salients in shaping the innovation trajectory
of other product classes. Developing a set of stylized facts about the role of ergonomics not only requires a retesting
of the hypotheses advanced in this article but also the development of new research questions. For example, it is im-
portant to know whether the significance of ergonomic features, relative to technology and aesthetics, varies over the
product lifecycle, and if so, what factors explain this. The demands on data collection, and problems of omitted vari-
able bias, for longitudinal studies such as this are demanding but potentially highly rewarding.
The analysis presented in this article has focused on design ergonomics. This is driven by a need to redress an im-
balance in recent scholarship in design (Stoneman, 2010; Eisenman, 2013), which has expanded our understanding
of the role of aesthetics in innovation but paid little or no attention to the role of ergonomics. These two areas of de-
sign are not exclusive. While aesthetics does not play a key role in the development of portable computers during the
era studied, an important future avenue for research is the development of case studies in which both ergonomic and
aesthetic design, along with technology, are analyzed as shaping factors in the innovation process.
Finally, our empirical findings hold important implications for managers. Successful product management re-
quires an understanding of the role(s) of design within innovation, and how the inputs of designers complement the
technological inputs of R&D. Designers have much to offer in determining market positioning, understanding and
creating demand, and in addressing and unlocking the latent needs of consumers. As Moody (1980) has stated, man-
agers must address the opposition of R&D engineers to industrial designers. Long-term competitiveness requires the
strategic harnessing and integration of inputs from both designers and R&D engineers. Successful companies focus
The importance of ergonomic design 17

their product innovation activities along well-defined design trajectories that carry a company’s recognizable
signature.
Product design is expected to become increasingly decisive for a company’s competitive advantage. For example,
figures from the UK Design Council (2010) indicate a 15% growth in real earnings over the period 2005–2010, des-
pite that country’s economic downturn in 2008 and recession in 2009. This represents a major shift in the manage-
ment of product innovation, which in many sectors has been the preserve of the R&D department. Understanding
this shift and developing new ways to create value are important challenges for both managers and academic scholars
of innovation.

Acknowledgements
The authors would like to thank Paul Stoneman, Peter Swann, Steven Klepper, and two anonymous referees for their insightful com-
ments on previous versions of this article. In addition, the authors offer our gratitude for suggestions and comments made by those
attending presentations at the DRUID society conference, and research seminar presentations at Imperial College and at Manchester
Institute of Innovation Research.

References
Ahamad, B. (1967), ‘An analysis of crimes by the method of principal components,’ Applied Statistics, 16, 17–35.
Ahamad, B. (1968), ‘A note on the interrelationships of crimes—a reply to Mrs Walker,’ Applied Statistics, 17, 49–51.
Alba, J. W. and J. W. Hutchinson (1987), ‘Dimensions of consumer expertise,’ Journal of Consumer Research, 13, 411–454.
Alba, J. W. and J. W. Hutchinson (2000), ‘Knowledge calibration: what consumers know and what they think they know,’ Journal of
Consumer Research, 27, 123–156.
Baker, T. A. (1997), ‘Quality-adjusted price indexes for portable computers,’ Applied Economics, 29(9), 1115–1123.
Baldwin, C. Y. and K. B. Clark (2000), ‘Design rules,’ in The Power of Modularity, Vol. 1. MIT Press: Cambridge, MA.
Berndt, E. R., Z. Griliches and N. J. Rappaport (1995), ‘Econometric estimates of price indexes for personal computers in the 1990s,’
Journal of Econometrics, 68(1), 243–268.
Berndt, E. R. and N. J. Rappaport (2001), ‘Price and quality of desktop and mobile personal computers: a quarter-century historical
overview,’ American Economic Review, 91(2), 268–273.
Bhaskarabhatla, A. and S. Klepper (2014), ‘Latent submarket dynamics and industry evolution: lessons from the US laser industry,’
Industrial and Corporate Change 23(6), 1381–1415.
Bloch, P. H. (1995), ‘Seeking the ideal form: product design and consumer response,’ Journal of Marketing, 59(3), 16–29.
Boff, K. R. (2006), ‘Revolutions and shifting paradigms in human factors & ergonomics,’ Applied Ergonomics, 37(4), 391–399.
Brown, T. (2008), ‘Design thinking,’ Harvard Business Review, 86(6), 84–92.
Castaldi, C., R. Fontana and A. Nuvolari (2009), ‘Chariots of fire: the evolution of tank technology, 1915-1945,’ Journal of
Evolutionary Economics, 19(4),545–566.
Chow, G. (1967), ‘Technological change and the demand for computers,’ American Economic Review, 57, 1117–1130.
Chwelos, P. (2003), ‘Approaches to performance in measurement in hedonic analysis: price indexes for laptop computers in the
1990’s,’ Economics of Innovation and New Technology, 12(3), 199–224.
Court, A. T. (1939), ‘Hedonic price indexes with automotive examples’, in The Dynamics of Automobile Demand. General Motors
Corporation: New York, NY, pp. 99–117.
Cox, G. (2005), Cox Review of Creativity in Business: Building on the UK’s Strengths. HM Treasury: London.
Cross, N. (2001), ‘Designerly ways of knowing: design discipline versus design science,’ Design Issues, 17(3), 49–55.
Dedehayir, O. (2009), ‘Bibliometric study of the reverse salient concept,’ Journal of Industrial Engineering and Management, 2(3),
569–591.
Design Council (2010), Design Industry Research 2010, http://www.designcouncil.org.uk/our-work/insight/research/design-indus
try-research-2010/. accessed 15 January 2011.
Eisenman, M. (2013), ‘Understanding aesthetic innovation in the context of technological evolution,’ Academy of Management
Review, 38, 332–351. doi:10.5465/amr.2011.0262.
Epple, D. (1987), ‘Hedonic prices and implicit markets: Estimating demand and supply functions for differentiated products,’ Journal
of Political Economy, 95(1), 59–80.
Gatignon, H. and T. S. Robertson (1989), ‘Technology diffusion: an empirical test of competitive effects,’ Journal of Marketing, 53,
35–49.
Gould, W. W. (1992), ‘Quantile regression with bootstrapped standard errors,’ Stata Technical Bulletin, 9, 19–21.
Gould, W. W. (1979), ‘Interquantile and simultaneous-quantile regression,’ Stata Technical Bulletin, 38, 14–22.
18 P. Windrum et al.

Green, P. E., and Y. Wind (1973), Multiattribute Decisions in Marketing: A Measurement Approach. Dryden Press: Hinsdale, IL.
Griliches, Z. (1957), ‘Hybrid corn: an explanation of the economics of technological change,’ Econometrica, 25, 501–522.
Griliches, Z. (ed.) (1971), Price Indices and Quality Change: Studies in New Methods of Measurement. Harvard University Press:
Cambridge, MA.
Hauser, J. R. and V. R. Rao (2004), ‘Conjoint analysis, related modeling, and applications,’ International Series in Quantitative
Marketing, 1, 141–168.
Hainmueller, J., D. J. Hopkins and T. Yamamotoz.(2014), ‘Causal inference in conjoint analysis: understanding multi-dimensional
choices via stated preference experiments,’ Political Analysis, 22(1), 1–30.
Henderson, R. M., and K. B. Clark (1990), ‘Architectural innovation: the reconfiguration of existing product technologies and the
failure of established firms,’ Administrative Science Quarterly, 35(1), 9–30.
Hill, A. M. (1998), ‘Problem solving in real-life contexts: An alternative for design in technology education,’ International Journal of
Technology and Design Education, 8(3), 203–220.
Hughes, T. P. (1983), Networks of Power: Electrification in Western Society, 1880-1930. Johns Hopkins University Press: Baltimore,
MD.
Hughes, T. P. (1987), ‘The evolution of large technological system,’ in W. Bijker, T. P. Hughes and T. Pinch (eds), The Social
Construction of Technological Systems: New Directions in the Sociology and History of Technology. MIT Press: Boston, MA.
Karwowski, W. (2005), ‘Ergonomics and human factors: the paradigms for science, engineering, design, technology and management
of human-compatible systems,’ Ergonomics, 48(5), 436–463.
Keller, K. L. (1993), ‘Conceptualizing, measuring, and managing customer-based brand equity,’ Journal of Marketing, 57, 1–22.
Koenker, R. and K. F. Hallock (2001), ‘Quantile regression,’ Journal of Economic Perspectives, 15, 143–156.
Koenker, R. and G. Bassett (1978), ‘Regression quantiles,’ Econometrica, 46, 33–50.
Lancaster, K. J. (1966), ‘A new approach to consumer theory,’ Journal of Political Economy, 14, 133–146.
Lancaster, K. J. (1971), Consumer Demand: A New Approach. Columbia University Press: New York, NY.
Lawson, B. (2006), How Designers Think: the Design Process Demystified. Architectural Press: Oxford.
Lien, A., C. Cai, R. A. John, E. Galligan and J. Wilson.(2001), ‘16.300 QSXGA high resolution wide viewing angle TFT-LCDs based
on ridge and fringe-field structures,’ Displays, 22(1), 9–14.
Linden, D. and T. B. Reddy (2001), Handbook of Batteries, 3rd edn. McGraw-Hill: New York, NY.
Mayhew, D. J. (1999), The Usability Engineering Lifecycle. Academic Press: San Diego, CA.
Miles, I., and L. Green (2008), Hidden Innovation in the Creative Industries. NESTA: London.
Moenaert, R. K. and W. E. Souder (1990), ‘An information transfer model for integrating marketing and R&D personnel in NPD pro-
jects,’ Journal of Product Innovation Management, 7(2), 91–107.
Moody, S. (1980), ‘The role of industrial design in technological innovation,’ Design Studies, 1(6), 329–339.
Moore, G. E. (1965), ‘Cramming more components onto integrated circuits,’ Electronics, 38, 5. www.intel.com/research/silicon/more
slaw.htm.
Murmann, J. P. and K. Frenken (2006), ‘Towards a systematic framework for research on dominant designs. technological innova-
tions, and industrial change,’ Research Policy, 35(7), 925–952.
Nelson, R. A., T. L. Tanguay and C. D. Patterson (1994), ‘A quality-adjusted price index for personal computers,’ Journal of Business
and Economic Statistics, 12(1), 23–31.
Pakes, A. (2003), ‘A reconsideration of hedonic price indexes with an application to PC’s,’ American Economic Review, 93(5),
1578–1596.
Park, C. S. and V. Srinivasan (1994), ‘A survey-based method for measuring and understanding brand equity and its extendability,’
Journal of Marketing Research, 31, 271–288.
Perks, H., R. Cooper and C. Jones (2005), ‘Characterizing the role of design in new product development: an empirically derived tax-
onomy,’ Journal Product Innovation Management, 22(2), 111–127.
Postrel, V. (2003), The Substance of Style: How the Rise of Aesthetic Value Is Remaking Commerce, Culture, and Consciousness.
HarperCollins: New York.
Ragaswami, A., R. R. Burke and A. T. Olivia (1993), ‘Brand equity and the extendibility of brand names,’ International Journal of
Research in Marketing, 10, 61–75.
Requena-Silvente, F. and J. Walker (2006), ‘Calculating hedonic price indices with unobserved product attributes: an application to
the UK car market,’ Economica, 73, 509–532.
Rosen, S. (1974), ‘Hedonic prices and implicit markets: product differentiation in pure competition,’ Journal of Political Economy,
82(1), 34–55.
Rosenberg, N. (1969), ‘The direction of technological change: inducement mechanisms and focusing devices,’ Economic
Development and Cultural Change, 18(1), 1–24.
Safire, W. (1988), ‘On language; sit on my laptop’, New York Times, October 23
Salvendy, G. (ed) (2012), Handbook of Human Factors and Ergonomics, 4th edn. Wiley and Sons: New Jersey.
The importance of ergonomic design 19

Sanders, M. M., and E. J. McCormick (1993), Human Factors in Engineering and Design, Chapter 4, 7th edn. McGraw-Hill: New
York, NY, pp. 91–117.
Saviotti, P. P. and J. S. Metcalfe (1984), ‘A theoretical approach to the construction of technological output indicators,’ Research
Policy, 13, 141–151.
Saviotti, P. P. (1985), ‘An approach to the measurement of technology based on the hedonic price method and related methods,’
Technological Forecasting and Change, 29, 309–334.
Saviotti, P. P. (1996), Technological Evolution, Variety and The Economy. Edward Elgar: Cheltenham.
Schon, D. A. (1983), The Reflective Practitioner: How Professionals Think in Practice. Basic Books: New York, NY.
Shneiderman, B. (1992), Designing the User Interface: Strategies for Effective Human-Computer Interaction, Second Edition,
Addison-Wesley: Reading, MA.
Stevens, J. (1992), Applied Multivariate Statistics for The Social Sciences. Lawrence Erlbaum: New Jersey.
Stone, R. (1956), Quantity and Price Indexes in National Accounts. OECD: Paris.
Stoneman, P. (2010), Soft Innovation. Economics, Product Aesthetics, and the Creative Industries. Oxford University Press: Oxford.
Suh, N. P. (2001), Axiomatic Design: Advances and Applications. Oxford University Press: Oxford, UK.
Triplett, J. E. (2006), Handbook on Hedonic Indexes and Quality Adjustments in Price Indexes: Special Application to Information
Technology Products, OECD: Paris.
US Military Standard (1999), United States Military Standard: Department of Defense Design Criteria Standard. Mil-Std-1472F,
Boff K. and J. Lincoln (eds), U.S. Department of Defense: Washington DC.
van Schalwijk, W.A. and B. Scrosati (eds) (2002), Advances in Lithium-Ion Batteries. Kluwer Academic/Plenum: London.
Verganti, R. (2009), Design Driven Innovation: Changing the Rules of Competition by Radically Innovating What Things Mean.
Harvard Business Press: Boston, MA.
von Hippel, E. (1986), ‘Lead users: a source of novel product concepts,’ Management Science, 32(7), 791–806.
Ware, C. (2004), Information Visualization: Perception for Design, 2nd edn. Morgan Kaufmann: San Mateo, CA.
Wilson, J. E. (2006), Vintage Laptop Computers: First Decade 1980-89. Outskirts Press: Colorado.
Windrum, P. (2005), ‘Heterogeneous preferences and new innovation cycles in mature industries: the amateur camera industry 1955–
1974,’ Industrial and Corporate Change, 14(6), 1043–1074.
Woolgar, S. (1991), ‘Configuring the user: the case of usability trials,’ in J. Law (ed.), A Sociology of Monsters: Essays on Power,
Technology and Domination. Routledge: London.
Woolgar, S. (1994), ‘Rethinking the dissemination of science and technology’, CRICT Discussion Paper. CRICT (Centre for Research
into Innovation, Culture and Technology): Brunel University.
 Brand Meaning Management
Brand Remixing: 3D Printing the Nokia Case
Aric Rindfleisch Matthew O’Hern
Article information:
To cite this document: Aric Rindfleisch Matthew O’Hern . "Brand Remixing: 3D
Printing the Nokia Case" In Brand Meaning Management. Published online: 05 May
2015; 53-81.
Permanent link to this document:
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

http://dx.doi.org/10.1108/S1548-643520150000012003
Downloaded on: 15 February 2016, At: 20:01 (PT)
References: this document contains references to 0 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 69 times since NaN*
Users who downloaded this article also downloaded:
Eda Sayin, Zeynep Gürhan-Canl#, (2015),"Feeling Attached to Symbolic Brands within
the Context of Brand Transgressions", Review of Marketing Research, Vol. 12 pp.
233-256 http://dx.doi.org/10.1108/S1548-643520150000012009
Allison R. Johnson, Matthew Thomson, Jennifer Jeffrey, (2015),"What Does Brand
Authenticity Mean? Causes and Consequences of Consumer Scrutiny toward a Brand
Narrative", Review of Marketing Research, Vol. 12 pp. 1-27 http://dx.doi.org/10.1108/
S1548-643520150000012001
Pankaj Aggarwal, Megha Agarwal, (2015),"Linear versus Step-Function Decision
Making: The Moderating Role of Relationship Norms on Consumer Responses to
Brand Transgressions", Review of Marketing Research, Vol. 12 pp. 207-232 http://
dx.doi.org/10.1108/S1548-643520150000012008

Access to this document was granted through an Emerald subscription provided by

emerald-srm:149425 []
For Authors
If you would like to write for this, or any other Emerald publication, then please
use our Emerald for Authors service information about how to choose which
publication to write for and submission guidelines are available for all. Please visit
www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society.
The company manages a portfolio of more than 290 journals and over 2,350 books
and book series volumes, as well as providing an extensive range of online products
and additional customer resources and services.
 Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner
of the Committee on Publication Ethics (COPE) and also works with Portico and the
LOCKSS initiative for digital archive preservation.

*Related content and download information correct at time of download.

Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)
 BRAND REMIXING: 3D PRINTING
THE NOKIA CASE
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Aric Rindfleisch and Matthew O’Hern

ABSTRACT

Purpose To identify, conceptualize, and analyze a newly emerging

form of consumer-initiated, brand-altering activity that we term “brand
remixing.”
Methodology A content analysis of 92 remixes of the Nokia Lumia
820 smartphone case.
Findings We find that nearly 40% of the remixed versions of Nokia’s
case retained at least one element of its standard template. The remixed
cases contained considerable congruency with the design elements in the
standard template, a high degree of personalization, and no negative
brand imagery.
Implications Our research is the one of the first examinations of the
role of 3D printing upon marketing activities. It has important implica-
tions for marketing scholarship by showing that 3D printing empowers
consumers to physically alter the brands they consume. Our research
also suggests that practitioners interested in using this technology to
develop and enhance their brands should accept the notion that firms are
no longer fully in control of their brand assets. Hence, we believe that

Brand Meaning Management

Review of Marketing Research, Volume 12, 53 81
Copyright r 2015 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1548-6435/doi:10.1108/S1548-643520150000012003
53
 54 ARIC RINDFLEISCH AND MATTHEW O’HERN

brand managers should develop co-creation platforms that allow custo-

mers to easily modify, remix, and share various aspects of their brands
with their peers.
Originality We identify and label an important emerging branding
practice (i.e., brand remixing). This practice has the potential to drama-
tically alter the branding landscape.
Keywords: 3D printing; Thingiverse; branding; brand community;
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

co-creation; remixing

Brands are powerful assets that are carefully cultivated, tightly controlled,
and strongly protected by the companies that create them. For instance,
Coca Cola’s brand managers have systematically crafted the Coke brand
using consistent design elements that are nearly a century old (i.e., red and
white color scheme, signature cursive lettering, classic contour bottle) and
have protected the brand by keeping its secret formula locked in a vault in
Atlanta since 1925. As a result of this consistent and careful management,
Coca Cola’s brand value is estimated to be worth nearly $80 billion
(Interbrand, 2013). Although few brands have achieved Coke’s level of suc-
cess, most companies follow a similar strategic approach in terms of treat-
ing their brands as assets that must be carefully managed by executives
within the firm (Aaker, 2009). In addition to being a popular managerial
practice, this top-down approach to brand management is also strongly
endorsed by a number of prominent branding scholars (Aaker, 2009;
Keller, 2012). For example, Keller’s (2000) influential Brand Report Card
grades brands based on their success in terms of “building and properly
managing brand equity” (p. 147). Thus, this strategic branding paradigm is
widely accepted by marketing scholars and practitioners alike.
Although this type of strategic brand management may be conventional
wisdom, a growing chorus of voices has begun to question this gospel.
Thus far, the loudest dissent has come from anti-branding advocates such
as Lasn (1999) and Klein (1999), who view strategic branding efforts as lar-
gely manipulative and disingenuous. These advocates recommend that con-
sumers resist brands either passively (by refusing to purchase particular
brands) or actively (by manipulating various brand elements). A popular
example of this type of resistance is Adbuster’s Joe Chemo posters, which
 Brand Remixing: 3D Printing the Nokia Case 55

depict Camel cigarette’s Joe Camel sadly lying in a hospital bed with an
intravenous drip in his arm. These forms of anti-brand resistance have
recently gained the attention of a growing number of branding scholars,
who document how consumers either avoid or alter a firm’s strategic
branding efforts by engaging in boycotts (Kozinets & Handelman, 2004),
using brands in idiosyncratic ways (Holt, 2002), and crafting oppositional
brand imagery (Thompson, Rindfleisch, & Arsel, 2006). Thus, a growing
number of consumers are no longer content to simply choose and use
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

brands in the form in which they are offered; instead, they wish to play a
more active role in crafting the brand itself.
Our paper seeks to contribute to this growing body of research by identi-
fying, conceptualizing, and documenting a newly emerging form of
consumer-initiated, brand-altering activity that we term “brand remixing.”
Borrowing from Lessig (2008), we define brand remixing as a process of
taking elements from an existing brand and changing or recombining them in
a manner that substantially alters the brand. Brand remixing bears some
similarity to Thompson et al.’s (2006) doppelganger brand image (i.e.,
disparaging brand images created and circulated by anti-brand activists).
For example, propagators of both a doppelganger brand image and a
remixed brand often seek to share their modifications with others.
However, these two concepts also have some notable differences. First,
brand remixing is not necessarily a form of consumer resistance and may,
in fact, be positive in nature. Second, brand remixing activities are not lim-
ited to the message and/or visual images surrounding a brand but may also
include other elements such as a brand’s physical features. This physical
aspect of brand remixing is the focus of our paper.
A good example of what we mean by brand remixing is the recent case
of FedEx Furniture, in which a young college student (i.e., Jose Avila) used
discarded FedEx boxes to construct furniture for his apartment, including
a bed, sofa, and coffee table and shared these remixes with others via his
(now defunct) website, FedExFurniture.com (Berthon, Pitt, McCarthy, &
Kates, 2007). At present, physical brand remixing is the exception rather
than the norm, as this type of activity typically requires a certain level of
technical knowhow and access to the appropriate tools. However, we sug-
gest that this type of remixing is quickly gaining in popularity, due to the
rapid diffusion of free or low-cost design tools (e.g., computer-aided draw-
ing (CAD) software) and accessible manufacturing hardware (e.g., desktop
3D printers).
In addition to having access to the appropriate knowledge and tools,
brand remixing also requires that both managers and consumers change
 56 ARIC RINDFLEISCH AND MATTHEW O’HERN

their perceptions of brands from objects that are fixed and centrally con-
trolled to resources that are fluid and open to the masses. Thus, in the next
section of this paper, we contrast these two competing views of brands
(i.e., as communal vs. private assets). We then provide a brief review of the
remixing literature. This review informs our analysis of a recent example of
brand remixing, the Nokia case. This example is particularly interesting, as
Nokia openly posted the design for its Lumia 820 series smartphone case
for anyone to digitally remix (using 3D design software) and physically
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

produce (via 3D printing technology). We explore both the motives behind

this corporate-sponsored brand remixing campaign as well as users’
response to this novel initiative. Based on this review and examination,
we conclude with a set of implications for brand managers and branding
scholars.

BRANDS: PRIVATE VERSUS COMMUNAL ASSETS

As noted earlier, extant marketing thought and practice largely views
brands as valuable assets that must be tightly protected and closely mana-
ged. As documented by Madden, Fehle, and Fournier (2006), brands are
financially valuable assets that reduce cash flow volatility. Likewise, a
healthy body of work in the brand equity domain provides compelling evi-
dence that brands with higher levels of consumer equity are able to com-
mand a number of benefits, such as price premiums and customer loyalty
(Aaker & Keller, 1990; Keller, 1993, 2000; Park, Jaworski, & MacInnis,
1986; Yoo, Donthu, & Lee, 2000). According to this literature, building
and maintaining strong brand equity is highly dependent upon a firm’s
ability to systematically manage its brands and achieve consistency across
its various brand elements. For example, Yoo et al. (2000) recommend
that, “brand equity should be managed over time by maintaining brand
consistency, protecting the sources of brand equity … and fine tuning the
supporting marketing program” (p. 197). Likewise, Park et al. (1986) note
that, “the long-term success of a brand depends on marketers’ abilities to
select a brand meaning prior to market entry, operationalize the meaning
in the form of an image, and maintain that image over time” (p. 135).
Hence, according to this view, brand management is primarily a firm-
driven, top-down process in which brand managers create meaning and
communicate this meaning to current and prospective customers. Thus,
managers have a strong incentive to protect their valuable brand
 Brand Remixing: 3D Printing the Nokia Case 57

investments from external agents who may seek to dilute or alter their
brand’s essence.
In contrast to this traditional perspective of brands as private assets that
are developed and maintained by the firms that market them, a growing
body of research suggests that brands are actually communal assets that
are often shaped by those who purchase and use them. Indeed, the brand
community literature provides numerous examples of consumers collec-
tively engaged in various brand-altering activities (e.g., McAlexander,
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Schouten, & Koenig, 2002; Muniz & O’Guinn, 2001; Muniz & Schau,
2005). For example, Muniz and Schau (2005) document how members of
the Apple Newton community provide technical support and create new
applications for an electronic device that was discontinued and no longer
supported by the firm that launched it. Likewise, McAlexander et al. (2002)
show how some Jeep enthusiasts remove the doors of their vehicle to
enhance their driving experience. Although these activities are typically
initiated and conducted by a brand’s customers, they seem to have positive
spillover effects for its manufacturer (e.g., increased word of mouth,
enhanced loyalty, new ideas for product development).
Beyond the brand community literature, the communal nature of brands
is also clearly illustrated by related research on brand resistance, which
focuses on the variety of ways in which consumers alter brand meanings
through either consumption practices or anti-branding activities. For exam-
ple, Thompson et al. (2006) illustrate how anti-brand activists crafted a
“doppelganger” image of Starbucks as a rapacious corporate poseur and
how this alternative image competes against its strategically managed
brand image as a respite for corporate bohemians. Although this compet-
ing imagery was created by a loosely organized network of individuals
without the benefit of a marketing budget, it had a substantial, negative
impact on Starbuck’s brand equity and was cited by Howard Schultz as a
major threat to its brand franchise (Hofman, 2007). As noted by O’Hern
and Rindfleisch (2010), this type of anti-branding activity appears to be on
the rise and “has led several large firms, including Wal-Mart, Nike, and
McDonalds to be more cognizant of and open to consumer input” (p. 87).
In addition to these types of direct challenges, consumers may also indir-
ectly subvert strategic branding initiatives by using brands in idiosyncratic
and unexpected ways that may alter a brand’s meanings and intended use.
For example, Duck Tape, which was originally designed to seal holes in
WWII ammo boxes, has been transformed by creative users into a material
for crafting a variety of fashion accessories, including hand bags, prom
dresses, and footwear. As noted by Holt (2002) a growing number of
 58 ARIC RINDFLEISCH AND MATTHEW O’HERN

customers are seeking to “individuate market offerings and avoid market

influences” (p. 83). This type of individuating activity bears a close resem-
blance to von Hippel’s (2005) notion of innovative lead users, who modify
products to better suit their idiosyncratic needs.1 As documented by von
Hippel (2005), these modifications often provide manufacturers with
unique and valuable insights for future new product development efforts.
According to Poetz and Schreier (2012), these types of consumer-generated
new product concepts often score substantially higher in terms of novelty
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

than product concepts designed in-house by a firm’s employees.

In sum, despite their best efforts to carefully cultivate and protect their
valuable brand assets, firms are increasingly facing a world in which consu-
mers treat these assets as communal property that can be altered in a vari-
ety of ways. These alterations may subvert a brand’s intended imagery
and/or usage and thus challenge managers’ efforts to maintain consistency
and control over a brand. However, these unexpected challenges may
also lead to a variety of unanticipated benefits in the form of enhanced mar-
ket sensing and improved product functionality (Thompson et al., 2006;
von Hippel, 2005). Thus, a growing number of firms are actively encoura-
ging their customers to help them co-create certain elements of their brands
(O’Hern & Rindfleisch, 2010). We suggest that for these types of firms,
brand remixing may provide a new and valuable tool for cultivating com-
munal contributions as a way of enhancing their strategic branding efforts.

THE RISE OF REMIXING

At its core, remixing is the process of taking elements from existing offer-
ings and altering or recombining them to create something new (Lessig,
2008). Although remixing is most commonly associated with songs and
videos, this process can be applied to nearly any cultural offering. For
example, Andy Warhol employed remixing to create his acclaimed pop art
paintings (e.g., multicolored Campbell Soup cans and stylized Coca Cola
bottles). Indeed, Knobel and Lankshear (2008) note that, “Remix has not
simply emerged with digitization. It has always been a part of any society’s
cultural development” (p. 22).
At present, remixing activity is most prevalent in the entertainment
industry; remixed songs and videos permeate YouTube, CCMixter, and
other file-sharing websites. For digitized objects such as recorded songs
and videos, this process largely entails cutting, copying, and pasting slices
of these works to form a unique composition (Navas, 2010).2 The term
 Brand Remixing: 3D Printing the Nokia Case 59

“remix” originated in US hip-hop culture in the mid-1970s as DJs such as

Grandmaster Flash and DJ Kool Herc used multimixers to combine slices
of existing songs into a new sound (Manovich, 2005; Russo & Coppa,
2012). Since then, remixing has moved to the mainstream, as the multi-
mixer has been replaced by free and easy-to-use digital editing toolkits
(e.g., Apple Garage Band, Windows Movie Maker) that enable a song or
video to be copied, cut, and pasted by nearly anyone over the age of 10
with access to a computer (Lessig, 2008).
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Prior Research

Academic scholarship on remixing is thin and scattered across a wide range

of disciplines, including education, law, and media studies (Knobel &
Lankshear, 2008; Lessig, 2008; Manovich, 2005, 2007). The definitive work
in this domain, thus far, is Lessig’s (2008) book, Remix, which outlines the
foundations of remixing activity and its economic, social, and legal implica-
tions. As part of this treatise, Lessig (2008) draws a distinction between
read/write (RW) versus read/only (RO) cultures. Specifically, he suggests
that cultural offerings have historically been produced in RW cultures in
which a song or a story is remixed as part of its natural development.
However, during the 20th century, new technologies such as the radio and
television gave rise to cultural offerings “read” by many but “written” by
only a chosen few. As a result, “The twentieth century was the first time in
the history of human culture when popular culture had become professio-
nalized, and when the people were taught to defer to the professional”
(Lessig, 2008, p. 29). Our new century appears to be reversing this trend, as
the tools to design, edit, and distribute cultural offerings have become
democratized. Thus, in a larger sense, remixing represents a return to a
RW culture in which the production of cultural objects is placed in the
hands of the many rather than the few.
Although remixed cultural offerings are now fairly common, little is
known about remixing activity and the few studies that have been con-
ducted largely focus on music remixing. One of the earliest studies in this
domain is Poschardt’s (1998) historical overview of DJ culture, in which
remixing plays a key role. In this study, Poschardt (1998) examines the
motives underlying remixing activity and suggests that, “In most cases the
remix is prompted by love of the original. The remix seeks to give the origi-
nal new life, to reinforce its influence and preserve its idea” (p. 33). This
view of remixing as a form of tribute and adulation is also shared by
 60 ARIC RINDFLEISCH AND MATTHEW O’HERN

Cheliotis and Yew (2009), who conduct an empirical assessment of remix-

ing activity among members of the ccMixter community.3 They find that
approximately 40% of the original songs (vs. only 10% of remixes) posted
on ccMixter ultimately get remixed. Cheliotis and Yew (2009) also suggest
that remixing provides users with an outlet for personal expression and
enhanced social relations. This portrait of remixing (i.e., adulation of an
original offering, forum for personal expression, enhanced social relations)
bears a close resemblance to the motives underlying consumer participation
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

in brand communities (Muniz & O’Guinn, 2001). Thus, there appears to be

good reason for creators to encourage remixing activity among their users.

The Impact of Remixing

As remixing is democratized and becomes more accessible to the average

user, musicians and creators of other cultural offerings appear to be increas-
ingly aware of its potential benefits. Knobel and Lankshear (2008) note that
some musicians such as Jay Z actually “encourage remixing and make
mixable versions of their work available for downloading and tinkering”
(p. 24). Moreover, remixing appears to be influencing the manner in which
an original offering is composed, as “musicians now often conceive their
works beforehand as something that will be remixed, sampled, taken apart,
and modified” (Manovich, 2007, p. 4). This use of remixing as a strategy has
also been adopted by video game manufacturers (e.g., Microsoft’s Halo 3
and Sony’s Little Big Planet), who provide toolkits that allow users to mod-
ify their games and share these modifications with fellow enthusiasts
(Merges, 2007). As noted by Knobel and Lankshear (2008), “each new
remix becomes a meaning-making resource for subsequent remixes” (p. 26).
By employing this principle, Sony was able to turn a game (i.e., Little Big
Planet) that left the factory with five built-in levels into a game that has over
6 million levels, nearly all of which were created through customer-led
remixing (Softpedia, 2012). Thus, at least in some industries, remixing
appears to be increasingly viewed as a beneficial activity for both creators
and users.
This positive view of remixing appears at odds with the extant branding
literature, which suggests that maintaining control over a brand and
crafting consistent branding campaigns is paramount (Aaker & Keller,
1990; Keller, 1993, 2000). Remixing flies in the face of this well-established
branding paradigm. Because it is difficult to predict how consumers will
alter a brand, remixing seems antithetical to good branding practice and is
 Brand Remixing: 3D Printing the Nokia Case 61

often regarded as an activity that should be actively discouraged, and per-

haps, legally prosecuted. This may be one reason why, outside of cultural
offerings such as music, videos, and games, examples of brand remixing are
few and far between. As noted by Senevirante and Monroy-Hernandez
(2010) remixing is most likely to occur when “people build on the work of
those creators who want their work to be reused.” The next section of our
paper examines a creator (i.e., Nokia) that closely fits this description and
who actively cultivated remixing as part of its branding strategy.
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

THE NOKIA CASE

Nokia, which had long been the world’s leading provider of mobile phones,
was a late mover to the touchscreen smartphone market and has seen its
market share drop precipitously as a result (Spence, 2013). As a response, it
recently teamed up with Microsoft to launch a new line of smartphones
(and tablets) under the “Lumia” sub-brand.4 The first Lumia smartphone
was launched in November 2011 and a series of additional versions have
been introduced since then. Our examination focuses on one of these mod-
els, the Lumia 820 smartphone, which was launched in September 2012 (see
Fig. 1). This phone employs the Windows 8 operating systems, has a high-
speed processor, a long-lasting battery, a high-resolution camera, and cur-
rently retails for approximately $250. In contrast to competing offerings,
such as the Apple iPhone, the Lumia 820 is designed with a removable case
in which the back of the phone can be easily opened for battery replacement
by users. In addition, the stock Lumia 820 case can be swapped with a new
case with a different color, design, or function (i.e., a shell with a wireless
charging option). We focus on this case because Nokia recently launched an
interesting new initiative that allowed users to remix a template of the
Lumia 820 case using desktop 3D printing technology. Before discussing the
specifics of this initiative, we provide a brief overview of 3D printing, as this
technology is still quite novel and may not be familiar to most readers.

3D Printing Overview

In brief, a 3D printer is a technological device that allows users to create an

object by layering thin (300 microns or less) slices of a material in an addi-
tive manner. Although commercial 3D printers have existed for about three
 62 ARIC RINDFLEISCH AND MATTHEW O’HERN
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Fig. 1. Nokia Lumia 820 Smartphone. Source: http://www.microsoft.com/en/

mobile/phone/lumia820/specifications/

decades, mass-market desktop 3D printers have only recently become avail-

able and, at present, are still in an early stage of diffusion (Lipson &
Kurman, 2013).5 The current leader in the desktop 3D printing market is
the MakerBot Replicator (see Fig. 2). This printer is about the size of a
microwave oven, has a build capacity of about 400 cubic inches and prints
 Brand Remixing: 3D Printing the Nokia Case 63
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Fig. 2. MakerBot Replicator 3D Printer. Source: Personal photo.

objects using a thermoplastic called PLA (polylactic acid). The Replicator

2 was introduced in September 2012 and currently retails for approximately
$2,000.6 Users can print objects that are designed using free or low-cost 3D
modeling software (e.g., Google Sketchup), captured via a 3D scanner
(e.g., MakerBot Digitizer), or downloaded from an online 3D file-sharing
platform (e.g., Thingiverse.com). Although desktop 3D printers have a lim-
ited range of printing materials (mainly thermoplastics and resins), users of
this technology have been able to create a wide variety of objects including
the whimsical (e.g., Stephen Colbert’s head), the practical (e.g., replacement
parts for old appliances), and combinations of the two (e.g., t-rex shower-
head) (see Fig. 3). Smartphone cases are one of the earliest and most popu-
lar categories of 3D printed objects. For example, Thingiverse.com boasts
nearly 4,000 design files labeled “iPhone case,” all of which were designed
by iPhone users, rather than Apple itself.

Customizing the Lumia 820

In January 2013, Nokia announced the launch of a “3DK kit” which con-
tained “everything someone versed in 3D printing needs to print their own
custom Lumia 820 case” (Willans, 2013). This kit included a 3D design
template, a set of design ideas, hardware and software specifications, and a
list of recommended materials for printing a customer-designed case.
 64 ARIC RINDFLEISCH AND MATTHEW O’HERN
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Fig. 3. Examples of Objects Made Using a 3D Printer. Sources: (a) http://www.

thingiverse.com/thing:9104; (b) http://www.thingiverse.com/thing:9581; (c) http://
www.thingiverse.com/thing:329596
 Brand Remixing: 3D Printing the Nokia Case 65
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Fig. 4. Customizable Nokia Lumia 820 Case App. Source: http://www.thingiverse.

com/thing:52507

Nokia’s basic concept was that their Lumia 820 customers who had access
to a 3D printer could download this design template, modify it as desired,
and then create their own customized case. Nokia followed this initial
release by launching a “Customizable Nokia Lumia 820 Case” on
Thingiverse on February 25, 2013. This customizable design took advan-
tage of Thingiverse’s newly launched Customizer app, which allows users
with no programming skills to remix various elements of the standard
design template using a series of pull-down menu options. Nokia’s Lumia
820 standard design template is displayed in Fig. 4. The Customizer app
allows users to easily modify various elements of this design, including the
shape of its design pattern, the message printed at the bottom of the case,
as well as its font. These elements represent concrete product features
(Park, Milberg, & Lawson, 1991), and thus, are the focus of our analysis.

Analysis of Remixing Activity by Users

As of December 31, 2013, Thingiverse users had uploaded 92 remixes of

Nokia’s Customizable Lumia 820 Cases (approximately one every 3.4
days).7 These 92 remixes were submitted by 63 users. Although most of
 66 ARIC RINDFLEISCH AND MATTHEW O’HERN

these users submitted only a single remix, there were 11 (17%) who sub-
mitted multiple remixes.8 We examined each of these remixes and also
recorded the username of the individual responsible for each remix as well
as the number of views and downloads that each remix received. Because
our primary emphasis is physical brand remixing, we focus on the pattern,
font type, and message displayed on each design. As shown in Fig. 4, the
standard design template offered by Nokia employed a hexagonal design
pattern and displayed the message “NOKIA” at the bottom in all caps
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

using the brand’s recognizable futuristic font. Given the paucity of research
on brand remixing, we do not pose any propositions regarding the degree
or nature of the remixing activity. Thus, our analysis is largely descriptive
and exploratory in nature. Specifically, we examine the alterations that
users made in terms of remixing the case’s text, font type, and design pat-
tern, as well as the amount of diffusion (i.e., views and downloads) attained
by these remixes.
Out of the 92 designs submitted, only 22 remixes (24%) employed the
message “NOKIA.” However, several others featured messages that
acknowledged-related aspects of the brand. For example, while the app’s
template specified the corporate brand, five remixes employed the model’s
brand name (i.e., “LUMIA”), one displayed both the corporate and model
brand (i.e., “Nokia Lumia”), and three other remixes acknowledged the
phone’s relation to its brand partner, Microsoft (e.g., “MICROSOFT,”
“<3 WindowsPhone”). Also, three more remixes featured personalized ver-
sions of the brand name (i.e., “Raman’s NOKIA,” “BLAGO’S LUMIA
820”) or messages of brand adoration (i.e., “Luv My Nokia”). In total, 34
remixes (37%) employed some version of the Nokia (or its partner
Microsoft’s) brand name.
Among the two thirds of remixes that employed a non-branded alterna-
tive message, 25 (27%) had the personal name of an individual
(e.g., “Demitri!,” “JOHN,” “MARTA”), 17 (18%) displayed a username
or some type of nickname (e.g., “Datasmasher,” “KING ATHI,”
“SLAPDRAGON”), and 13 (14%) remixes featured no message whatso-
ever. The remaining three (4%) remixes were difficult to classify, as they
represent a hodgepodge of statements (i.e., “#paypal it”), imperatives
(i.e., “CALL ME!”), and perhaps one attempt at remixing the Nokia brand
to combine it with a user’s name (i.e., “NIKKIA”).
In sum, when provided with the opportunity to modify the message on
the Lumia 820 case, about half of the users who created their own remixes
and uploaded their designs on Thingiverse chose to replace “NOKIA” with
 Brand Remixing: 3D Printing the Nokia Case 67

a more personally relevant message (such as their name), while slightly

more than one third chose to retain some form of the Nokia brand mes-
sage. While one out of three may seem like a modest degree of brand
adherence, it should be noted that the users who crafted these remixes
could have entered any message they desired. Thus, the fact that nearly
40% elected to design a case that portrayed a pro-Nokia branded message
actually seems quite remarkable.
As noted earlier, the Nokia Lumia 820 Customizer App allowed users not
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

only to alter the message that appeared on the case, but also to make
changes to both the font and the geometric pattern that appeared above this
message. Specifically, users had three font options (i.e., basic, futuristic, and
fancy) and 12 pattern options (three-sided to twelve-sided pattern, thirty-
sided pattern, or no pattern). Of the 79 remixes that had a message, 54
(68%) displayed this message using the futuristic font that Nokia uses in its
brand messaging. This percentage is significantly higher (p < .05) than what
would be expected by chance (i.e., 33%). Likewise, 20 (22%) of remixes
employed a hexagonal pattern. This percentage is significantly higher (p <
.05) than what would be expected by chance (i.e., 8%). Of the remaining
72 remixes, 56 (61%) employed an alternative pattern, while 16 (17%) did
not display any pattern at all. In aggregate, 42% of the 92 remixes had at
least one design element (i.e., message, font, or pattern) that adhered to
Nokia’s design template. Again, considering that users were free to custo-
mize each of these elements to their liking, this seems like a notable degree of
congruence to Nokia’s standard design. Hence, these results indicate that a
substantial amount of brand remixing activity may closely adhere to the
suggested brand elements found in the remixing tools that a firm provides
to its users.
Finally, we assessed the number of views and number of downloads for
both Nokia’s base design template as well for as each of its 92 remixes. By
the end of 2013, Nokia’s Lumia 820 template received 13,003 views and
was downloaded 832 times. In contrast, on average, each of the 92 remixes
garnered 222 views and 54 downloads. While these individual averages pale
in comparison to those of the base design, the aggregate total of views
(20,429) and downloads (4,949) across all 92 remixes substantially exceed
what Nokia achieved with its base design template. Even if these totals are
discounted to the 42% of the designs that contained at least one design ele-
ment (i.e., message, font, or pattern) that adhered to the default settings,
the adjusted total number of views (8,580) and downloads (2,079) are still
quite impressive.
 68 ARIC RINDFLEISCH AND MATTHEW O’HERN

Impact of Nokia’s Remixing Initiative

Although the ultimate impact of this remixing initiative remains to be seen,

several factors suggest that this initiative was successful for Nokia. First,
the immediate response to this initiative from the business press appears to
have been uniformly positive (Lomas, 2013; Meyer, 2013; Spence, 2013).
For example, according to Meyer (2013), “Nokia has made one of its
smartest moves in ages … it’s invited its users to effectively tailor an ele-
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

ment of its smartphone hardware to their individual needs.” Likewise,

Lomas (2013) added, “It’s certainly a bold move for a big corporate com-
pany to allow users to remix its design without any checks and balances on
what they produce.”
In addition to this positive reaction from the business press, this remix-
ing initiative also provided Nokia with apparent benefits in terms of custo-
mer response. As noted earlier, by allowing modification of the case’s
textual message, Nokia’s customers were able to attain a high degree
of personalization and/or unique brand expression. Prior research suggests
that these types of personal modifications strengthens customers’ degree
of brand attachment (Franke, Schreier, & Kaiser, 2010; Moreau &
Herd, 2010).
Moreover, this remix initiative also allowed Nokia to multiply the diffu-
sion of its downloadable case by a factor of nearly six (i.e., 833 downloads
of Nokia’s base template vs. 4,949 downloads of the 92 remixes). In addi-
tion to this increased diffusion, by allowing users to remix their cases,
Nokia succeeded in obtaining 92 free prototypes, some of which could
prove helpful in future product development efforts (von Hippel, 2005).
Nokia was also able to track which designs were downloaded most fre-
quently, thereby gaining access to potentially valuable consumer feedback
(Ogawa & Piller, 2006). Indeed, these consumer response and prototyping
benefits appear to be key motives behind Nokia’s decision to launch this
customizable app (Willans, 2013).
To complement the Customizable Lumia 820 Case that appeared on
Thingiverse, Nokia posted a separate, lead-user oriented remixing contest
on its website called the 3D Printing Challenge. In this contest, developers
from around the world were asked to alter the basic design for the Nokia
820 case to create new product prototypes that were highly creative, had
the potential to lead to new commercial products, and would be widely
embraced by the Nokia user community (Nokia, 2013). A five-member jury
of Nokia and MakerBot employees evaluated 109 design submissions and
ultimately selected five winners. Winning designers received an all-expense
 Brand Remixing: 3D Printing the Nokia Case 69

paid trip to the 2013 Mobile World Congress in Barcelona, Spain, and an
opportunity to meet with the jury to discuss their designs.
An examination of the winning submissions reveals that these designs
focused heavily on creating tangible new product features rather than
altering aesthetic features. For example, a 3D printed prototype by Victor
Johansson cleverly added a slot to the basic design that allowed users to
hide a key within the case (see Fig. 5). Another physical prototype, created
by designer Aditya Damar, went one step further by adding a larger
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

compartment that allowed the case to be used as a wallet to store cash,

credit cards, etc. In both cases, the winning designers used 3D printing
technology to create working product prototypes that were quite innova-
tive and judged to be highly appealing to Nokia users. Moreover, these
products incorporated tangible new features not typically seen in mobile
phone cases. Hence, Nokia’s 3D Printing Challenge succeeded in providing
Nokia designers with several innovative product prototypes, while also
enhancing the value of the Nokia brand by reinforcing its association
with a highly innovative and disruptive technology (i.e., 3D printing).
This association is clearly evident in the response of Victor Johannson
(one of the winners of Nokia’s 3D Printing Challenge), whom we inter-
viewed as a means of collecting additional insights into this aspect of
Nokia’s remixing initiative:

I already had a positive attitude towards Nokia as a company before this competition
but it (the contest) certainly reinforced that view. Nokia seems more like a collective of
people working towards a goal than a company, and this was very apparent at MWC
when comparing the Nokia and Samsung booth, for example. (Johansson, 2014)

Fig. 5. Nokia 3D Printing Challenge Winner: Key Holder. Source: http://www.

coroflot.com/vijo/Nokia-3D-printing-challenge. r Victor Johansson, objectsand
interactions.com
 70 ARIC RINDFLEISCH AND MATTHEW O’HERN

Johansson’s reflections suggest that Nokia succeeded in using its remixing

campaign to put a human face on its brand and to position itself as an
authentic and forward-thinking company.
In sum, for a relatively meager investment, Nokia’s multi-faceted remix-
ing campaign succeeded in attracting renewed attention to its brand,
enhancing customer connection with the company, and strategically lever-
aging a cutting edge technology. In the next section of our paper, we build
on the insights from this case study and explore the broader implications
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

that the remixing of physical products has for both brand managers and
branding scholars.

IMPLICATIONS FOR BRAND MANAGEMENT

THOUGHT AND PRACTICE

At present, the Nokia case, and brand remixing more broadly, is clearly the
exception rather than the norm. Most consumers use products that come
ready made and few have the capability to change a brand’s physical char-
acteristics. Nonetheless, as recently revealed in a large-scale, national sur-
vey of consumer innovation in industrial countries such as Japan, the
United Kingdom, and the United States, millions of individuals tinker with
mass marketed products to better adapt them to their idiosyncratic needs
and preferences (von Hippel, Ogawa, & de Jong, 2011).
Retrofitting existing products post-purchase is a challenging and uncer-
tain task. As a result, the percentage of tinkerers is currently rather modest
(about 5% of the population) (von Hippel et al., 2011). However, the
advent of 3D printing, scanning, and design has the potential to bring the
ability to easily (and reliably) alter the structure of physical products to a
much broader segment of the population (Anderson, 2012). As noted by
Ratto and Ree (2012) 3D printing allows for “the materialization of digital
information.” The digitizing of material objects is a groundbreaking tech-
nological development and provides the “opportunity for consumers to
compete with producers” (Lessig, 2008, p. 37). Aided by this new technol-
ogy, consumers who would like to tinker with a brand’s physical character-
istics can use existing (and increasingly affordable) 3D printing technology
to remix a digital template of the product and actually produce the remixed
product in their own homes. This is exactly what Nokia encouraged its
users to do with its Lumia 820 case and we expect that more brands will
soon follow. This advent of brand remixing presents a number of implica-
tions for both branding scholarship and practice.
 Brand Remixing: 3D Printing the Nokia Case 71

Implications for Brand Scholars

We believe that the rise of brand remixing will dramatically alter how both
managers and consumers think about brands. At present, brands are lar-
gely viewed as intangible entities that are tightly controlled by marketing
managers (Aaker, 2009; Keller, 2012). Brand remixing challenges this per-
spective by ceding a portion of this control to a brand’s user base and
allowing them to tinker with the basic DNA of the brand. As shown in the
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Nokia case, when afforded this capability, users create an interesting set of
“mutations,” some of which reveal authentic expressions of brand love
(e.g., “LUV MY NOKIA,” “<3 WindowsPhone”) that might seem disin-
genuous and inauthentic if suggested by a marketing manager.
Nokia’s Lumia 820 remixing campaign was revolutionary, as it allowed
users to actually alter key elements of the Nokia brand. For example, users
were free to retain the Nokia brand name on their case while changing the
official Nokia font. Alternatively, users had the power to remove the Nokia
brand name completely, to replace it with a personalized message, or to
combine the Nokia brand name with their own message. Although mass
customization tools (e.g., NikeID) allow users to modify certain aspects of
a product (e.g., colors, textures, patterns), brand remixing is more powerful
than mass customization, because it gives consumers the power to alter key
design elements that constitute and lend meaning to a brand.
To illustrate this difference, imagine a remixing tool that allowed
NikeID users to remove the Nike swoosh from their shoes or an Apple-
sponsored tool that enabled users to create their own “MacBook Bro” label
for their laptops. Given that much of Apple and Nike’s success can be
traced to each company’s ability to create design elements that are highly
consistent with their brands, perhaps it is not surprising that neither firm
allows customers to engage in the type of brand remixing that Nokia
openly encouraged. Hence, in terms of ceding control over the brand to
customers, Nokia’s Lumia 820 remixing initiative can be seen as a bold
departure from traditional branding practice. An interesting question for
future research is the degree to which brand managers can release control
to remixers and still achieve their brand’s organizational objectives.
Firms that adopt a pro-remixing stance may also be able to garner bene-
fits that more traditional minded brands find difficult to achieve. By pro-
viding users with a vehicle for authentic self-expression, brands that
embrace remixing should be less likely to be targets of doppelganger brand
imagery or other forms of anti-brand activism, as these attacks largely
focus on brands perceived as lacking authenticity and controlled by the
 72 ARIC RINDFLEISCH AND MATTHEW O’HERN

corporate elite (Holt, 2002; Kozinets & Handelman, 2004; Thompson

et al., 2006). As noted by Holt (2002), “brands will be more valuable if they
are offered not as cultural blueprints but as cultural resources, as useful
ingredients to produce the self as one chooses” (p. 83). Indeed, the remixing
literature strongly suggests that remixed cultural products are not simply
plagiarized forms of an original blueprint but authentic and novel creations
(Manovich, 2005, 2007; Poschardt, 1998).
Relatedly, brand remixing also challenges established views regarding
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

consumer-brand relations (Fournier, 1998). Traditionally, this literature

regards brands as relational partners that help consumers establish or
change their identity. For example, a Harley Davidson motorcycle may
confer an image of rebellion and masculinity upon its owner (Schouten &
McAlexander, 1995). As famously quipped by former Harley Davison
CEO, Richard Teerlink, “What we sell is the ability for a 43-year-old
accountant to dress in black leather, ride through small towns and have
people be afraid of him” (Dourado & Blackburn, 2005). In essence, the
extant brand relations literature has focused on how brands help form the
identity of their users. Brand remixing inverts this perspective by focusing
on how users help form the identity of their brands. As shown from the
Lumia 820 case, brand remixing appears to be a mechanism for users to
personalize brands with a sense of their own identity. Hence, the impact of
brand remixing on traditional self-brand relations is an interesting avenue
for future research.
Brand remixing may also hold implications for research on brand com-
munities (McAlexander et al., 2002; Muniz & O’Guinn, 2001; Muniz &
Schau, 2005). The brand community research is replete with examples of
consumers engaged in a variety of brand-related practices such as customiz-
ing and evangelizing (Schau, Muniz, & Arnould, 2009). These practices are
believed to provide many positive benefits to a brand, including strength-
ened consumer engagement and higher levels of brand usage (Schau et al.,
2009). Remixing activity can be viewed as a communal activity, as remixed
versions usually credit the original source and are often freely shared with
others in the user community (Cheliotis & Yew, 2009; Poschardt,
1998). However, in contrast to most known forms of communal brand
practice, brand remixing appears to be considerably more fragmented
in nature and potentially lacking in terms of the collective consciousness
typically associated with most brand communities (Muniz & O’Guinn,
2001). Thus, some interesting questions for future research include the
ways in which production-based community activities, such as brand
remixing, differ from more prevalent consumption-based activities, whether
 Brand Remixing: 3D Printing the Nokia Case 73

production-based or consumption-based activity is more effective in

strengthening brand community attachment, and whether the same set of
mechanisms underlie these activities.
Our research focused on one type of remixing activity (i.e., modifications
of an original creation). Although this is the prototypical form of remixing,
users can also engage in mixing (i.e., splicing together two or more original
creations) and sampling (pulling thin slices of several original creations into
a new composition) (Poschardt, 1998). Thus, future research could enrich
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

our work by assessing the degree to which these other forms of remixing
practice occur in various brand communities and assessing their differential
impact on consumer response.
In addition to its implications for how we think about brands, remixing
also challenges our view of brand users. Although recent research has
acknowledged that users may exercise some degree of control over brand
meanings and consumption practices, the branding literature largely
focuses on users as consumers (Aaker, 2009; Keller, 2012). However, brand
remixing provides users with considerable control over both the tangible
and intangible elements that comprise a brand offering. Thus, users who
engage in remixing activity are doing much more than simply consuming a
brand; in a very real sense, remixers are involved in co-creating a brand.
Poschardt (1998) provides an eloquent analogy from the domain of music
remixing and suggests that a DJ “is at once a composer and a listener”
(p. 378). In addition to identifying this dual role, Poschardt (1998) also sug-
gests that playing both roles profoundly affects the manner in which music
is consumed: “As soon as the music fan becomes a DJ, his way of listening
to records changes. Perception becomes more precise” (p. 379). Future
research could provide a valuable contribution by identifying the extent to
which this dual role perception occurs among brand remixers as well as the
degree to which these dual roles (i.e., producer vs. consumer) affect how a
brand is evaluated and consumed.

Implications for Brand Managers

Brand managers face an array of intriguing questions when considering

the implications of brand remixing. Perhaps the first question is what
types of products (and brands) should be offered for physical remixing?
Unfortunately, current examples are few and far between. In addition to its
Lumia 820 case, Nokia has also made the case for its Lumia 520 available
for remixing (via Thingiverse). In January 2014, Samsung followed Nokia’s
 74 ARIC RINDFLEISCH AND MATTHEW O’HERN

lead by teaming up with 3D Systems to allow its customers to remix cases

for its smartphones (Volpe, 2014). In addition to smartphone cases, sculp-
tures represent another prominent category for remixing. For example,
both the New York Metropolitan Museum of Art and the Chicago Art
Institute have allowed 3D scans of some of their most prominent sculp-
tures, such as the Met’s Reclining Nyad and the lion sculptures that stand
guard in front of the Art Institute, to be uploaded on Thingiverse with the
understanding that these scans would likely be modified and reinterpreted
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

by Thingiverse users (Andrew, 2012). As an example of how these sculp-

tures have been remixed, one user sliced the head off one of the Art
Institute lions and remixed it to function as the top of a Pez dispenser!
As 3D printing technology moves up the adoption curve, we expect that
a variety of other products (and brands) will be interested in joining the
fray. As noted by Holmes (2012), “as 3D printing and scanning technolo-
gies become more ubiquitous and accessible, this mashup mindset will emi-
grate from the digital realm of music files and online videos into the land of
physical objects.” However, some brands may be better suited for remixing
than others. According to Park et al. (1991), brand consistency appears to
be more important for prestige brands than for functional ones. Thus,
brand remixing may be less risky for mass-market brands such as Nokia
and Samsung than for luxury brands such as Prada and Rolex.
Nonetheless, as 3D scanners and printers become more accessible and
obtain greater capability, users will soon be able to scan and remix ele-
ments of nearly any brand. Hence, even if a firm does not directly partici-
pate in remixing, its offerings may be remixed without its consent. For
example, a scan of the head of Stephen Colbert was uploaded on
Thingiverse on June 8, 2011. Since this time, this scan has been remixed 30
times into a variety of manifestations, including being placed on the body
of a t-rex and being hollowed out to serve as a coffee mug. Thus, brand
managers should be prepared for the possibility that the physical aspects of
their brand may soon be remixed and will need to decide if they will
embrace this act of user creativity, ignore it, or seek to squelch it (Berthon
et al., 2007).
Although some firms may employ the threat of litigation to prevent con-
sumers from creating and distributing remixes of their brands, such heavy-
handed tactics are likely to result in reducing a company’s brand equity
(via negative word of mouth) rather than effectively protecting its intellec-
tual property (Berthon et al., 2007). Hence, as brands move from the
domain of private assets to that of communal assets, decisions relating to
sustaining and enhancing brand equity are likely to become increasingly
 Brand Remixing: 3D Printing the Nokia Case 75

complex. For example, firms that position themselves as being highly

customer-oriented and/or highly innovative may find it particularly difficult
to prohibit unwanted remixing activity for fear of alienating their consumer
base and jeopardizing their brand equity.
In fact, our case study suggests that managers may have other options
for ensuring that brand remixing is consistent with their other brand-
building initiatives. Although branding scholars acknowledge that man-
agers are virtually powerless in terms of dictating how customers react to
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

their brands on social media (Fournier & Avery, 2011), evidence from the
Nokia case suggests that firms may be better able to guide and manage the
content that their customers create within the confines of a brand remixing
campaign. Consider, for example, that a relatively high percentage of cus-
tomers in our sample (i.e., 37%) created messages that explicitly referenced
the Nokia brand or Microsoft, its partner brand. Similarly, 68% of cases
that contained a message used Nokia’s official futuristic font to display this
message. These results strongly suggest that certain elements of the brand
remixing tool itself (e.g., the tool’s default settings) can influence the type
of content that users create and the degree to which these creations are con-
gruent with existing brand elements. Thus, brand remixing has the poten-
tial to offer consumers substantial room for self-expression while at the
same time being able to influence both the kind of content that is produced
and the way that other users view this content.
Successfully achieving these dual goals likely starts by pinpointing a tar-
get market of highly involved users that are capable and motivated to
engage in brand remixing. Nokia accomplished this goal by placing its
design template on the 3D design repository, Thingiverse.com. This step is
crucial, as brand remixing campaigns will depend on these customers to
create a critical mass of compelling content that can be shared with other
users to build online buzz. Selecting an appropriate target market helps
ensure that the resulting remixes are less likely to take on a doppelganger
character and reduces the risk of users condemning or disparaging the focal
brand.
In addition, managers have the ability to define the parameters that gov-
ern the kind of content created via a brand remixing initiative. For exam-
ple, if a remixing campaign takes the form of a firm-sponsored contest, the
rules and guidelines surrounding that contest could be instrumental in
ensuring that the majority of the content created reflects positively on the
firm and that most users create content that corresponds to an overarching
campaign theme endorsed by the firm. Again, providing a coherent struc-
ture for remixers will likely result in more content consistent with the focal
 76 ARIC RINDFLEISCH AND MATTHEW O’HERN

brand. In addition, managers can draw attention to user-generated remixes

that are especially creative yet at the same time successfully build on the
image of the focal brand. Highlighting and promoting these kinds of
remixes should encourage other users to emulate the work done by their
peers and provide an additional mechanism for managers to exert some
influence over the way in which a remixing campaign unfolds.
In terms of overall branding strategy, another important consideration
for brand managers relates to the optimal timing for encouraging (or per-
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

haps not discouraging) brand remixing. Park et al. (1986) suggest that
brand managers think of their brand concepts as evolving through three
key stages: introduction, elaboration, and fortification. Most firms would
likely seek to maintain tight control over a brand at both the introduction
and fortification stages (Park et al., 1986). Thus, remixing may be most
effective during the elaboration stage, as the richness offered by remixed
versions of brand may enhance brand concept elaboration and may help
differentiate a brand from competing offerings. According to Park et al.
(1986), “As markets and needs change, elaborating the brand concept
becomes important” (p. 144). However, they also note that when respond-
ing to these changes, “a firm must consider its resource capabilities” (p.
144). Remixing may help address these concerns by providing an effective
means for achieving elaboration by leveraging the capabilities of a brand’s
user base.
Assuming that brand managers are interested in incorporating remixing
into their branding strategy, what is the best means of encouraging users to
devote their time and energy to remix a particular brand? Prior research on
the remixing of songs and videos provides some helpful hints. First of all,
remixing is enhanced by modularity (Holmes, 2012; Manovich, 2005).
Breaking an object into divisible modules makes the task of remixing easier
for users, while also providing a firm with greater control over the part of a
brand that gets remixed. For example, Nokia divided the template for its
Lumia 820 case into two essential modules focusing on text and design pat-
terns. 3D printing technology greatly facilitates this task by digitizing phy-
sical products into smaller discrete units that can be easily modularized
(Ratto & Ree, 2012). Once a product is modularized and a firm has decided
which modules it would like to have remixed, the next step is to place the
modularized product on a hosting platform that consumers can use to
access the template. Although a firm could post this template on its own
webpage, a more productive approach may be to post it on a third-party
platform specifically designed for this purpose. At present, the best example
of this type of platform is Thingiverse.com. Designs posted on this website
 Brand Remixing: 3D Printing the Nokia Case 77

are governed by a Creative Commons license which provides the creator

with some control over the initial creation while allowing “other people to
use part of their work and make it new” (Manovich, 2005). It is likely that
new hosting sites for brand remixing will soon arise, with some of these
sites devoted to specific product categories and/or types of users.
Perhaps the final and most important question facing firms that embrace
remixing (or find their brands being remixed without their consent) is what
is their role in this new world order? As noted earlier, users who remix a
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

brand’s offerings take on some of the roles traditionally conducted by the

brand’s creator. Thus, remixing has important implications for the roles
and responsibilities of brand managers. Holmes (2012) suggests that firms
should adapt to this change in roles by focusing more on developing gen-
eral platforms rather than on creating specific objects. This is essentially
what Nokia has done with its Lumia 820 case by giving users a template to
alter, rather than a final design to adopt. Indeed, Nokia envisions a future
in which these templates may be offered as part of their product line:
“Perhaps in addition to our own beautifully-designed phones, we could sell
some kind of phone template … You want a waterproof, glow-in-the-dark
phone with a bottle opener and a solar charger? … you can print it your-
self!” (Willans, 2013). In addition to providing users with the benefit of
being able to customize an offering to better suit their needs and prefer-
ence, this template approach should also reduce a firm’s marketing research
and product development costs (von Hippel, 2005).

CONCLUSION

Brands have traditionally been created by firms and marketed to consu-

mers. This division, which has served as the foundation for much of the
extant branding literature, is partly a technological artifact, as the means of
brand creation have largely been held by firms rather than consumers. The
development and diffusion of such technologies as the personal computer,
digital design software, and the Internet has allowed consumers to remix a
brand’s message and have forced iconic brands such as Starbucks to alter
their marketing practices (Thompson et al., 2006). We believe that the
development and diffusion of low-cost 3D printing, scanning, and design
technology will usher in a similar shift in terms of remixing a brand’s key
design elements including its physical properties. The Nokia case provides
early insights into both the managerial motives for remixing as a product
 78 ARIC RINDFLEISCH AND MATTHEW O’HERN

strategy as well as the reactions to this approach by a brand’s user base.

The insights gleaned from this case suggest that, despite having to release
some control of their brand, remixing may provide a firm with positive ben-
efits in terms of increased product diffusion, greater user empowerment,
enhanced new product development, and more personalized consumer-
brand connections.
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

NOTES

1. According to von Hippel (2005), lead users modify products in an attempt to

enhance their functionality and are unlikely to share their modifications with others.
In contrast, brand remixers, like those shown in the example of the Nokia Lumia
820 case, often modify brands in ways that add little functionality and seek to
broadcast their modifications to others.
2. For an example of this process, see how GirlTalk remixes music in the follow-
ing video (http://www.youtube.com/watch?v=VlPkIS-uNMk).
3. ccMixter is a site dedicated to music remixing (ccmixter.com).
4. Nokia’s devices division was subsequently acquired by Microsoft in September
2013.
5. It is estimated that approximately 100,000 desktop 3D printers have been sold
over the past 3 years (Dougherty, 2013).
6. In January 2014, MakerBot announced three new desktop 3D printers, ran-
ging in price from $1,400 to $6,500.
7. These 92 remixes were the ones uploaded and available on Thingiverse on
December 31, 2013. There may have been more remixes that were created but not
uploaded by their makers.
8. The largest number of remixes submitted by any user was 12 (by 3dxcel).

REFERENCES

Aaker, D. (2009). Managing brand equity. New York, NY: Free Press.
Aaker, D., & Keller, K. L. (1990). Consumer evaluations of brand extensions. Journal of
Marketing, 54, 27 41.
Anderson, C. (2012). Makers: The next industrial revolution. New York, NY: Crown Business.
Andrew. (2012). Met MakerBot Hackathon Art Now On Thingiverse. MakerBot Blog.
Retrieved from http://www.makerbot.com/blog/2012/06/01/met-makerbot-hackathon-
art-now-on-thingiverse/
Berthon, P. R., Pitt, L. F., McCarthy, I., & Kates, S. M. (2007). When customers get clever:
Managerial approaches to dealing with creative customers. Business Horizons, 50,
39 47.
Cheliotis, G., & Yew, J. (2009). An analysis of the social structure of remix culture. New York,
NY: ACM.
 Brand Remixing: 3D Printing the Nokia Case 79

Dourado, P., & Blackburn, P. (2005). Tattoo you: Super loyalty. Stakeholder, 2(3)(July),
26 29.
Fournier, S. (1998). Consumers and their brands: Developing relationship theory in consumer
research. Journal of Consumer Research, 24, 343 374.
Fournier, S., & Avery, J. (2011). The uninvited brand. Business Horizons, 54, 193 207.
Hofman, M. (2007). Memo to Howard Schultz … Retrieved from http://www.inc.com/staff-
blog/2007/03/07/memo_to_howard_schultz.html
Holmes, K. (2012). Remix culture comes to physical objects with the remix project. Retrieved
from http://thecreatorsproject.vice.com/blog/remix-culture-comes-to-physical-objects-
with-the-iremixi-project
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

Holt, D. B. (2002). Why do brands cause trouble? A dialectical theory of consumer culture
and branding. Journal of Consumer Research, 29, 70 90.
Interbrand. (2013). Best global brands 2013. Retrieved from http://www.interbrand.com/en/
best-global-brands/2013/Best-Global-Brands-2013.aspx
Johansson, V. (2014). Email correspondence on 19th January 2014.
Keller, K. L. (1993). Conceptualizing, measuring, and managing customer-based brand equity.
Journal of Marketing, 57(1), 1 22.
Keller, K. L. (2000). The brand report card. Harvard Business Review, 78, 147 154.
Keller, K. L. (2012). Strategic brand management: Building, measuring, and managing band
equity (4th ed.). Upper Saddle River, NJ: Prentice Hall.
Klein, N. (1999). No logo: Taking aim at the brand bullies. New York, NY: Alfred A. Knopf.
Knobel, M., & Lankshear, C. (2008). Remix: The art and craft of endless hybridization.
Journal of Adolescent & Adult Literacy, 52(1), 22 33.
Kozinets, R. V., & Handelman, J. M. (2004). Adversaries of consumption: Consumer move-
ments, activism, and ideology. Journal of Consumer Research, 31, 691 704.
Lasn, K. (1999). Culture jam: The uncooling of America. New York, NY: William Morrow.
Lessig, L. (2008). Remix: Making art and commerce thrive in the hybrid economy. New York,
NY: Penguin Press.
Lipson, H., & Kurman, M. (2013). Fabricated: The new world of 3D printing. Indianapolis:
Wiley.
Lomas, N. (2013). Nokia releases 3D print files for Lumia 820 smartphone. Retrieved from
http://techcrunch.com/2013/01/18/nokia-releases-3d-print-files-for-lumia-820-smart-
phone-got-a-3d-printer-custom-print-your-own-removable-shell/
Madden, T. J., Fehle, F., & Fournier, S. (2006). Brands matter: An empirical demonstration
of the creation of shareholder value through branding. Journal of the Academy of
Marketing Science, 34(2), 224 235.
Manovich, L. (2005). Remixing and remixability. Retrieved from http://www.manovich.net/
TEXTS_07.HTM#articles
Manovich, L. (2007). What comes after remix? Retrieved from http://www.manovich.net/
TEXTS_07.HTM#articles
McAlexander, J. H., Schouten, J. W., & Koenig, H. F. (2002). Building brand community.
Journal of Marketing, 66(January), 38 54.
Merges, R. P. (2007). Locke remixed. UC-Davis Law Review, 40, 1259 1273.
Meyer, D. (2013). Why Nokia’s 3D-printing move embraces the future. Retrieved from http://
gigaom.com/2013/01/18/why-nokias-3d-printing-move-embraces-the-future/
Moreau, C. P., & Herd, K. B. (2010). To each his own? How comparisons with others influ-
ence consumers’ evaluations of their self-designed products. Journal of Consumer
Research, 36(5), 806 819.
 80 ARIC RINDFLEISCH AND MATTHEW O’HERN

Muniz, A. M., Jr., & O’Guinn, T. C. (2001). Brand community. Journal of Consumer
Research, 27(March), 412 432.
Muniz, A. M., Jr., & Schau, H. J. (2005). Religiosity in the abandoned apple newton brand
community. Journal of Consumer Research, 31, 737 747.
Navas, E. (2010). Regressive and reflexive mashups in sampling culture. In S. Sonvilla-Weiss
(Ed.), Mashup cultures (pp. 157 177). New York, NY: Springer.
Nokia. (2013). 3D printing challenge. Retrieved from https://ideasproject.com/web/3d-printing
Ogawa, S., & Piller, F. T. (2006). Reducing the risks of new product development. MIT Sloan
Management Review, 47(2), 65 71.
O’Hern, M. S., & Rindfleisch, A. (2010). Customer co-creation: A typology and research
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)

agenda. In N. K. Malholtra (Ed.), Review of marketing research (Vol. 6, pp. 84 106).

Armonk, NY: M.E. Sharpe.
Park, C. W., Jaworski, B. J., & MacInnis, D. (1986). Strategic brand concept image manage-
ment. Journal of Marketing, 50, 135 145.
Park, C. W., Milberg, S., & Lawson, R. (1991). Evaluation of brand extensions: The role of
product feature similarity and brand concept consistency. Journal of Consumer
Research, 18, 185 193.
Poetz, M. K., & Schreier, M. (2012). The value of crowdsourcing: Can users really compete
with professionals in generating new product ideas. Journal of Product Innovation
Management, 29(2), 245 256.
Poschardt, U. (1998). DJ culture. London: Quartet Books.
Ratto, M., & Ree, R. (2012). Materializing information: 3D printing and social change. First
Monday, 17. Retrieved from http://firstmonday.org/ojs/index.php/fm/article/view/3968/
3273
Russo, J. L., & Coppa, F. (2012). Fan/remix video (a remix). Transformative Works and
Cultures, 9. Retrieved from http://journal.transformativeworks.org/index.php/twc/
article/view/431/304
Schau, H. J., Muniz, A. M., Jr., & Arnould, E. J. (2009). How brand community practices
create value. Journal of Marketing, 73, 30 51.
Schouten, J. W., & McAlexander, J. H. (1995). Subcultures of consumption: An ethnography
of the new bikers. Journal of Consumer Research, 22, 43 61.
Senevirante, O., & Monroy-Hernandez, A. (2010). Remix culture on the web: A survey of con-
tent reuse on different user-generated content websites. Retrieved from http://dig.csail.
mit.edu/2010/Papers/WebScience/paper.pdf
Softpedia. (2012). Sony says that little big planet has more than 6 million user-made levels.
Retrieved from http://news.softpedia.com/news/Sony-Says-that-Little-Big-Planet-Has-
More-than-6-Million-User-Made-Levels-249933.shtml
Spence, E. (2013). What will the 3D printing crowd do with Nokia’s templates for the Lumia
820? Forbes.
Thompson, C. J., Rindfleisch, A., & Arsel, Z. (2006). Emotional branding and the strategic
value of the doppelgänger brand image. Journal of Marketing, 70, 50 64.
Volpe, J. (2014). A chip off the printing block: Samsung partners with 3D systems for custom
note 3 cases. Retrieved from http://www.engadget.com/2014/01/07/samsung-3d-sys
tems-printing-case/
von Hippel, E. (2005). Democratizing innovation. Cambridge, MA: MIT Press.
von Hippel, E., Ogawa, S., & de Jong, J. P. J. (2011). The age of the consumer-innovator.
Sloan Management Review, 53, 27 35.
 Brand Remixing: 3D Printing the Nokia Case 81

Willans, J. (2013). Everything you need to know about the Lumia 820 and 3D printing. Retrieved
from http://conversations.nokia.com/2013/01/18/everything-you-need-to-know-about-
the-lumia-820-and-3d-printing/
Yoo, B., Donthu, N., & Lee, S. (2000). An examination of selected marketing mix elements
and brand equity. Journal of the Academy of Marketing Science, 28(2), 195 211.
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)
 This article has been cited by:

1. 2015. Branding as risk management. Strategic Direction 31:10, 12-14. [Abstract]

[Full Text] [PDF]
Downloaded by Chinese University of Hong Kong At 20:01 15 February 2016 (PT)
 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

Fr om N a u t ilu s t o N a n obo( a ) t s: Th e Visu a l Con st r u ct ion of N a n oscie n ce

Br igit t e N e r lich
Copyright AZoM.com Pt y Lt d.
This is an AZo Open Access Rewards Syst em ( AZo- OARS) art icle dist ribut ed under t he t erm s of t he AZo–OARS
ht t p: / / www .azonano.com / oar s.asp which perm it s unrest r ict ed use provided t he original work is properly cit ed but is
lim it ed t o non- com m er cial dist ribut ion and r eproduct ion.

Subm it t ed: May 24 t h , 2005

Post ed: Decem ber 22 nd 2005

DOI : 10.2240/ azoj ono0109

Topics Cove r e d
Abst ract
Background
Nanoscience and Technology
Nanosubm arines
Popular Cult ure and Nanoscience
Nanocult ure and Nanowrit ing
Nanobot s, Nanom achines and Science Fict ion
The Various Guises of Nanobot s
Nanoboat s Break Down Barriers bet ween Fict ion and Realit y
Nanot echnology in Children’s Cult ure
Realit y is Cat ching up t o Science Fict ion
Nanosubm arines in The Press
Nanoboat s in Defence Applicat ions
An I conography of The Nano- Boat : From Nem o t o Nano
Evolut ion of t he Nano- Subm arine
Envisioning t he Fut ure of Nanot echnology
The Root s of Nanot echnology – Twent y Thousand Leagues under The Sea
Nanoboat s Know No Bounds
Naut ilus, A Tem plat e for Fut urist ic Travel
The Naut ilus Myt h Merges wit h The Nano Myt h
Naut ilus Meet s t he I ndust rial Revolut ion
The Nanoboat Begins I t s Voyage
Fant ast ic Voyage Meet s Fut uram a
A React ion t o Fant ast ic Voyage Meet s Fut uram a
I nner Space and Nanom edicine
Molecular Nanot echnology Nearly I nvent ed in 1942
There is Plent y of Room at The Bot t om
Fict ional Nano- Machines becom e Part of The Fact ual Nano- Discourse
Science Fict ion Evolves by Applying Miniat urizat ion t o Exist ing I deas
Science Fict ion and Moore’s Law
Journey of The Naut ilus Cont inues
The Naut ilus Turns int o A Nano- Space- Ark
Nano- Space Travel
Self Replicat ion
Speculat ion About The Fut ure of Nanot echnology
Conclusion
Acknowledgem ent
References
Cont act Det ails

Volum e 1 | Decem ber 2005 Page 1 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

Abst r a ct

The view t hat nanot echnology will lead t o t iny robot ic subm arines navigat ing our bloodst ream
is ubiquit ous but t here is an alm ost surreal gap bet ween what t he t echnology is believed t o
prom ise and what it act ually delivers. Vehicular ut opias, such as Jules Verne’s voyage under
t he sea, and dream m achines, such as nanorobot s, have t ended t o fill t his surreal gap and
have had an im m ediat e and long last ing hold on public im aginat ion. They cont inuously serve t o
sciencefict ionalise science fact and blur t he boundaries bet ween cult ural visions and scient ific
realit y. This paper exam ines t he visual and verbal im agery surrounding t he various
‘subm arines’ t hat have t ravelled t hrough popular im aginat ion, from Jules Verne’s Naut ilus,
driven by Capt ain Nem o, up t o t he m ost iconic and m ost recent represent at ion of
nanot echnology, from t he j ourney t hrough t he hidden space of t he world’s oceans on board
t he hidden space of a luxurious subm arine, t o expedit ions int o t he hidden space of t he hum an
body as port rayed in film s such as Fant ast ic Voyage, I nner Space and beyond. The paper aim s
t o show t hat popular cult ure and im aginat ion do not sim ply follow and reflect science. Rat her,
t hey are a crit ical part of t he process of developing science and t echnology; t hey can inspire
or, indeed, discourage researchers t o t urn what is t hinkable int o new t echnologies; and t hey
can fram e t he ways in which t he ‘public’ react s t o scient ific innovat ions. Fict ional im ages, be
t hey lit hographs in children’s books or st ills from popular sci- fi film s, play an im port ant part in
t his process.
I n m em ory of t he cent enary of Jules Verne’s deat h in 2005
“ Ah! Science never goes fast enough for us! ”
( Art hur Rim baud, The I m possible)

Ba ck gr ound

“ Nanot echnology is a rapidly advancing and t ruly m ult idisciplinary field involving subj ect s such
as physics, chem ist ry, engineering, elect ronics and biology. I t s aim is t o advance science at
at om ic and m olecular level in order t o m ake m at erials and devices wit h novel and enhanced
propert ies. The pot ent ial applicat ions are except ionally diverse and beneficial, ranging from
self- cleaning windows and clot hes t o ropes t o t et her sat ellit es t o t he eart h’s surface, t o new
m edicines” [ 1] . Som e, however, have claim ed t hat nanot echnology m ight lead t o
nanoassem blers and pot ent ially self- replicat ing nanom achines swam ping life on eart h. This so-
called ‘grey goo’ scenario was first described by Drexler in his book Engines of Creat ion [ 2] , a
scenario he now t hinks unlikely ( as of 11 June, 2004) , but which has been int egrat ed int o
fict ional narrat ives such as Michael Cricht on’s novel Prey [ 3] , has provoked com m ent s by t he
Prince of Wales, lead t o inquiries by t he Royal Societ y and t he Royal Academ y of Engineering
in t he UK, and, in general, t aps int o wide- spread fears about loss of cont rol over t echnologies
[ 4, 5] . Anot her popular im age of nanot echnology, t he view t hat it will lead t o t iny robot ic
subm arines t ravelling t hrough t he hum an bloodst ream and repairing or healing our bodies is
equally ubiquit ous.

N a n oscie nce a nd Te ch n ology

However, t here is, as Richard Jones point ed out in a recent art icle on “ The Fut ure of
Nanot echnology” , an alm ost a surreal gap bet ween what t he t echnology is believed t o prom ise
( or t hreat ens t o creat e) and what it act ually delivers [ 4] – a gap t hat cert ain science fict ion
scenarios can easily fill, be t hey of t he dyst opian, grey goo t ype, or t he ut opian, spect acular
voyage, t ype. This view was echoed by López in his 2004 art icle “ Bridging t he Gaps: Science
fict ion in nanot echnology” , in which he explores how nanoscient ist s t hem selves have em ployed
devices from sci- fi lit erat ure t o argue t he case for nanoscience. He com es t o t he conclusion
t hat “ t he relat ion bet ween SF narrat ive elem ent s and NST [ nanoscience and t echnology] is not
ext ernal but int ernal. This is due t o NST’s radical fut ure orient at ion, which opens up a gap
bet ween what is t echnoscient ifically possible t oday and it s inflat ed prom ises for t he fut ure.”
[ 6] .
Vehicular ut opias [ 7] , from Jules Verne’s Naut ilus voyaging under t he sea t o nanom achines

Volum e 1 | Decem ber 2005 Page 2 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

navigat ing t hrough our bloodst ream s, have, for a long t im e, filled t his surreal gap bet ween t he
t echnologically possible and t he t echnologically real and can have an im m ediat e and long
last ing hold on public im aginat ion. They cont inuously serve t o sciencefict ionalise science fact
and blur t he boundaries bet ween cult ural visions and scient ific realit y [ 8,9] .

N a nosubm a r ine s
This paper exam ines t he visual and verbal im agery surrounding t he various ‘subm arines’ t hat
have t ravelled t hrough popular im aginat ion from Verne’s Naut ilus, driven by Capt ain Nem o, up
t o t he m ore recent represent at ions of nano- subm ersibles, som e of t hem fact , som e of t hem
fict ion and m any m ore float ing bet ween t he t wo. I t ret races t he voyages of iconic subm arines
from t he 1870s, when Verne wrot e Twent y Thousand Leagues under t he Sea, up t o t he
present , from t he j ourney t hrough t he hidden space of t he world’s oceans on board t he hidden
space of a luxurious subm arine, t o expedit ions int o t he hidden space of t he hum an body and
beyond t o out er space.

Popu la r Cult u r e a n d N a n oscie nce

The aim of t his paper is t o dem onst rat e t hat popular cult ure and im aginat ion do not sim ply
follow and reflect science. Rat her, t hey are a crit ical part of t he process of developing science
and t echnology; t hey can inspire or, indeed, discourage researchers t o t urn what is t hinkable
int o new t echnologies and t hey can fram e t he ways in which t he ‘public’ react s t o scient ific
innovat ions. Popular cult ure t alks about space- rocket s before t here are space- rocket s, t est -
t ube babies before t here are t est - t ube babies and clones before t here are clones. Before
scient ist s do anyt hing t here is oft en a ready m ade public percept ion of how good or how bad it
is going t o be, derived from t his social, lit erary and cult ural precognit ion. So when science
does t hese t hings for real, t heir im age has already been form ed - for good or ill. Fict ional
im ages, be t hey lit hographs in 19 t h - cent ury children’s books, st ills from popular sci- fi film s or
nano- illust rat ions produced by professional science illust rat ors play an im port ant part in t his
process. I t m ight be going t oo far t o say t hat science can only discover what has already been
creat ed in im aginat ion, but it is cert ainly t he case t hat science can only flourish in societ y when
popular im aginat ion does not st rongly oppose it s developm ent . This m ight be why som e
nanoscient ist s have also becom e nano- visionaries, act ively involved in creat ing an im aginat ive
and im aginary space for nanoscience in m odern societ y t hrough writ ing and illust rat ions.
I n t he following I shall first sum m arise som e insight s int o t he way science and fict ion int eract
inside nanoscience, t hen provide an overview of t he iconographical voyage of progress
accom plished in and by t he Naut ilus, t o be followed by a sect ion in which t he Naut ilus m eet s
ot her science fict ion influences and m erges wit h nanoscience proper, only t o be t aken, finally,
from t he inside of t he hum an body int o out er space. I shall t hen t ry t o draw som e conclusions
from t his voyage t hrough visual space and t im e.

N a n ocu lt ur e a nd N a n ow r it ing
A recent ly published book ent it led Nanocult ure: I m plicat ions of t he new t echnoscience begins
wit h t he sent ence “ I m agine a world …” [ 10] and goes on t o describe a world, our world, in
which nanoscience and nanofict ion have begun t o int erpenet rat e each ot her in m yriad ways
and where t he boundaries bet ween t he lit eral and t he m et aphorical and t he real and t he
im aginary have becom e ut t erly blurred.

N a n obot s, N a n om a chin e s a n d Scie n ce Fict ion

There is one discursive knot in what t he cont ribut ors t o t he book call ‘nano- writ ing’ where t his
blurring becom es m ost obvious: t he so- called nanobot . Nanobot s have long been t he st uff of
science fict ion and t hey have, m ore recent ly, becom e t he st uff of science- fict ionalisat ion when
som e nanoscient ist s t alk about t hese not - yet - exist ing but soon- t o- exist nanom achines as if
t hey were as real as t he m et aphorical m ot ors, m achines or pum ps t hat float t hrough our body
in t he shape of enzym es or part s of bact eria. As one nano- writ er point s out in an art icle
ent it led “ Of silicone and subm arines” : “ Aft er all, t he m ost successful nanom achines ever
creat ed are t hose operat ing inside every cell” . [ 11] Hence, writ es Bensaude- Vincent , “ t he

Volum e 1 | Decem ber 2005 Page 3 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

a nanom achine?’ However t he not ion of m achine is it self polysem ic, so t hat it can support
dissim ilar views of living syst em s and t each quit e different lessons t o nanoscient ist s and
engineers” [ 12; see 13] . I t can also support dissim ilar im ages of t he fut ure and t each quit e
different lessons t o t he scient ifically int erest ed public. The nanom achine I am int erest ed in in
t his art icle is t he nanosubm ersible, which has m ainly posit ive associat ions, unlike t he nano-
assem bler t hat replicat es and dest roys t he eart h. These t wo visions of a ut opian or dyst opian
nano- fut ure seem t o m at ch ont o t he different discourses of hope and fear associat ed wit h
eit her m edical GMOs ( genet ically m odified organism s) , which are regarded as rat her beneficial,
and environm ent al GMOs, i.e. foods and crops, which are not [ 14] .

Th e Va r ious Gu ise s of N a n obot s

Nanom achines, in t he shape of nanobot s, flip not only bet ween t he corporeal and t he
m echanical, t he m et aphorical and t he lit eral and t he fict ional and t he fact ional in an alm ost
quant um - m echanical way, t hey also flip bet ween good and evil, t he past and t he present and
t he present and t he fut ure. They can invade or heal t he body, dest roy t he world or m ake it a
bet t er place. They are part of a fut ure t hat will, as m any nanowrit ers say, inevit ably becom e
t he present and t hey have been part of our past im aginat ion for m any years. One incarnat ion
of t he nanobot in part icular, t he nano- subm arine or, as one can call it , nanoboat , has, as we
will see, a long and illust rious fict ional and visual ancest ry, which links t he fut ure t o t he past ,
woodcut s t o com put er generat ed im ages and lit hographs t o nanolit hography [ 15] .
The m ot or of m et aphorical im aginat ion which powers fict ional im aginat ion, be it verbal or
visual, nam ely seeing som et hing as som et hing else, is all- im port ant in t he process of flipping
bet ween body and m achine, fact and fict ion, past and fut ure, hope and fear. Machines are
seen as biological phenom ena, biological phenom ena, including hum an bodies are seen as
m achines, sm all obj ect s are seen as large obj ect s and large obj ect s are seen as sm all ones,
t he out side is seen in t erm s of t he inside and t he inside in t erm s of t he out side, science is seen
in t erm s of fict ion and fict ion in t erm s of science. This reciprocal m et aphorisat ion of t he real
and t he not - yet real in nanowrit ing is quit e unlike t radit ional uses of m et aphor and im agery in
science and of science in fict ion where scient ist s use m et aphors or m et aphorical m odels t o
describe as yet unknown aspects of t he real world and where science fict ion writ ers use
science as a st art ing point beyond which t o proj ect im aginary scenarios – from t he known t o
t he unknown and from t he esot eric world of t he lab t o t he lay world of ordinary discourse. By
cont rast :
The blurring of pot ent ialit y and act ualit y in t he nanoworld, and t he lack of general knowledge
about nanot echnology, creat e a fert ile im aginary around t he new discipline. A com m on
night m are speculat es t hat , wit h t he aid of nanot echnology, researchers will build
nanost ruct ures capable of replicat ing t hem selves like nano- robot s. UCLA Professor Jam es
Gim zewski relat es t hat when he worked at I BM “ a newspaper called t he Bild print ed a front
page st ory saying ‘I BM creat es nanobot s t hat can cure cancer’ wit h a pict ures of t hem
swim m ing inside t he hum an body and describing it as having a cancer- killing unit t hat used
lasers t o ‘blast away’ t he cancer cells.” I m m ediat ely, t here were people from all over t he world
calling I BM and asking how t o get t hese nano- bot s. [ 16]

N a n oboa t s Br e a k D ow n Ba r r ie r s be t w e e n Fict ion a nd Re a lit y

This st ory was not t rue ( at t he t im e) but it is indicat ive of how iconic and at t he sam e t im e real
nanoboat s have becom e in t he public im aginat ion, in t his case reproduced in t he shape of a
t abloid pict ure and nano- hyperbole [ 17] . Nano- writ ing, be it in scient ific m agazines, in novels
or in t abloids, “ rem oves” , as Milburn point s out “ all int ellect ual boundaries bet ween organism
and t echnology” and “ causes ‘t he dist inct ion bet ween hardware and life… t o blur’ – and hum an
bodies becom e post hum an cyborgs, inext ricably ent wined, int erpenet rant , and m erged wit h
t he m echanical nanodevices already inside of t hem .” [ 8] The im aginary, not yet real, nanoboat
swim m ing inside our body is concept ualised as if it were already real, as ‘real’ as t he biological
‘m achines’ swim m ing inside our body on which t he nanoboat is m odelled.

Volum e 1 | Decem ber 2005 Page 4 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

N a n ot e ch nology in Ch ildr e n ’s Cu lt u r e
Whereas som e part s of public/ adult im aginat ion about nanot echnology have been nurt ured by
t he t abloids and by popular novels ( where nano has been port rayed m ost ly in dyst opian ways) ,
children’s nano- im aginat ion has been nurt ured by cart oons, com ics, novels and com put er
gam es, in which nano has perhaps not only negat ive associat ions ( but m ore research is
needed here) . What links t he t wo, adult and children’s im aginat ion, are im ages t aken from
children’s lit erat ure and film s, such as Verne’s novel and it s Disney film adapt at ion Twent y
Thousand Leagues under t he Sea, which were t hen t ransferred t o adult film s, such as Fant ast ic
Voyage and I nner Space and t ransferred back again t o children’s m edia, such as Dr Who and
I nvader Zim ( see t able 1) , wit h t he ( nano) subm ersible weaving forwards and backwards
bet ween m edia and audiences. This process of int erpenet rat ion also involves a give and t ake
bet ween ‘éducat ion et récreat ion’ ( t he m ot t o under which Verne’s novels were published in t he
19t h cent ury) , bet ween fact and fict ion and bet ween rat ionalit y and im aginat ion. This seem s t o
cont radict Pet er Weingart ’s claim t hat “ [ s] ociet y obviously does not consider science a m at t er
of am usem ent ” [ 18] . When it com es t o nanobo( a) t s it cert ainly does.
Modern children’s fict ion refers t o nanoscale creat ures and m achines so frequent ly t hat flippant
rem arks about ‘nano’ have becom e perm issible. I n a recent ly shown episode of t he Disney
cart oon Kim Possible one charact er point s t o som et hing and calls it ‘nano’. The ot her charact er
asks what t hat m eans and is t old: “ Sm all, m ini, t iny, m inut e.” Asked why he t hen didn’t say
‘m ini’, he replies: “ because nano sounds a t housand t im es bet t er, why else?” I n short : Nano is
cool.

Re a lit y is Ca t ch in g up t o Scie n ce Fict ion

But j ust as language, especially t eenage slang and ‘t eenage’ t echnology in t he form of t he iPod
Nano, is cat ching up wit h im aginary nanobo( a) ts, so realit y m ay be cat ching up wit h t hem ,
t oo. I n a recent art icle about nanoscience ( chosen am ongst m any) we are t old about a new
I nst it ut e at Leeds Universit y dealing wit h nanot echnology:
Molecular- scale ‘t rains’ and ‘subm arines’ t hat will carry loads such as t iny doses of drugs and
virt ual realit y soft ware t o enable operat ors t o cont rol m at t er on t he nanoscale are proj ect s
planned by t he I nst it ut e.
[ …]
Professor Pet er St ockley [ …] said: “ [ …] I n t he fut ure we could im agine an engineered ‘nano-
subm arine’ swim m ing around a pat ient ’s bloodst ream t o t he sit e of a t um our t oo sm all t o be
t ackled by surgery.” [ 19]

N a n osu bm a r in e s in Th e Pr e ss
I n t he year 2000 an im age of a t iny robot ic subm arine t ravelling t hrough a hum an art ery
t ravelled round t he world and was even pict ured in t he UK t abloid newspaper The Mirror
( Thursday, 7 Sept em ber, 2000) under t he t it le “ Fant ast ic Voyage 2” . This t iny 4m m long craft
was said t o be able t o save lives before t he end of t he decade, t o be able t o cruise t hrough
blood vessels using sensors t o check for signs of illness and cancer and, it was report ed, m ay
one day be able t o repair art eries and heart s. I t was t here t o dem onst rat e in a visible way
what could already be achieved in t he field of m icro- or nanot echnology. A sim ilar pict ure was
also exhibit ed at t he Hannover Expo 2000 [ see 11] .
The year 2000 im age of a prot ot ype of a nano- subm arine, which circulat ed widely in t he press,
appeared a cent ury aft er a pict ure of Jules Verne’s Naut ilus had graced t he visit ors’ guide t o
t he Paris Expo in 1900 [ 20] , a sign t hat t he Naut ilus had becom e part of m odern m yt hology
( see figure 1 for anot her exam ple) . We seem t o have com e a long way in a cent ury, but we
can’t seem t o leave t he Naut ilus m yt h quit e behind us yet – and, as we shall see, it would soon
m erge wit h t he nano- m yt h. ( I t should be st ressed however t hat t he year 2000 im age was not
t he first . An im age of a nanosubm arine swim m ing t hrough a capillary and at t acking a fat
deposit , such as norm ally m ay accom pany an art eriosclerot ic lesion, was published as early as
1988 in an art icle for t he Scient ific Am erican, for exam ple [ 21] ) .

Volum e 1 | Decem ber 2005 Page 5 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

Figur e 1 . Microsubm arine.

M icr o- su bm a r in e s in t he body, concept ual com put er ar t work. Microscopic m achines such as t hese m ay be
int roduced int o t he body t o augm ent t he body's im m une syst em . The subm arines could be program m ed t o find and
dest roy t um our cells, for exam ple, or t o repair defect s in organs and t issues. CONEYL JAY / SCI ENCE PHOTO
LI BRARY

N a n oboa t s in D e fen ce Applica t ion s

Moving away from nano- m edicine, t he year 2003 brought news t hat ‘nano- fish- boat s’ could be
invent ed t o spy on ( real/ real- size) subm arines:
The concept of “ nano under t he sea” is not so far- fet ched. The m ove t o Advanced SEAL
Delivery Syst em ( ASDS) m inisubs, t o t orpedo- t ube- launched unm anned undersea vehicles
( UUVs) for m inefield surveillance and ot her dangerous dut y, and t he planned arm ed unm anned
undersea com bat vehicle MANTA, can be viewed collect ively as a t rend t oward m iniat urizat ion
serving t win purposes: increasing st ealt h, and m inim izing casualt ies if st ealt h were
com prom ised. [ 22]
Before t hese various m edical or m ilit ary fut ures cat ch up wit h t he present , I would like t o look
at t he past and explore t he genealogy and iconography of im aginary nano- bo( a) t s, especially
nano- subm arines, t hat have nurt ured popular im aginat ion and m ight st ill be nurt uring t he
im aginat ion of art ist s who produce nano- illust rat ions deposit ed at t he ‘Science Phot o Library’,
for exam ple, from which som e of t he illust rat ions for t his art icle have been t aken. Such
im ages, which draw on bot h t echnical and aest het ic expert ise, depict t he progress of
nanoscience while at t he sam e t im e driving it forwards.

Volum e 1 | Decem ber 2005 Page 6 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

An I con ogr a ph y of Th e N a no- Boa t : Fr om N e m o t o N a no

As shown in previous st udies [ 23, 24, 25, 8] popular cult ure and im aginat ion do not sim ply
follow and reflect science; rat her, t hey lead and ant icipat e developm ent s in science and
t echnology. This is nowhere m ore apparent t han in nanoscience, especially as far as t he
nanoboat is concerned. Let us now look m ore closely at t he way it has t ravelled bet ween
popular and scient ific and adult and j uvenile im aginat ion.

Evolu t ion of t h e N a no- Subm a r in e

The following ( rat her incom plet e) t able chart s t he origins and developm ent of t he nano-
subm arine t hat has enchant ed readers and viewers for m ore t han a cent ury and a half:
Ta ble 1 . Nano- sciencefict ionalisat ion ( The shaded it em s are not direct ly relat ed t o
nanosubm arines) .

( Som e ) Fict iona l/ Visu a l ( Som e ) Fict iona l/ N a r r a t ive N a n oscie nce m e e t s
in flue n ce s on na noscie n ce in flue n ce s on na noscie n ce n a nofict ion

1 8 6 9 Jules Verne: 20,000

Leagues under t he sea:
popularises lit hographs of t he
Naut ilus
1 9 1 6 silent film version of
Verne’s book
1 9 5 4 Disney version of t he 1 9 4 2 Heinlein publishes
book: popularises new im age Waldo: popularises fict ional
of t he Naut ilus im age of t iny m echanical
surgeons
Toys, gam es, m ock- ups, 1 9 4 7 Eric Frank Russell 1 9 5 9 Feynm an gives
com ics, m any different edit ions publishes Hobbyist : “ at om fed speech: “ There is plent y
of Disney’s 20,000 leagues t o at om like brick aft er brick t o of room at t he bot t om ” –
swam p children’s im aginat ion build a house.” launches nanoscience –
1 9 5 5 Russell publishes serial possibly influenced by
Call him dead: “ surgical and Heinlein
m anipulat ory inst rum ent s so
t iny t hey can be used t o
operat e on a bacillus.”
1 9 6 6 Fant ast ic Voyage: film
popularises im age of nano-
surgeons and nano-
subm arines; im age of sub
called Prot eus is based on
popular 1954 im age of 1 9 7 3 Mahr Beware t he
Naut ilus Microbot s
1 9 7 7 Episode of Dr Who: The
invisible enem y parodies 1 9 8 5 Bear Blood Music ( A
Fant ast ic Voyage scient ist has been carrying out
1 9 8 2 FV released as com put er research behind t he back of t he
gam e biot ech firm he works for.
When t he com pany finds out ,
he is fired and ordered t o
dest roy his work. Sm uggles it
out in his bloodst ream )
1 9 8 7 I nner Space – parodies 1 9 8 6 Drexler published
Fant ast ic Voyage Engines of Creat ion
1 9 9 5 Goonan st art s writ ing 1 9 9 9 1999 Freit as

Volum e 1 | Decem ber 2005 Page 7 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

sci- fi st ories about nano; published Nanom edicine,

St ephenson publishes Diam ond first m edical
Age nanorobot ics book.
2 0 0 1 Episode of Fut uram a 2 0 0 2 Cricht on’s novel Prey 2 0 0 0 pict ure of nano
parodies Fant ast ic Voyage and put s nano and ‘grey goo’ on t he sperm sort er
Dr Who episode; episode of narrat ive m ap
“ I nvader Zim ” , called 2 0 0 0 iconic im age of a
NanoZim parodies I nner Space nano- subm arine
2 0 0 4 Marlow publishes Nano- a 2 0 0 4 fact / fict ion im age
nano- ‘swarm ’ t hriller of nanobot appears on
Physicsweb
I n a recent art icle on nano- hyperbole, Chris Toum ey wrot e:
One of t he ways people t ry t o envision t he fut ure of nanot echnology is t o t ell st ories about t he
past , expect ing t hat t he fut ure will cont inue cert ain feat ures of t he past . I f one t ells st ories
which em phasize t hat t he founders of nanot echnology past were heroic geniuses, for exam ple,
t hat kind of em phasis would bless nanot echnology present and fut ure as a noble effort whose
heroic qualit ies endure. [ 17]

En vision ing t h e Fu t ur e of N a n ot e ch n ology

Anot her way in which people t ry t o envision t he fut ure of nanot echnology is t o show and use
im ages of t he past , including im ages of heroic geniuses, t heir vehicles and voyages, expect ing
t hat t he fut ure will cont inue t hese im ages of t he past in t he sam e glorious, spect acular and
fant ast ic way.

Th e Root s of N a not e ch n ology – Tw e n t y Thou sa n d Le a gue s u nde r The Se a

I f we want t o find t he root s of som e of t he m ost popular im ages of nanot echnology, we have
t o look back t o t he 19t h cent ury, t he cent ury of indust rial revolut ion and scient ific progress, in
part icular t o t he year 1869 when Jules Verne published Twent y Thousand Leagues under t he
Sea [ 26] ( and one should not forget t hat som e describe nanoscience as t he next indust rial
revolut ion [ 27] ) . The novel describes t he advent ures of Professor Aronnax ( a nat uralist and
scient ist ) , his servant Conseil ( nam ed aft er an invent or of a real subm arine which was t est ed
in Paris in 1859) and Ned Land, a whale hunt er, st randed on board t he subm arine Naut ilus,
st eered by t he nam eless and t im eless, m isant hropic Capt ain Nem o, t he forefat her of m odern
superheroes.
This novel, illust rat ed wit h drawings by Alphonse de Neuville and Edouard Riou ( a st udent of
Gust ave Doré’s, t he m ost popular and successful French book illust rat or of t he m id- 19t h
cent ury) , was t he fift h in t he series of novels ent it led Voyages Ext raordinaires, which also
cont ained t he fam ous novels I n 80 days around t he World, From t he Eart h t o t he Moon, and A
Journey t o t he Cent re of t he Eart h. Twent y Thousand Leagues under t he Sea becam e one of
t he m ost popular of t hese advent ure and discovery novels designed t o ‘t each’ t he reading
bourgeois public, and m ost im port ant ly children, in an am using way about science and
t echnology. Before being published as print ed books, t he Voyages Ext raordinaires first
appeared in t he form of feuillet ons in t he Magasin d’Educat ion et de Récréat ion, published by
P. J. Het zel in Paris. Pict ures of ot herwise invisible places, port rait s of heroic t ravellers, im ages
of sensat ional vehicles t ravelling on land, t hrough t he air and under t he sea, m aps,
represent at ions of spect acular fauna and flora, et c. becam e a m eans of bringing t he unknown,
unexplored, gigant ic or invisible ( at t he bot t om of t he ocean, in t he cent re of t he eart h, on t he
m oon, et c.) t o a curious readership. Bet ween 1920 and 1980 alone Vingt m ille lieues sous les
m ers was reprint ed 12 t im es wit h it s original illust rat ions, unt il 1919 by t he publisher Het zel,
aft er t hat by Hachet t e, and 15 t im es wit h new illust rat ions adapt ed t o t he t ast e of t he
audiences of t he t im e. From 1930 onwards Hachet t e com m issioned illust rat ions t hat would
appeal m ore specifically t o children; from 1960 onwards a general increase in int erest in
science fict ion ( especially aft er t he m oon landing) prom pt ed Hachet t e t o com m ission
illust rat ions wit h a m ore avant - garde or m odernist feel [ 28, 29] .

Volum e 1 | Decem ber 2005 Page 8 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

I n t his process t he Naut ilus becam e, t hrough various incarnat ions in print and film , t he ‘I con of
t he spaceship’, as Gary Wolfe has point ed out in his book The Known and t he Unknown: The
iconography of Science Fict ion [ 30] ( however, I would argue, since 1966, in iconographic
com pet it ion wit h t he St arship Ent erprise) .

Figur e 2 . Original illust rat ion of t he Naut ilus.

N a n oboa t s Kn ow N o Bou nds

The as yet virt ual space explored by fut ure nanoboat s would be t he inside of t he hum an body.
I t all began however wit h t he explorat ion of real, but largely im aginary, spaces on eart h. The
Voyages Ext raordinaires begin an explorat ion of places of im aginat ion t hat would event ually
span t he physical ( from unexplored lands t o out er space) , t he virt ual and, finally, nano, where
t he unknown “ is folded wit hin t he known obj ect s and spaces we inhabit in our everyday life”
[ 16] , including our own bodies. By cont rast , in Verne’s case, 19t h- cent ury everyday living ( of
t he upper classes) is folded wit hin t he explorat ion of t he unknown out side. The luxurious inner
space of t he Naut ilus, inhabit ed by it s heroic invent or Nem o, becam e even bet t er known t han
it ’s out er shape:

Volum e 1 | Decem ber 2005 Page 9 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

Figur e 3 . Original illust rat ion of t he int erior of t he Naut ilus.

The voyages ext raordinaires explore worlds known and unknown: t he int erior of Africa, t he
int erior of t he Eart h, t he deeps of t he sea, t he deeps of space. Charact erist ically, Verne’s
voyagers t ravel in vehicles t hat are t hem selves closed worlds- his im aginat ion proj ect s it self in
t erm s of “ inside” and “ out side” - from which t he im m ensit y of nat ure can be appreciat ed in
upholst ered com fort . The Naut ilus is t he m ost fam iliar of t hese com fort able, m obile worlds;
inside all is cozy elegance, t he epit om e of t he civilized and hum an, while out side t he oceans
gleam or rage in inhum an beaut y or m yst ery. Roland Bart hes finds t he principle at t he heart of
Verne’s fict ions t o be t he “ ceaseless act ion of secluding oneself.” The known and enclosed
space, t he com fort able cave, is safe while “ out side t he st orm , t hat is, t he infinit e, rages in
vain.” The basic act ivit y in Verne is t he const ruct ion of closed and safe spaces, t he
enslavem ent and appropriat ion of nat ure t o m ake a place for m an t o live in com fort . “ The
enj oym ent of being enclosed reaches it s paroxysm when, from t he bosom of t his unbroken
inwardness, it is possible t o wat ch, t hrough a large window- pane, t he out side vagueness of t he
wat ers, and t hus define, in a single act , t he inside by m eans of it s opposit e.” [ 7]

Volum e 1 | Decem ber 2005 Page 10 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

Figur e 4 . Nem o looking out t hrough window of Naut ilus.

N a u t ilu s, A Te m pla t e for Fu t u r ist ic Tr a ve l

Transferred from out er space t o inner space, t he Naut ilus has becom e t he icon of t he nano-
boat and at one and t he sam e t im e t he icon of any ‘fant ast ic’ or ‘im aginary’ voyage
what soever. I t becam e a generic space- ship on which ‘m icronaut s’ began t o t ravel t hrough
‘inner space’, in an effort t o appropriat e ‘nat ure t o m ake a place for m an t o live in com fort ’. I t
also cam e t o st and iconically for t he ‘progress of science’, where t he im age of science as a
voyage or j ourney is m apped ont o 19t h- cent ury hopes invest ed in t he posit ive out com es of
t hat j ourney, hopes which are st ill being conj ured up by m any biot ech or nanot ech
ent repreneurs.

Th e N a u t ilu s M yt h M e r ge s w it h Th e N a n o M yt h
Since 1869 t he Naut ilus has been represent ed in various shapes and form s, first in black and
whit e t hen in colour in m yriads of children’s’ books and com ics t oget her wit h it s heroic
inhabit ant , Capt ain Nem o. But it also began t o ‘m ove’: first in a rat her silly adapt at ion by
George Méliès in 1907, t hen in a silent m ovie in 1916 and, finally, in 1954, t he Naut ilus
resurfaced in it s m ost canonical form in Disney’s m ovie adapt at ion of Verne’s novel. I t should
be st ressed t hat Walt Disney had t he m ost fam ous Hachet t e edit ion of Twent y Thousand
Leagues under t he Sea on his bookshelf ( as report ed by t he French newspaper Le Figaro in
1952) and had carefully st udied t he original illust rat ions by Riou and de Neuville [ 29] . Nem o
and t he Naut ilus were now t ransposed from t he Vict orian age of indust rial progress t o t he
at om ic age of post - war fear and st art ed t o assum e t he double- edged associat ions of hope and
fear which it would m aint ain once it had assum ed nano- shape.

N a u t ilu s M e e t s t he I n du st r ia l Re volu t ion

The age of t he I ndust rial Revolut ion was t aken up and used as a m et aphor for t he elect ronic
and at om ic revolut ions t hat followed World War I I . As Am ericans looked forward wit h
excit em ent t o t he 1954 launching of Hym an Rickover’s “ At om Sub” significant ly called t he

Volum e 1 | Decem ber 2005 Page 11 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

U.S.S. Naut ilus, Disney was bringing t o life t he nam esake of t he Naut ilus in Capt ain Nem o’s
elect ric subm arine boat and subt ly alt ering it s ideological t ext ure t o fit a new age. [ 7]
From 1954 onwards t he Naut ilus was no longer a Vernian creat ion but a Disney one, sprout ing
num erous reincarnat ions in t he shape of t oys, m ock- ups, com ics, illust rat ed children’s books,
gam es, am usem ent rides and sequels.

Th e N a n oboa t Be gin s I t s Voya ge

I n t he m eant im e, in 1966, t he film Fant ast ic Voyage was released which incorporat ed a
Naut ilus- shaped subm arine, t he Prot eus, int o it s script . The Prot eus was creat ed by t he sam e
m an who had designed t he fam ous Naut ilus for Walt Disney's film adapt at ion of Verne’s novel:
Harper Goff. I n Fant ast ic Voyage t he Naut ilus, under t he nam e of Prot eus, ent ered t he nano-
age by being t ransposed from t he out side t o t he inside of t he body. I n t his film , based on a
sci- fi st ory by t he renowned sci- fi writ er I saac Asim ov ( who had dubbed Jules Verne ‘t he
world’s first science fict ion writ er’) , a group of scient ist s and doct ors m iniat urize t hem selves,
are placed in a m iniat urized sub and are inj ect ed int o t he body of a dying m an in order t o
perform life- saving surgery. This is how t he nanoboat began it s iconic voyage of
popularisat ion.

Figur e 5 . I llust rat ion of t he Fant ast ic Voyage, apparent ly t he original book cover art .
Repr int ed w it h t he perm ission of Prof. Raoul Kopelm an; Laborat ory websit e, Bio- Analyt ical NanoScale Chem ist ry &
Mat er ials, Univ ersit y of Michigan ht t p: / / www .um ich.edu/ ~ k oplab/ r esearch2/ analy t ical/ NanoScaleAnaly sis.ht m l

Fa n t a st ic Voya ge M e e t s Fut u r a m a
Fant ast ic Voyage and it s plot inspired num erous parodies and m et a- parodies. I n 1977, for
exam ple, an episode of Dr Who, “ The invisible enem y” m erged speculat ions about cloning,
which were around since t he m id 1960s [ 31] , and speculat ions about m iniat urizat ion in a
past iche of Fant ast ic Voyage: t he Doct or and his assist ant Leela are cloned, and t he clones are
m iniat urized and inj ect ed int o t he Doct or t o defeat invading parasit es. The voyage leads him t o
t he int erior of his own brain. Anot her parody of bot h Fant ast ic Voyage and Dr Who appeared
as a 2001 episode of Fut uram a ( a cart oon inspired by Mat t Groening, t he creat or of The
Sim psons) , ent it led “ Parasit es Lost ” ( a play on Milt on’s Paradise Lost ) . I n t his parody im ages of
t he nano- sub are m erged wit h im ages of robot ic droids – t he heroes are Fry and Leela ( ! ) .
When Fry eat s an egg- salad sandwich from a vending m achine at a gas st at ion, he begins t o

Volum e 1 | Decem ber 2005 Page 12 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

have st range side- effect s - - he becom es st ronger and sm art er. Professor Farnswort h m akes a
diagnosis and concludes t hat Fry had ingest ed " int elligent worm s" t hat have set up shop in his
body. Because t hese worm s are so sm art , t he regular m eans of flushing t hem out will not
work. The Planet Express crew m ust shrink t hem selves t o m icroscopic form and ent er Fry’s
body t o fight off t he int ruders. Meanwhile Leela is left t o divert Fry’s at t ent ion, but finds herself
m ore and m ore at t ract ed t o t he new m an Fry has becom e. [ 32]

A Re a ct ion t o Fa n t a st ic Voya ge M e e t s Fu t u r a m a
A m edical st udent who wat ched t his episode recorded his react ion in a web blog:
I wat ched Fut uram a t oday, and I t hought it was a pret t y cool episode. First off, it was probably
t he first t im e som eone had point ed out how absurd it would be [ t o] act ually m iniat urize people
in order t o ent er a body ( a la “ I nnerspace” ) . I nst ead what t hey did was creat e a bunch of
nano- droid replicas and cont rolled t hem via VR gear and m ade t he nano- droids get int o a
nano- spaceship. Second, when t hey finally did get int o t he body, it was pret t y realist ic. Well,
as realist ic as a cart oon can get . I never t hought I ’d hear a cart oon charact er ut t er t he phrase
“ pelvic splanchnic nerve” . They even got t heir hum an anat om y right , as t hey ent ered t hrough
t he ear, m aking a m icrohole t hrough t he t ym panic m em brane ( which quickly sealed up
because Fry- - t he guy t hey had ent ered- - had been infest ed wit h worm s t hat quickly healed all
his inj uries) , and apparent ly t ravelling down t he Eust achian t ube t o em erge in t he
nasopharynx. From t here, t hey went int o t he nasal cavit y, punched t hrough a capillary, and
got carried all t he way t o t he heart . They even got t he shape of eryt hrocyt es right . By
following t he circulat ion, t hey event ually m ade it t o t he st om ach, where t hey get pursued by
som e worm s ( Worm s! ) flying som e TI E fight er- like craft [ which appear in St ar Wars, BN] . The
heroes' nano- spaceship barely m akes it t hrough t he pyloric sphinct er ( t hat 's anot her t hing I 'd
never t hought would be ut t ered in a cart oon) , leaving t heir pursuers t o crash. And get t his.
You know what t heir obj ect ive was? To irrit at e t he pelvic splanchnic nerves enough so t hat t he
m ot ilit y of Fry’s bowel would increase, flushing t he worm s out . ( I n t he Professor’s words,
“ Aft er t his bowel m ovem ent , he’ll be lucky if he has any bones left .” ) . [ 33]

I n n e r Spa ce a nd N a nom e dicine

This blog refers t o I nner Space, t he 1987 film t hat had, it self, also been a parody of Fant ast ic
Voyage. This film t ook up t he narrat ive of t he nano- subm arine once again when it t old t he
st ory of a heroic sergeant who is m iniat urized and inj ect ed inside t he body of a hypochondriac
and he, t he hypochondriac and t he leading lady get involved in various m is- advent ures ( which
have no longer anyt hing t o do wit h m edical issues as such or wit h visions of nanosurgery, as
conj ured up in Fant ast ic Voyage) .
Post ers of I nner Space show a ‘Naut ilus’ t hat has sprout ed robot ic arm s and legs swim m ing
out of t he open m out h of t he hero. Bet ween 1954 and 1966, when Fant ast ic Voyage was
released, a different t ype of nano- m achine had appeared on t he fict ional horizon and m erged
wit h t he im age of t he Naut ilus. As Freit as explains in an art icle on nano- fict ions which
appeared in Nanom edicine:

M ole cu la r N a n ot e ch nology N e a r ly I nve n t e d in 1 9 4 2

The lat e science fict ion aut hor Robert A. Heinlein nearly invent ed t he concept of m olecular
nanot echnology in 1942 when he suggest ed a process for m anipulat ing m icroscopic st ruct ures.
Heinlein envisioned t he ext ensive use of life- size t eleoperat or hands, called “ waldoes” ,
com plet e wit h sensory feedback for full, rem ot e- cont rolled t elepresence. His fict ional hero,
Waldo, used a collect ion of t hese m echanical t eleoperat ed hands for building and operat ing a
series of ever- sm aller set s of such m echanical hands. The sm allest m echanical hands, “ hardly
an eight h of an inch across” , were equipped wit h m icro- surgical inst rum ent s and st ereo
“ scanners” , and were used t o “ m anipulat e living nerve t issue, [ t o exam ine] it s perform ance in
sit u” , and t o perform neurosurgery. Eric Frank Russell’s 1947 st ory “ Hobbyist ” described a
fabricat ion process wit h “ at om fed t o at om like brick aft er brick t o build a house” . I n 1955,
Russell’s serial “ Call Him Dead” , also published in Ast ounding Science Fict ion, had a virus-
based alien int elligence t hat spread t hrough cont act wit h blood or saliva; t he st ory feat ures a

Volum e 1 | Decem ber 2005 Page 13 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

“ m icroforger” , a m an who m akes “ surgical and m anipulat ory inst rum ent s so t iny t hey can be
used t o operat e on a bacillus” . Also in t he m echanical t radit ion, I saac Asim ov’s “ Fant ast ic
Voyage” in 1966 t ook it s m iniat urized hum an crew in a m iniat urized subm arine t hrough t he
bloodst ream of a hum an pat ient on a m ission of repair. [ 34]

Th e r e is Ple n t y of Room a t Th e Bot t om

This was t he ( fict ional) cont ext in which Richar d Feynm an, renowned physicist and Nobel prize
winner, was im m ersed when he wrot e his fam ous speech “ There is plent y of room at t he
bot t om ” [ 35] in which he explained t hat t here is no reason why t he m iniat urizat ion of
product ion processes cannot be cont inued down t o t he level of at om s. The explorat ion of t he
bot t om of t he ocean ( 20,000 leagues under t he sea) is here replaced by t he explorat ion of t he
bot t om of t he at om ic scale. I n his speech Feynm an “ [ t ] ells st ories about t iny writ ing, t iny
com put ers, t he act ual visualizat ion of an at om , hum an surgery accom plished by ‘swallow[ ing]
t he surgeon,’ and “ com plet ely aut om at ic fact ories” [ 8] .

Fict ion a l N a no- M a chin e s be com e Pa r t of Th e Fa ct ua l N a n o- D iscou r se

Aft er m erging t he m echanical im ages of Heinlein wit h t he vehicular ones of Verne, and
t ransposing t he t echnology from t he out side t o t he inside of t he hum an body, fict ional nano-
m achines becam e part of t he fact ual nano- discourse on t he one hand and of m odern ( sem i-
fact ual/ fict ional) nano- illust rat ions on t he ot her. These fict ional nano- m achines becam e
inst rum ent al in shaping t he nano- origin ‘m yt h’ spun around t he very incept ion of t he field and
are st ill used t o shaping it s ( scient ific, financial and polit ical) fut ure. I t is highly probable,
according t o Milburn [ 8] , t hat Feynm an him self was influenced in his t hinking by fict ional
st ories about nano- surgeons “ circulat ing in t he discourse of science fict ion long before science
‘grabbed t he idea’” . Milburn com es t o conclusion t hat : “ I f we really want t o locat e an origin t o
nanot echnology, it is not t o Feynm an t hat we m ust look, but t o science fict ion.” [ 8]

Scie n ce Fict ion Evolve s by Applyin g M in ia t u r iza t ion t o Ex ist in g I de a s

This voyage from fict ional t o fact ual discourse, facilit at ed by one fam ous subm arine, t he
Naut ilus, seem s t o follow a general t rend in science fict ion writ ing t hat Vos Post and Kroeker
have discussed in t heir art icle “ Writ ing t he Fut ure: Com put ers in Science Fict ion” . They claim ed
t hat m any of t he leaps in science fict ion were m ade by applying m iniat urizat ion t o exist ing
ideas, in our case large subm arines swim m ing in t he oceans t hat becom e m ini- subm arines
swim m ing in t he hum an body:

Scie n ce Fict ion a n d M oor e ’s La w

Generally speaking, science fict ion has adhered t o a kind of Moore’s law of it s own, wit h each
successive generat ion of writ ers at t em pt ing t o out t hink earlier generat ions’ t echnologies in
t erm s of bot h form and funct ion. Oft en, out doing earlier fict ions sim ply ent ailed im agining a
device sm aller or m ore port able or wit h great er funct ionalit y - - exact ly t he kind of
enhancem ent s at t he heart of com pet it ion in t he com put ing m arket place t oday. [ 36]

Jou r n e y of Th e N a u t ilu s Con t in ue s

And, I would add: t he nanot echnology m arket place as well. But t his is not where t he j ourney
of t he Naut ilus ends. Aft er having brought t he Naut ilus from t he eart h’s oceans t o t he veins
and art eries of t he hum an body, vehicular nanoboat s were proj ect ed int o out er space,
especially inside t he im aginat ion of t he physicist Michio Kaku. The Naut ilus as t he ‘icon of t he
spaceship’ began t o boldly t ravel where nobody, not even Nem o, had gone before: t o t he final
front iers of size and space.

Th e N a u t ilu s Tu r n s in t o A N a no- Spa ce - Ar k

I n 1998 Kaku, an int ernat ionally recognized aut horit y in t heoret ical physics and t he
environm ent , wrot e a book ent it led Visions: How Science Will Revolut ionize t he Twent y- First
Cent ury [ 37] in which he envisioned t hat soon swarm s of nano- robot s t he size of viruses would
scour our blood vessels, and t hat , wit h knowledge gleaned from t he on- going Genom e Proj ect ,
we would be able t o t weak our recalcit rant genes.

Volum e 1 | Decem ber 2005 Page 14 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

N a no- Spa ce Tr a ve l
On Novem ber 22, 2000, Chris Wallace of ABC News int erviewed Kaku. Kaku st at ed, “ I believe
t hat t he first st arship m ay be a nanoprobe, perhaps t he size of your fist , which will use
nanot echnology t o m iniat urize it s propulsion syst em s.” [ 38] This idea was furt her explored in
his book Parallel Worlds, published in 2004 [ 39] . Here Kaku goes int o m ore det ail about how
nano- space- t ravel m ight work for a new generat ion of hum ans – what one could call post -
hum ans. Merging ideas about cloning and about self- replicat ing m achines wit h spaceships,
space t ravel and space colonisat ion, he writ es:

Se lf Re plica t ion
Given t he ast ronom ical num ber of possible planet s in t he galaxy, a Type I I civilizat ion m ay t ry
a m ore realist ic approach t han convent ional rocket s and use nano t echnology t o build t iny,
self- replicat ing robot probes which can proliferat e t hrough t he galaxy in m uch t he sam e way
t hat a m icroscopic virus can self- replicat e and colonize a hum an body wit hin a week. Such a
civilizat ion m ight send t iny robot von Neum ann probes t o dist ant m oons, where t hey will
creat e large fact ories t o reproduce m illions of copies of t hem selves. Such a von Neum ann
probe need only be t he size of bread- box, using sophist icat ed nano t echnology t o m ake
at om ic- sized circuit ry and com put ers. Then t hese copies t ake off t o land on ot her dist ant
m oons and st art t he process all over again. Such probes m ay t hen wait on dist ant m oons,
wait ing for a prim it ive Type 0 civilizat ion t o m at ure int o a Type I civilizat ion, which would t hen
be int erest ing t o t hem . ( There is t he sm all but dist inct possibilit y t hat one such probe landed
on our own m oon billions of years ago by a passing space- faring civilizat ion. This, in fact , is t he
basis of t he m ovie 2001, perhaps t he m ost realist ic port rayal of cont act wit h ext ra- t errrest rial
int elligence.) [ 40]

Spe cu la t ion Abou t The Fu t u r e of N a not e chn ology

Here, as everywhere in speculat ions about t he fut ure of nanot echnology, scient ific visions m eet
science fict ional visions and popular science m eet s popular cult ure – t his t im e not only in t he
shape of popular im ages but also in t he shape of popular t unes. One can j ust hear St rauss’s
‘Also sprach Zarat hust ra’ playing in t he background. Ot her m usical backdrops t o nanoscience
st em from t he Beat les ( “ We all live in a nano- subm arine” [ 41] ) and Disneyland ( “ I t ’s a sm all
world” – used in t he t it le of an art icle by a nanoscient ist : “ I t 's a sm all, sm all, sm all, sm all
world” , [ 42] ) .

Con clu sion

During it s voyage t hrough popular im aginat ion, which began in 1869, t he ( nano- ) subm arine
has becom e one of t he m ost popular and m ost posit ive icons of nanot echnology and part of
our visual repert oire. As a visual m et aphor it has becom e part of West ern cult ure, which
m akes it especially useful in t he com m unicat ion of science. However, like all m et aphors, visual
m et aphors can highlight and hide aspect s of realit y; t hey can exaggerat e or downplay issues;
t hey can clarify or confuse argum ent s; new m et aphors can creat e new perspect ives on realit y;
old m et aphors can proj ect past prom ises and past cont roversies ont o new achievem ent s [ 43] .
The use of verbal and visual m et aphors and im ages should t herefore be m onit ored very
closely, especially at t he incept ion of a new scient ific field.
Recent ly, som e of t he m ost enchant ing and seduct ive im ages of nanosubm arines and
nanosurgeons have been produced by art ist s t rying t o im agine what a nano- fut ure m ight look
like or t rying t o illust rat e real advances in nano- science. The following im age, for exam ple,
shows art work of nanosub inside a hum an vein.

Volum e 1 | Decem ber 2005 Page 15 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

Figur e 6 . Nanobot swim m ing t hrough hum an vein.

Nanorobot . Art work of a nanorobot ( at upper left ) rem oving a blockage of plaque ( grey ) from t he wall of a hum an
blood vessel. The nanorobot is using rot ary blades t o break up t he block age and suck ing t he fragm ent s int o nozzles.
Around t he robot are disk- shaped red blood cells. At right is t he t ip of t he hypoderm ic needle used t o inj ect t he
robot int o t he blood vessel. This t ype of robot , only 0.1 m m long, is a possible applicat ion of nano- t echnology t o
m edicine. Plaque form s on t he insides of art eries t o cause at herosclerosis, rest rict ing t he flow of blood t o vit al
organs. Nanot echnology is t he science of const r uct ing m achines from m icroscopic com ponent s. JULI AN BAUM /
SCI ENCE PHOTO LI BRARY

Anot her Science Phot o, depict ing a sperm sort er, was used in t he New Scient ist t o illust rat e an
advance in m edical nanoscience achieved by Aust ralian researchers in t he year 2000. [ 44] I t
was also used in t he Germ an newspaper Bild which had so irrit at ed Professor Jam es Gim zewski
by publishing an art icle on nanobot s blast ing away cancer cells ( see above) . The phot o was
discussed on a Bild blogging sit e under t he heading: Sym bol( - phot o) of t he fut ure [ 45] .

Figur e 7 . Sperm sort er.

Medical nanorobot . Com put er art work of a m edical nanorobot holding a sperm cell. Microscopic robot t echnology
could be developed in t he fut ure t o t r eat disorders, such as infert ilit y , in new ways. This m achine has ident ified a
suit able sper m cell ( m ale reproduct ive cell) and is guiding it t owards t he egg ( fem ale reproduct ive cell, not seen)
where fer t ilisat ion will occur. VI CTOR HABBI CK VI SI ONS / SCI ENCE PHOTO LI BRARY

Just as t he Naut ilus becam e a posit ive sym bol of t he fut ure at t he end of t he 19t h cent ury, so

Volum e 1 | Decem ber 2005 Page 16 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

t he nanoboat has becom e a sym bol of t he fut ure at t he beginning of t he 21st cent ury. Like t he
Naut ilus, t he nanoboat is t here t o m ake t he invisible - an as yet ut opian fut ure - visible and, in
t he process it t urns science int o a spect acle prom ising spect acular cures for all sort s of evils.
What Buisine wrot e in 1974 about Verne’s descript ions of t he oceans applies j ust as well t o
m odern nanoboat s: “ Ce qui se donne com m e descript ion du visible n’est alors qu’une
représent at ion au sens t héât rale, une m ise en scène du discours scient ifique.” ( That which
present s it self as a descript ion of t he visible is in t he end not hing but a perform ance in t he
t heat rical sense, a t heat rical display of scient ific discourse) [ 46] . There are ot her cont inuit ies
bet ween t he Naut ilus and nanoboat s. I n bot h cases, fact and fict ion m erge. Just as Verne
proj ect ed what was t echnologically feasible at t he t im e int o a fict ional fut ure, so nano-
illust rat ors t oday proj ect what is nano- t echnologically feasible int o a fict ional fut ure. I n bot h
cases t he writ ers and illust rat ors believe ( or hope) t hat , som e day, t his fut ure will becom e t he
present . I n bot h cases, t oo, educat ion and recreat ion m erge. The reading and viewing public
are not only inform ed about science, science is present ed in t he shape of recreat ional im ages.
There is a difference however: t he im aginary Naut ilus, once it becam e real, say in t he shape of
t he U.S.S. Naut ilus, could be used in various ways, t o prot ect or t o dest roy life, but it could not
t ransform life direct ly. Nanom achines, if ever t hey becom e real, m ight well do t his [ 47] .
For Sam uel Taylor Coleridge, t he fam ous 19t h- cent ury English poet , im aginat ion was t he link
bet ween m an and t he world. I n t he case of nanobo( a) t s, im aginat ion penet rat es int o t he core
of t his link bet ween hum ans and t he world and m akes t he boundaries bet ween t he t wo fluid.
I m aginat ion feeds on t he past t o represent t he fut ure and feeds on t he fut ure t o m obilise t he
present . This art icle has t ried t o show how som e cult urally well- ent renched im ages are helping
t o const ruct and link visions of a post hum an fut ure t o t he present . Like genet ic engineering
and biot echnology, nanoscience ( as an am algam , som e m ight say, of bot h) conj ure a up
visions of bot h hope and fear and, depending on t he im ages used up by scient ist s, polit icians
and ent repreneurs, different fut ures can be const ruct ed on t he back of such im ages. Whereas
Drexler’s Engines of Creat ions have, in t he m ain, been narrat ively int egrat ed int o a discourse
of fear and dest ruct ion, t he Naut ilus, in it s various shapes and form s, has becom e an icon of
scient ific progress and, aft er m erging wit h ot her im ages, an icon of hope and healing in
m edical nanoscience.
As Hub Zwart [ 48] has recent ly point ed out wit h regard t o genom ics, previous analyses of
lit erary and visual sources have shown t hat t he public underst anding of scient ific research
t ends t o rely on a lim it ed num ber of basic, st ereot ypical im ages [ 49, 50] which t he French
philosopher Gast on Bachelard called ‘archet ypes’. They refer t o t ypical expect at ions t hat lay
audiences t end t o have vis- à- vis science, t o basic im ages t hat incit e bot h fascinat ion and
unease am ong t he public at large. The Naut ilus, once m iniat urized, links up wit h such
archet ypical im ages and wit h hum an beings’ et ernal fascinat ion wit h lit t le t hings t hat inhabit
hidden or secret worlds, from Lilliput and Tom Thum b t o Fant ast ic Voyage and Dr Seuss’s
Hort on Hears a Who ( where Hort on t he Elephant hears a cry for help from a speck of dust , and
t ries t o prot ect t he infinit esim al creat ures who live on it ) . Most im port ant ly however, it links up
wit h m yt hs and legends about int repid explorers in fant ast ic vehicles exploring what som e call
t he ‘endless front ier’ of nanoscience [ 51] .
Wit hout t he nanosubm arine ( evoking relat ively posit ive at t ribut es of nano, such as invisibilit y
and m icro- locom ot ion, but not self- replicat ion) , t he voyage of nanoscience m ight have ent ered
m uch m urkier wat ers of public im aginat ion and expect at ion. Wit h t he subm arine as an icon or
sym bol it is m uch easier t o ‘sell’ nano t o t he public ( j ust as t he use of t he ‘book of life’
m et aphor m ade selling t he hum an genom e proj ect easier, see [ 52] ) . This posit ive pot ent ial
has recent ly been exploit ed for exam ple by General Elect ric in t he Unit ed St at es. As Howard
Lovy not es on his Nanobot sit e: “ General Elect ric is working on real- life nanot echnology, but
som ebody in it s ad depart m ent knows t hat lect ures on t he com pany’s R&D in nanocom posit es
and nanost ruct ured opt oelect ronics will leave viewers running for t he fridge or t he rem ot e.
I nst ead, it chose t o t ry for t he im aginat ion, using cult ural icons and hum or” . [ 53] Using im ages
from Fant ast ic Voyage, General Elect ric invit ed t he public in 2004 t o ‘t ake a j ourney int o t he
hum an brain wit h GE’s I nst aTrak 3500 surgical navigat ion syst em from GE Healt hcare’.
Naut ilus m eet s nano m eet s neuro…… t he st ory cont inues.

Volum e 1 | Decem ber 2005 Page 17 of 19 DOI : 10.2240/ azoj ono0109

 Briget t e Nerlich

Ack now le dge m e n t

This art icle has been writ t en at t he I nst it ut e for t he St udy of Genet ics, Biorisks and Societ y,
Universit y of Not t ingham , UK, which is part ly funded by t he Leverhulm e t rust . I would like t o
t hank Robert Dingwall, Colin Milburn, Anit a and Malcolm Boshier, Cecily Palm er and Nick
Wright for t heir helpful com m ent s, and m y son Mat t hew for his inspiring suggest ions.

Re fe r e n ce s
1. St one, V., “ Developing a Safe Nanot echnology” , CaSE News ( Cam paign for Science & Engineer ing in t he UK, 44,
8, 2005.
2. Drex ler, E. K., Engines of Creat ion: The com ing era of nanot echnology. Anchor Books, New York, 1986.
3. Cricht on, M., Prey. Harper and Collins, London, 2002.
4. Jones, R., “ The Fut ure of Nanot echnology” , Physics World, August , 2004:
ht t p: / / www .physicsweb.org/ obj ect s/ new s/ 8/ 7/ 17/ j ones.pdf .
5. Jones, R., Soft Machines: Nanot echnology and Life, Oxford Univ ersit y Pr ess, Oxford, 2004.
6. Lopez, J., “ Br idging t he Gaps: Science Fict ion in nanot echnology” , Hy le – I nt er nat ional Journal for Philosophy of
Chem ist ry ( special issue on “ Nanot ech Challenges” ) , 10 ( 2) 129- 152, 2004.
7. Evans, A. B., “ The Vehicular Ut opias of Jules Verne” , Transform at ions of Ut opia: Changing Views of t he Perfect
Societ y, Shusser, G., AMS Pr ess, New York, 99- 108, 1999.
8. Milburn, C., “ Nanot echnology in t he Age of Post hum an Engineer ing: Science Fict ion as Science” , Hayles, K. N.,
ed., NanoCult ure. I m plicat ions of t he new t echnoscience, I nt ellect , London, 109- 129, 2004.
9. Milburn, C., Nanoscience in Com ic Books, in prep.
10. Hayles, K. N., ed., NanoCult ure. I m plicat ions of t he new t echnoscience, I nt ellect , London, 2004. 12
11. Moore, A., “ Of Silicone and Subm ar ines” , EMBO Report s 2 ( 5) 367–370, 2001; doi: 10.1093/ em bo-
report s/ kve104.
12. Bensaude- Vincent , B., “ Two Cult ures of Nanot echnology?” , Hyle – I nt ernat ional Journal for Philosophy of
Chem ist ry ( special issue on “ Nanot ech Challenges” ) , 10 ( 2) 65- 82, 2004.
13. Godsell, D., Liv ing Machinery. Bionanot echnology: Lessons from Nat ure, Wiley - Sm it h, New York, 2003.
14. PABE, Final Report : Public Percept ions of Agricult ural Biot echnologies in Europe, 2002. ht t p: / / www .pabe.net
15. Reit h lect ures, Lord Broers, 2005: The Trium ph of Technology, Lect ure 4: Nanot echnology and Nanoscience:
ht t p: / / www .bbc.co.uk / radio4/ reit h2005/ lect ure4.sht m l
16. de Souza e Silva, A., “ The I nvisible I m aginary: Museum spaces, hybr id realit y and nanot echnology” , Hayles, K.
N., ed., NanoCult ure. I m plicat ions of t he new t echnoscience, I nt ellect , London, 27- 46, 2004.
17. Toum ey, C., “ Narrat ives for Nanot ech: Ant icipat ing public react ions t o nanot echnology” , Techne 8 ( 2) , 88- 116,
2004.
18. Weingar t , P., “ Von Menschenzucht ern, Welt beherr schern und skrupellosen Genies – das Bild der Wissenschaft im
Spielfilm ” , Ret r ieved on 02/ 04/ 05 from : ht t p: / / www.scienceandfict ion.de/ en/ 04/ pdf/ 008Weingar t .pdf.
19. The Brav e New World of Nanot echnology, Br it ain Today, 2004. Ret riev ed on 12/ 04/ 05 from :
ht t p: / / www .uk inindia.com / m agazines/ brit aint oday/ BTI nnerpage.asp?I ssueI d= 77&m agzineI d= 3&Sect ionI d= 557
20. Robin, C., Un Monde Connu et I nconnu: Jules Ver ne, H.V.B. Edit eur - I m prim eur, Nant es, 1978. 133
21. Dewdney , A.K., “ Nanot echnology - - Wher ein Molecular Com put ers Cont r ol Tiny Circulat ory Subm arines” , Scient ific
Am er ican 258 ( January) 100- 103 ( 101) , 1988.
22. Buff, J., “ Towards Nano- Fish- Bot s?” , Milit ary .com , 12 Febr uary, 2003: Ret rieved on 12/ 04/ 05 from :
ht t p: / / www .m ilit ary.com / New Cont ent / 0,13190,Buff_021203,00.ht m l.
23. Ner lich, B., Clarke, D. D. and Dingwall, R. “ Fict ions, Fant asies, and Fears: The lit erary foundat ions of t he cloning
debat e” , Journal of Lit erary Sem ant ics, 30, 37- 52, 2001.
24. Ner lich, B., Johnson, S. and Clarke, D. D., “ The Role of Fram es, Clichés and Met aphors in t he Debat e about
Designer Babies” , Science as Cult ure, 12 ( 4) 471- 498, 2003.
25. Miksanek , T., “ Microscopic Doct ors and Molecular Black Bags: Science Fict ion’s prescript ion for nanot echnology
and m edicine” , Lit erat ure and Medicine 20 ( 1) , 55- 70, 2001.
26. Verne, J. ( 1869 onwards) . Vingt m ille lieues sous les m er s/ 20,000 leagues under t he sea. Var ious edit ions:
ht t p: / / hom e.at t .net / ~ karen.crisafulli/ 20000.ht m l; ht t p: / / j v .gilead.org.il/ fpwalt er/
27. Shult z, W., “ Nanot echnology : The next big t hing” , Chem ical and Engineering New s, 78 ( 18) 41- 77, 2000.
28. Ner lich, B., “ Voyage à t raver s l’im possible. Pièce en t rois act es de Jules Verne - Essai d’une int erprét at ion
fant ast ique” , Jules Ver ne 5, Raym ond, F., Let t r es Modernes, 117- 130, 1987.
29. Ner lich, B., Bedeut ung der I llust rat ionen für die Jules- Ver ne- Rezept ion. Schrift liche Hausarbeit im Rahm en der
Erst en St aat sprüfung für das Lehram t für die Sekundarst ufe I I vorgelegt dem St aat lichen Prüfungsam t für Erst e
St aat sprüfungen für Lehräm t er an Schulen, Düsseldorf, 1982.
30. Wolfe, G. K. The Know n and t he Unknow n: The I conography of Science Fict ion, Kent St at e UP, Kent OH, 1979.
31. Taylor, G. R., The Biological Tim e- Bom b, Tham es and Hudson, London, 1968.
32. Fut uram a, Par asit es Lost , 2001. Ret riev ed on 03/ 05/ 05 from :
ht t p: / / www .t vt om e.com / t vt om e/ servlet / GuidePageServlet / showid- 249/ epid- 15024/
33. Fat o Profugus, Hist ology and Mat t Groening, weblog, 22 January, 2001. Ret rieved on 15/ 05/ 05 from :
ht t p: / / fat oprofugus.net / j ournal/ 2001- 01- 22- 0006- 0600.ht m l
34. Freit as Jr., R. A., Nanom edicine, Volum e I : Basic Capabilit ies, Landes Bioscience, Georget own, TX, 1999:
ht t p: / / www .nanom edicine.com
35. Feynm an, R. P., “ There’s Plent y of Room at t he Bot t om ” , Engineer ing and Science 23, 22- 36, 1960.
36. Vos Post , J. and Kroeker, K. L. Writ ing t he Fut ur e: Com put ers in Science Fict ion. Com put er Magazine, January,
29- 37, 2000.

Volum e 1 | Decem ber 2005 Page 18 of 19 DOI : 10.2240/ azoj ono0109

 From Naut ilus t o Nanobo( a) t s: The Visual Const ruct ion of Nanoscience

37. Kaku, M., Visions: How science will revolut ionize t he 21 st Cent ury, Oxford Univ ersit y Pr ess, Oxford, 1998.
38. McGuire, N.K., “ Nano in t he News” , Chem ical I nnovat ion, 31 ( 11) 60–61, 2001. Ret r ieved on 20/ 05/ 05 from :
ht t p: / / pubs.acs.org/ subscr ibe/ j ournals/ ci/ 31/ i11/ ht m l/ 11inet .ht m l
39. Kaku, M., Parallel Worlds: A j ourney t hrough creat ion, higher dim ensions, and t he fut ur e of t he cosm os,
Doubleday , New York, 2004.
40. Kaku, M., “ The Physics of I nt erst ellar Travel” , 2005. Ret rieved on 15/ 04/ 05 from :
ht t p: / / www .m kaku.org/ ar t icles/ physics_of_space_t ravel.sht m l
41. Howard Lovy's NanoBot , “ We all liv e in a nano subm ar ine” , Sept em ber 4, 2004. Ret r ieved on 16/ 03/ 05 from :
ht t p: / / www .nanot ech- now .com / new s.cgi?st ory_id= 05438
42. Merkle, R. C., “ I t ’s a Sm all, Sm all, Sm all, Sm all world” , Technology Review 100 ( 2) 25- 32, 1997.
43. Lakoff, G. and Johnson, M., Met aphor s We Live By, Chicago Universit y Pr ess, Chicago, 1980.
44. Melbourne, R. N., “ Sper m Magnet t o Help I nfert ile Men” , New Scient ist , 8 January , 2005. Ret r ieved on 13/ 04/ 05
from : ht t p: / / w ww.newscient ist .com / ar t icle.ns?id= m g18524812.800
45. Bildblog: Not izen über eine große deut sche Boulevardzeit ung, January 6, 2005. Ret r ieved on 17/ 04/ 05 from :
ht t p: / / www .bildblog.de/ ?p= 415
46. Buisine, A., “ Un case lim it e de la descr ipt ion: L’énum érat ion. L’ex em ple de Vingt m ille lieux sous les m er” , La
Descr ipt ion: Nodier, Sue, Flaubert , Hugo, Verne, Zola, Alexis, Fénelon, Univer sit é de Lille I I I and Edit ions
univ ersit aires, Lille, 81- 102, 1974.
47. Milburn, C., “ Disint egrat ing t he Post biological Body” , New Lit erary Hist ory , 35, 2005 ( fort hcom ing) .
48. Zwart , H. and de Beauford, I ., “ Challenging t he St er eot ypes” , Cent r e for Societ y and Genom ics, NL, n.d.:
49. Turney , J., Frankenst ein’s Foot st eps. Science, genet ics and Popular Cult ure, Yale Universit y Press, New
Haven/ London: Yale Univer sit y Press, 1998.
50. Van Dij ck , J., I m agenat ion. Popular I m ages of Genet ics, New York Univer sit y Press, New York, 1998.
51. Glim ell, H., “ Gr and Visions and Lilliput Polit ics: St aging t he Explorat ion of t he ‘Endless Fr ont ier’” , in Bair d, D.,
Nordm ann, A. & Schum m er, J. ( eds.) , Discovering t he Nanoscale, I OS Pr ess, Am st erdam , 2004.
52. Avise, J. C., “ Evolving Genom ic Met aphors: A new look at t he language of DNA” , Science, 294 ( 5) 86- 87, 2001.
53. “ GE’s Fant ast ic Voyage” . Ret rieved on 02/ 06/ 05 from :
ht t p: / / www .boingboing.net / 2004/ 08/ 16/ ges_fant ast ic_voyage.ht m l.

Con t a ct D e t a ils

Brigit t e Nerlich
I nst it ut e for t he St udy of Genet ics, Biorisks and Societ y
Universit y of Not t ingham
Law and Social Sciences Building
West Wing, Universit y Park
Not t ingham , NG7 2RD
UK

E- m ail: Brigit t e.Nerlich@not t ingham .ac.uk

Volum e 1 | Decem ber 2005 Page 19 of 19 DOI : 10.2240/ azoj ono0109

8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

Michelin and Fives’ AddUp:

From Making Tires to Making
Metal 3D Printers

AddUp’s vice president of business development

Vincent Ferreiro discusses the story of Michelin and
Fives’ entry into metal 3D printing.

WRITTEN BY PUBLISHED READING TIME

Michael Molitch-Hou  Jan 31, 2017 ~7 mins

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 1/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

In 2015, an announcement was made signaling the entry of a new,

large player in the 3D printing industry. The Michelin Group and
industrial engineering firm Fives launched a joint venture to develop
metal 3D printing technology.
When companies as notable as Michelin and Fives, which have been
around for over 150 years, decide to enter the 3D printing industry, it
signals that important changes are afoot in additive manufacturing
(AM). By the end of 2016, the 50/50 joint venture, dubbed AddUp,
unveiled its first metal AM system at form next in Germany.
Now that the company is officially a player in the AM market,
ENGINEERING.com reached out to AddUp's vice president of
business development Vincent Ferreiro to learn more.
The Formation of AddUp
The decision by a tire maker to become a metal 3D printer
manufacturer may raise eyebrows. After all, making the leap from
producing tires to producing advanced manufacturing equipment
may seem like quite a big one. Ferreiro, who was previously in
charge of strategy and partnerships at Michelin, explained that the
company's decision to build AM systems actually came directly from
making tires.
“Michelin started to work on the metal AM at the beginning of 2000.
The idea was to be able to make new molds for tire production. We
had it in mind to make quicker, more innovative tires with greater
cost-savings by making new designs for tire molds. To produce these
molds, we strongly believed that AM was the best way to put very
new innovative tires on the market.”

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 2/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

Upon looking into AM further, Ferreiro said the company was able to
determine that it was possible to create better tires, with such
features as improved grip resistance. The next step was figuring out
the cost structure related to 3D printing tire molds, as well as
whether or not it was even feasible to mass produce tires using these
molds, leading the company to purchase some AM systems and
begin testing.
“When the company decides to put a new product line on the
market, we produce a high number of molds and metal parts,”
Ferreiro said. “To give you a rough idea, we make almost one million
metal parts per year for only one or two new product lines. Our goal
was definitely to be sure that, if we decided to move forward and
make new tires with 3D printing technology, we had to make sure
that we could produce such a high quantity of parts with high quality
and repeatability. We also wanted to see if the overall equipment
efficiency was high enough to match our technical and economic
criteria.”
Unfortunately, however, Michelin determined that the AM machines
on the market were not capable of meeting those criteria and unable
to produce parts with sufficient quality, accuracy and repeatability.
In turn, Michelin began building its own technology inhouse. The
tire giant then went onto use that technology to produce tire molds,
according to Ferreiro.

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 3/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

A 3D-printed mold for the Crossclimate tire. (Image courtesy of Michelin.)

As he explained, “Today we have a couple dozen of these machines

worldwide that produce more than one million metal parts per year.
The machines feature an automated system with automatic powder
feeding with which you are able to make exactly the same metal part
with the best accuracy and highest level of quality. Since 2015, we
have been in serial production with these systems. [With AM], we put
two new product lines on the market, the Premier product line in the
West and the Crossclimate product line—the first tires approved for
winter and summer—in Europe.”
Over the course of the last three years, Ferreiro was involved in
identifying new businesses outside of Michelin’s core tire
manufacturing business, and 3D printing was determined to be of
potentially very high value. Having designed and made a metal 3D
printer inhouse, the company sought partners to commercialize it,
Ferreiro said. Fives proved to be ideal, given its 166 years of
experience in industrial engineering.
The FormUp 350

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 4/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

The first 3D printer to be released by AddUp is the FormUp 350, a

powder bed fusion system described to have great quality control for
industrial applications. With one or two 500W Yb fiber lasers, the
machine has a build volume of 350 x 350 x 350 mm and can produce
layer thicknesses as fine as 20 μm on maraging steel. Other materials
currently compatible with the FormUp 350 are stainless steels, nickel
alloys, Inconel 718, titanium alloys and aluminum alloys.

The FormUp 350 3D printer from AddUp. (Image courtesy of AddUp.)

According to the company, the FormUp 350 is designed to be capable

of 24/7 operation and includes “unrivaled” surface quality, material
integrity and internal stress levels and the ability to 3D print objects
with very low angles relative to unsupported material, which
minimizes the need for support structures and post-processing.
Ferreiro pointed out that, for an added cost, customers can purchase
advanced quality control mechanisms that include thermal
monitoring, but he believes that the machines themselves are so
capable of producing quality parts, that such mechanisms aren’t
always necessary.

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 5/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

“Thanks to Michelin and Fives' industrial mindset and vision, we

were able to develop a new process that assures that 100 percent of
the final product will perform within the desired technological and
economic specifications,” Ferreiro said. “That means that we
understand the relationship between the machine and its
components, between all of the process conditions, and the raw
materials. And we developed the software process to ensure that all
of the parts meet exactly the same specifications in terms of
mechanical properties, roughness and a high-quality metallurgical
result.”
The company also provides significant training and consulting
services. In addition to a basic introduction to AM and its benefits,
AddUp trains customers in the use and maintenance of its machines.
More uniquely, the company will work with customers to
conceptualize, design or redesign parts for AM. Customers who
purchase the machine as a means of manufacturing goods for their
own clients can turn to AddUp to help in meeting the manufacturing
and economic requirements of a job. AddUp even helps its customers
develop a roadmap for the technology, beginning with designs and
creating a proof of concept to leading them into preserial
production.
The Future of AM for AddUp
As AddUp continues to grow, it will continue to develop its 3D
printing technology. With a program dubbed SOFIA (“SOlutions pour
la Fabrication Industrielle Additive métallique” or "Solutions for
Industrial Metal Additive Manufacturing”), AddUp has planned a six-
year research endeavor to develop the entire metal AM value chain,
including powders, machines and processes.

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 6/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

With a number of partners—Aubert & Duval, ESI Group, FUSIA,

Michelin, Safran, VOLUM-E, Zodiac Aerospace, the French National
Center for Scientific Research and a number of universities—AddUp
will develop new technological modules that can be integrated into
AM systems beginning in 2018. SOFIA will have a specific focus on
the aerospace industry and will address 1) improving metal powder
ranges, the productivity of the machines through material/process
pairs and new energy sources, 2) designing new optimized parts and
3) increasing knowledge related to health and safety risks in metal
AM.
While improving its own technology, AddUp will also be enabling
Michelin to improve its ability to create new tire designs. Ferreiro
explained, “Today, Michelin is working on trying to make more
complex molds because we are now able to make leaps in new design
that we were not able to make before AM. We now have experts who
can understand the relationship between new complex mold designs
and the direct development of tires. We have already started to put
new passenger car tires on the market; now, our mission is to expand
this application to trucks or two-wheeled vehicles, for example, and
to explore completely new designs because the land of opportunity
with AM is huge. So we have all of our experts trying to create these
new designs.”
For drivers out there, that may mean better tires, but for the 3D
printing industry, it could mean a potentially powerful player.
Because Michelin has already implemented the technology to
manufacture millions of metal parts, the quality of the machines
may be just what the industry is looking for.
To learn more about AddUp, visit the company website.

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 7/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 8/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

WRITTEN BY SEE MORE LIKE THIS

Michael Molitch-Hou  Follow Five Stories Indicating Auto 3D Printing Is

Kicking into High Gear
Michael Molitch-Hou is the Editor of
ENGINEERING.com's 3D printing section.
Everything That Was Announced at
formnext

Meet Michelin’s 3D-Printed Concept Tire

Engineering High-Performance Winter

Tires

RELATED TOPICS THIS WEEK IN MACHINERY: Grinding,

AEROSPACE AND DEFENSE Plasma, Lasers & More

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 9/10
8/20/23, 5:52 PM Michelin and Fives’ AddUp: From Making Tires to Making Metal 3D Printers | Engineering.com

https://www.engineering.com/story/michelin-and-fives-addup-from-making-tires-to-making-metal-3d-printers 10/10
 European Journal of Innovation Management
Digital-physical product development: a qualitative analysis
Stine Hendler,
Article information:
To cite this document:
Stine Hendler, (2018) "Digital-physical product development: a qualitative analysis", European Journal
of Innovation Management, https://doi.org/10.1108/EJIM-01-2018-0026
Permanent link to this document:
https://doi.org/10.1108/EJIM-01-2018-0026
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

Downloaded on: 19 September 2018, At: 22:06 (PT)

References: this document contains references to 39 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 8 times since 2018*
Users who downloaded this article also downloaded:
,"Exploring the digital innovation process: The role of functionality for the adoption of innovation
management software by innovation managers", European Journal of Innovation Management,
Vol. 0 Iss 0 pp. - <a href="https://doi.org/10.1108/EJIM-02-2018-0051">https://doi.org/10.1108/
EJIM-02-2018-0051</a>

Access to this document was granted through an Emerald subscription provided by emerald-
srm:380143 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.

*Related content and download information correct at time of download.

 The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1460-1060.htm

Digital-
Digital-physical product physical
development: a qualitative analysis product
development
Stine Hendler
Center for Industrial Production, Aalborg University, Aalborg, Denmark

Abstract Received 31 January 2018

Purpose – The purpose of this paper is to investigate how digital and physical product development can be Revised 21 June 2018
successfully coordinated and which new product development and contextual practices are suitable for the 1 August 2018
Accepted 16 August 2018
combined digital-physical product development process.
Design/methodology/approach – The paper is based on a multiple-case study within one company with
three digital-physical product development projects as the units of analysis. The data collection and analysis
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

are guided by an existing research model. The case study is used deductively to illustrate the model.
Findings – When combining digital and physical development processes, one or both need to change. This
may lead to sub-optimization of one or both of the processes but optimizes the combined digital-physical
process. Various development and coordination practices as well as contextual measures must be put into
place to improve fit to the digital-physical process characteristics and mixed materiality.
Research limitations/implications – The paper illustrates the research model with case evidence and
suggests tentative theory in the form of propositions. Further research needs to explore the impact of the
practices and contextual measures proposed.
Practical implications – This research proposes a range of conditions facilitating the successful
development of digital-physical products.
Originality/value – This paper is among the first to empirically explore the complex process of digital-physical
product development. Taking a process perspective and focusing on organizational and managerial practices and
the influence of context, organization theory is used as the theoretical lens.
Keywords New product development, Agile, Digital innovation, Digital-physical, Digitized products,
Smart products
Paper type Research paper

1. Introduction
In today’s fierce technology-driven competition with Internet of Things and Services,
Servitization and Smart Manufacturing/Industry 4.0, companies are increasingly engaging
with digital-physical product development by adding digital technology to previous
non-digital products (Yoo et al., 2010; Porter and Heppelmann, 2015; Ardito et al., 2017).
Digital-physical product development is the process of transforming ideas into commercial
products, which include both a software and a tangible component (Hendler and Boer,
forthcoming). A concrete example is the world’s first smart shoe from Digitsole®, which
automatically tightens, warms up or cools down your feet and connects to a mobile
application with sensor information on, e.g., distance traveled and temperature.
Digital-physical product development combines traditional product development
practices with software development practices, which are significantly different from
each other (e.g. Broy, 2005; Woodward and Mosterman, 2007; Svahn and Henfridsson, 2012;
Porter and Heppelmann, 2015; Lwakatare et al., 2016). Software development is optimized to
adapt to high degrees of uncertainty in product requirements and solution methods via agile
development methods. These agile methods enable fast and frequent feedback on
the developing product and the ability to adapt accordingly in multiple, short, iterative
development cycles of typically two weeks (Cohn, 2010). Once sufficient customer value has
been developed, the software is released and subsequently improved (Cohn, 2010). Due to
digital immateriality, i.e. no manipulation of tangible materials is needed, fixed costs are
limited, there is no cost associated with the number of units produced, no manufacturing, no European Journal of Innovation
Management
transportation time and no need to focus on the reuse of physical assets or scarce resources © Emerald Publishing Limited
1460-1060
such as manufacturing equipment and shelf space. Thus, immateriality enables reduced DOI 10.1108/EJIM-01-2018-0026
 EJIM cycle cost and time due to re-programmability (Yoo et al., 2010). Immaterial outputs enable
late binding of many design-decisions as no manipulation of tangible materials is needed
(Yoo et al., 2010) and can be delivered in small increments, which reduces the risk of product
market failure.
Physical product development, in contrast, is optimized for stable exploitation of
investments. It assumes high predictability of process outcomes and market demands. The
typical high cost of physical manufacturing processes requires that existing assets are
reused. Practices supporting that include the use of product platforms and incremental
product development (Svahn and Henfridsson, 2012). Lead times are long with extensive
up-front preparation and specifications are locked early (early binding) to reduce
uncertainty (Svahn and Henfridsson, 2012). The development process typically involves a
plan-driven approach with clear phases, such as a stage-gate process (Cooper, 1990).
The differences between the digital and physical development processes raise questions
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

such as: how to effectively coordinate the need for early specification with the need to keep
options open until late in the process, and how to effectively coordinate a focus on efficient
reuse with a focus on learning? Another set of challenges concerns the context supporting
digital-physical product development. There are significant differences between
manufacturing and software companies, such as organization design and HR policies
(Porter and Heppelmann, 2015). Can these contextual characteristics be combined effectively
to support an integrated hardware-software development process?
Based on a systematic literature review of digital-physical product development, Hendler
and Boer (forthcoming) conclude that little is known about the possibilities to combine the
two processes effectively. Based on their proposed research model (Figure 1), the objective
of the present paper is:
To identify effective development, coordination and contextual practices supporting the combined
digital-physical development process.
Although digital-physical product development is not a new phenomenon in practice, it is an
immature field of research within new product development theory, with significant
industrial relevance. This field is by some referred to as digital innovation (e.g. Yoo et al.,
2012; Nambisan et al., 2017; Holmström, 2018). Its immaturity calls for theory building

Physical product
development process
characteristics

Development
context

(Dissimilar) Digital-physical
development development
practices performance

Coordination
practices

Figure 1.
Research model of Software development
digital-physical process
product development characteristics
 (Nambisan et al., 2017; Holmström, 2018) through descriptive research aimed at developing Digital-
statements of associations in the form of models (Christensen, 2006) and the propositions or physical
hypotheses embedded in these models. There is ample theory in adjacent fields, including product
physical, digital and service development and innovation theory, which is used to inspire the
present research. As the research focuses on organizational and managerial practices from a development
process perspective and the influence of context, organization theory is used as a lens
through which the phenomenon is studied. In-depth case study evidence is used to provide
empirical support for the research model (Figure 1) in the form of a range of propositions.
Furthermore, practical implications are inferred, which inform innovation managers of key
areas that require attention when combining digital and physical product development.
Section 2 summarizes the theoretical background and the research model. Section 3
presents the case study method. Section 4 presents the results and develops propositions.
Section 5 discusses the results. Section 6 concludes the paper with a summary of its
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

contribution, a discussion of the limitations and implications for further research.

2. Theoretical background
According to a literature review by Hendler and Boer (forthcoming), theory on
digital-physical product development is emerging, scattered, descriptive and rarely
considers the full complexity of the phenomenon. The literatures on software and physical
product development have been developed relatively independently (Karlsson and Lovén,
2005; Svahn and Henfridsson, 2012). Literature on embedded software development
focusses primarily on technical aspects (Haghighatkhah et al., 2017). Research on topics that
may include digital-physical product development, such as complex product, product-
service system, mechatronics and cyber-physical system development does uncover
relevant themes such as design for product evolvement after launch, multi-disciplinarity,
systems engineering and life-cycle costs (e.g. Baines et al., 2007; Wolfenstetter et al., 2016;
Bialasiewicz, 2017; Maleki et al., 2017). However, theory on how to organize and manage the
process is underdeveloped. The specific gaps relating to the objective guiding this research
are explicated below.

2.1 Differences between digital and physical development

The literature indicates that the digital-physical product development process involves two
separate development subprocesses with different development methods (e.g. Broy, 2005;
Woodward and Mosterman, 2007; Svahn and Henfridsson, 2012; Porter and Heppelmann,
2015; Lwakatare et al., 2016). According to Boer and During (2001, p. 86), “[t]he success of an
innovation depends on the extent to which the “innovation manager” is able to fit the
organization of the process to the demands created by [its] characteristics.” Uncertainty
refers to the extent to which people are informed about the future (Boer and During, 2001)
and may concern goals, methods, people and the influence from an organization’s context
(Boer, 1991). Diversity refers to the variety of the work that needs to be done in terms of the
number of competences needed to perform the innovation process. Interdependence is
defined as the extent to which (groups of ) people depend on one another for their output
(Boer and During, 2001). Complexity refers to the difficulty with which the work can be
understood and has also been referred to as, e.g., comprehensibility and analysability
(Boer and During, 2001).
Hendler and Boer (forthcoming) find that digital and physical product development have
different characteristics and are supported by significantly different practices. The software
development process needs to make binding decisions late to exploit emergent options and
utilizes agile methods that facilitate effective learning. The physical product development
process needs to make binding decisions early, resulting in early maturation of the concept
and specifications while often guided by a linear, staged and gated process with high cost of
 EJIM change later in the process. Finally, the two processes use different vocabularies and
management techniques, such as detailed end-to-end project planning (physical) vs
prioritized product backlogs (digital).

2.2 Alignment and coordination

Part of the literature concerning digital-physical product development focuses on the
differences between the two subprocesses and observes challenges when combining them
that may result in performance detriments (Hendler and Boer, forthcoming). Some describe
examples or suggestions of how to coordinate the two processes, i.e. using various
coordination mechanisms from organization theory such as integration milestones or
implementing fully cross-functional teams (Könnölä et al., 2016; Eklund and Berger, 2017),
however, without fully exploring their effects. Others provide examples or suggestions of
how digital and physical product development practices are or could be changed and
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

compromised to reduce the differences between them to create successful coordination

(e.g. Joglekar and Rosenthal, 2003; Evans, 2009; Eklund and Bosch, 2012). Again, the
performance effects are not well explored. To reduce combination challenges digital
development can adapt to physical, physical can adapt to digital or both can adapt to each
other (Hendler and Boer, forthcoming).
2.2.1 Adapting digital to physical development. Some authors observe that the long-term,
plan-driven, staged and gated approach typically used in physical product development is
used as the primary coordination mechanism combining all the needed development
activities (Cordeiro et al., 2007; Eklund and Bosch, 2012; Lwakatare et al., 2016). This
requires software development to adhere to an early and extensive planning phase with
early binding, prepare the required stage-gate related documentation and adapt to
governance structures and role descriptions ( Joglekar and Rosenthal, 2003; Karlström and
Runeson, 2006; Eklund et al., 2014). Adapting software development to the physical
development methods may, however, result in reduced product performance due to software
development becoming slower (Eklund and Berger, 2017) and less learning and adaptability
focused (Diegel et al., 2008; Evans, 2009; Eklund and Bosch, 2012).
Other authors suggest software development to start up after the mechanical concept
has been locked. This reduces the risk of the software becoming “old” before release (Eklund
and Bosch, 2012) but limits the ability to optimize and co-create the full product concept
(Diegel et al., 2008; Evans, 2009; Eklund et al., 2014).
2.2.2 Adapting physical to digital development. Instead of reducing the differences
between the two development processes by adapting software development to a typical
staged and gated process, several authors propose making the physical product development
process more agile. This would allow more effective coordination and enable the physical
product to be adapted to new information more easily (Huang et al., 2012). It could also enable
a process in which software design decisions drive physical product design decisions, which
supports the notion that the competitive advantage of digital-physical products increasingly
comes from software (Evans, 2009). While some authors report successful use of agile
development methods for both digital and physical product development (e.g. Huang et al.,
2012; Cooper, 2016), other authors report challenges such as team composition (Eklund and
Berger, 2017), management of non-interchangeable competences and a need for different
coordination frequencies due to different development cycle lengths (Könnölä et al., 2016).
Other practices suggested for aligning physical product development practices with
digital development involve accepting a gradual growth of requirements (Karlström and
Runeson, 2006), accepting incomplete components for the digital-physical prototypes
(Eklund and Berger, 2017), and using platform components to help speed up physical
development cycles (Eklund and Berger, 2017).
 2.2.3 Alignment and coordination practices are not well understood. Most of the literature Digital-
reports examples of the digital-physical product development involving two separate physical
development processes, with different characteristics and development practices, and with
the digital process adapting to the physical process. Other sources propose that the physical
product
process is adapted to the digital process. Some authors focus on aligning development development
practices (e.g. agility), others on organizing (teams) or managing (e.g. platforms) the
digital-physical process, yet others on coordination of the two subprocesses. However, how
to effectively coordinate the largely dissimilar and sometimes conflicting practices of digital
and physical product development processes is not yet well understood in literature
(Hendler and Boer, forthcoming).

2.3 New practices and contextual changes that fit the digital-physical process characteristics
Another part of the literature focuses on introducing new development practices or capabilities
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

and creating a development context that suits the new digital-physical development process.
2.3.1 Suitable development practices. Development practices that could be suitable for the
characteristics of digital-physical product development include horizontal and
combinatorial innovation (Yoo et al., 2012), distributed innovation (Yoo et al., 2010),
system interoperability (Porter and Heppelmann, 2015), rethinking existing product
architectures (Yoo et al., 2010; Lee and Berente, 2012), designing for continued product
enhancement after launch (Broy, 2005; Porter and Heppelmann, 2015; Eklund and Berger,
2017) and designing for security and big data (Porter and Heppelmann, 2015). However, little
systematic research has been reported aimed at understanding, explaining and developing
practical theory on, the specific practices needed to effectively accommodate the process
characteristics of digital-physical product development.
2.3.2 The wider development context. More operational knowledge is needed on how
companies can effectively accommodate digital-physical product development with a suitable
context. Porter and Heppelmann (2015) suggest that manufacturing industries should learn
from characteristics and practices from the software industry, such as culture, structure,
strategy, HR policies and business processes outside the product development process.

2.4 Research model

Hendler and Boer (forthcoming) propose a research model, reflecting the notions presented
above. Figure 1 depicts the part of the model that is relevant for the purposes of this paper.
The model includes four key constructs:
(1) Development practices, which are (partly) dissimilar due to the different
characteristics (uncertainty, interdependence, diversity and complexity) of the
digital and the physical subprocesses. Aligning these practices, i.e. adapting them to
each other, may affect development performance directly.
(2) Coordination practices, which are considered to moderate the relationship between
the digital-physical development process and its performance.
(3) The context in which the digital-physical development process takes place may also
moderate the relationship between the development process and its performance.
(4) Development performance, measured in process criteria such as process cost and
lead-time, and product criteria such as product quality, cost and performance
(e.g. Kuwashima and Fujimoto, 2013).
Using this research model as an analytical framework, this paper aims to reduce the
identified gaps in literature by elaborating on the constructs and relationships proposed by
the model.
 EJIM 3. Method
The present research is based on a multiple-case study design with three digital-physical
product development projects as the units of analysis. This method enables exploring and
retaining the holistic and meaningful characteristics of a complex and contemporary
social phenomenon embedded in its context (Yin, 2009). While providing empirical
evidence for investigating the research objective, the case study is used deductively to
illustrate the research model (Figure 1) by giving concrete examples and illustrating the
nature and scope of the conceptual relationships. This is a less common but valid use of
case studies for conceptual theory building (e.g. Wacker, 1998; Siggelkow, 2007; see
Caniato et al., forthcoming).
The case company (hereafter labeled COMP) is a highly successful global company, which
develops, manufactures and markets consumer products for educational purposes and
entertainment. A leading company in its markets, the company was selected as it is involved
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

in several digital-physical product development projects to which the researcher has daily
access including internal communication and documents. Its elaborate and mature core
product development process includes more than 100 milestones across a front-end and an
execution phase. The process is orchestrated by a project team supported by a large number
of functional departments working on many concurrent projects, to deliver high levels of
product quality, without delays, produced in a stable and lean manufacturing system.
Exploiting the opportunity to collect robust evidence for analytical generalization, all
three digital-physical product development projects active at the time of data collection,
hereafter named Project A, B and C, were selected for study. These projects were the first to
deliver on the company strategy of digitally augmenting its manufactured products.
All projects involved extensive collaboration with world-class external software partners
using state-of-the-art technology and agile development practices. The resulting products
include physical and software components, which are integrated in use by the user via a
third-party device, for example a mobile phone using combination technology such as visual
recognition. Project A had one product launch cycle, while Projects B and C launched
products over several years.
Data were collected from November 2014 to November 2017 through 50 interviews with
key functional areas from each project, two hours of workshop observation, 19 documents,
notes from informal conversations and 110 h of participation in process design-related
project tasks, which all enabled triangulation. Table I presents a detailed overview of the
collected data per project. Most interviews were carried out during the unfolding projects
and were semi-structured with open-ended questions to allow for exploration of emerging
topics and hypothesizing about cause-effect relationships. Each interview lasted one hour
on average and started with a collection of background information including the
interviewee’s project role and work experience. Ten of the interviewees were interviewed
twice to follow-up on key topics. The interviews were guided by the questions:
(1) Which challenges/opportunities have you experienced when combining the digital
and the physical development processes?
(2) What were the actual mechanisms encountered?
(3) How did you cope with challenges/exploit opportunities, if you did?
(4) What key knowledge was gained for next time?
Before the end of each interview, the key summary statements were presented back to the
interviewee to correct for misunderstandings. The interviews were recorded and
transcribed. As illustrated in Figure 2, the collected data were condensed and ordered
into a database of statements from which data were extracted and summarized into two
tables per project, one focusing on the dissimilarities between digital and physical
 Project A Project B Project C
Digital-
physical
Role of Product designer A (1) (1 h) Technical project manager Digital project manager A (1) product
interviewee Product designer B (1) (1 h) A (2) (1 h) (1,5 h)
in project Digital project manager A (2) Technical project manager Digital project manager B (1) development
(number of (1 h) B (1) (1 h) (1 h)
interviews) Project manager B (1) (1 h) Line manager (1) (1 h) Digital project manager C (2)
(duration of Marketing A (2) (1,5 h + 0,5 h) Marketing (2) (1+1,5 h) (1 h)
each Marketing B (1) (1 h) Project management (1) (1 h) Project manager A (1) (1 h)
interview) Internal digital producer A Digital marketing (1) (1 h) Project manager B (1) (0,5 h)
(2) (1 h) Digital producer (2) (1 + 0,5 h)
Project manager C (1) (2 h)
Internal digital producer B Designer A (1) (1 h) Digital line manager (2) (1 h)
(1) (1,5 h) Designer B (1) (1 h) Digital designer A (2) (1 h)
Line manager A (1) (1 h) External digital producer A Digital designer B (1) (1 h)
Line manager B (1) (1 h) (1) (2 h) Digital producer A (1) (1 h)
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

Consumer insights (1) (1 h) External digital producer B Digital producer B (2) (1 h)

External digital producer (1) (1) (0,5 h) Digital producer C (1) (1 h)
(2 h) Designer A (1) (1 h)
Designer B (1) (1 h)
Designer C (1) (1 h)
Marketing (1) (1 h)
External digital producer
(1) (1 h)
Other Archival data (escalation Archival data (process Archival data (“lessons
collected data report, “lessons learned” suggestions from a product learned” report for full project,
report for full project, “lessons designer, project observations “lessons learned” report for
learned” report for from a technical project marketing, project
collaboration between COMP manager, “lessons learned” organization charts, product
and the digital vendor, project report for partnership concept diagrams, project plan)
plan) capability building, project Observation of a 2 h “lessons
Approx. ten informal organization and role chart, learned” workshop
conversations with project project schedule) Approx. ten informal
management and digital Approx. five informal conversations with project
project members conversations with digital management and digital
Notes from facilitating an project participants project members
end-to-end process design for Notes from participating
digital-physical product in the design of parts of a
development with project digital-physical front-end
management (approx. 100 h) process early in the project Table I.
together with project Overview of the
management (10 h) collected data

development and the alignment and coordination practices needed to overcome these
differences, the other on new practices including the contextual changes needed to
accommodate the new characteristics and materiality of the combined digital-physical
product development process. Next, the data were analyzed through the lens of the research
model by grouping paper slip representations of the table rows, first within and, then, across
the individual projects. Finally, the paper slips were grouped to extract propositions. To
avoid misinterpretation, continuing informal conversations with project members allowed
quick feedback during the data analysis.

4. Results
4.1 The digital and physical product development process characteristics
The research is based on the fundamental assumption that digital and physical development
have different characteristics. The aim of this subsection is to verify this assumption.
 EJIM Grouping of rows/paper slips
to develop propositions
Alignment and coordination practices

Columns: Proposition

Interviews
• Statement ID
• Conflicting process practices
Database of statements
• Example
Columns: • Coordination practice to create fit
Proposition
• Local performance effect
• Statement ID
• Overall performance effect
• Source
• Further solution suggestions
• Project
Documents

• Change/challenge/
opportunity New development and contextual
practices
• Example Proposition
• Performance impact Columns:
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

•Tested/suggested • Statement ID
• New process practices needed
• Conflict with existing practice
Conversation

Proposition
• Example
notes

Figure 2. • Performance impact

Analytical coding • Tested practice to create fit
process

Projects A, B and C all involved a collaboration with digital partners with largely similar
agile development methods. Relative to COMP’s physical product development process, their
digital product development processes are adapted to a higher degree of uncertainty, a lower
degree of diversity and similarly high degrees of interdependence and complexity. The high
degree of uncertainty stems from a rapidly evolving digital technology, industry structure,
competitor landscape and consumer preferences. One of the digital partners explained: “Our
R&D team continuously looks into improving our technology, which needs to be leading
edge to be competitive. If we were to lock months before release [like COMP does], we would
be [several] months behind.” The diversity is considerably lower for digital development with
3–4 competence areas represented in the digital cross-functional teams, whereas approx.
50 distinct competences are involved in the physical product concept design.
The digital and physical development practices can to a large extent be explained by
these process characteristics and materiality. For the digital partners, digital immateriality
enables short and low-cost development cycles, several releases per product and late
binding which, coupled with high degrees of complexity and uncertainty, goes a long way to
explain the iterative and emergent agile development process with short up-front planning
and the possibility to quickly adapt to new information via empowered decision-making,
cross-functional teams, a floating scope (product requirements) and highly formalized two-
week sprints. The lower degree of diversity enables the use of smaller cross-functional
teams, easier mutual adjustment and the use of one or a few product vision holders as there
is no need to re-interpret the vision across multiple knowledge domains, all of which enable
fast adaptation to maximize consumer value.
In comparison to digital development, COMP’s product development process is
characterized by higher diversity, lower uncertainty and physical materiality, resulting in
one long development cycle with a limited number of product spirals, higher cycle costs and
early binding due to logistics and manufacturing. Consequently, COMP’s development process
is designed to launch products reliably in accordance with extensive up-front planning via a
highly formalized, high-level schedule-bound plan with hierarchical decision making, which
leaves little room for exploration and adaptation in the latter part of the process. Unforeseen
changes are primarily managed by adjusting project staffing and/or work load. The focus on
 consumer value is balanced with schedule, cost and risk considerations. High degrees of Digital-
diversity and complexity result in multiple product vision holders to ensure translation across physical
domains, consensus decision making and coupled with low market uncertainty, a largely product
functionally-oriented organization. The digital development requires approx. one year whereas
the physical process is approx. twice as long. Thus, optimal execution of digital-physical development
product development requires two different ways of organizing them:
P1. Compared to physical product development, the digital development process is
characterized by a higher degree of uncertainty, a lower degree of diversity and
digital immateriality, which allows for an adaptability-optimized development
process of short, iterative development cycles with predominantly cross-functional,
teams, empowered decision-making, one product vision holder, a floating scope,
short development cycles, late binding, short up-front planning and several releases
per product.
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

P2. Compared to digital development, the physical product development process is

characterized by a lower degree of uncertainty, a higher degree of diversity and a
physical materiality, which allows for a stability-optimized development process of
one long development cycle with a large extent of functional unit grouping,
hierarchical and consensus driven decision-making, multiple product vision holders,
a highly formalized and high-level, schedule-bound development process, early
binding, extensive up-front planning and one launch per product.

4.2 Combining digital and physical development through alignment

This subsection aims at identifying attempts made to align the digital and physical
subprocesses by adapting one to the other or both to each other.
All three projects engaged with external software companies after concept lock, which
is later than the physical development teams had desired in hindsight. Reasons included a
late realization of the positive impact from collaborating in the front-end, difficult contract
negotiations and shorter digital development lead-time. Nevertheless, the late entry left
enough time for the development of digital experiences. At the time of entry, physical
development was getting ready to start handing over specifications to manufacturing and
marketing material development, which meant that the digital partners had to adapt
quickly to the physical development process due to many design interdependencies.
Partner C explained, “We had to adapt, and we did this time around, but with much
difficulty, […] extra cost and lower product quality.” The physical development process
also had to allow a large amount of exceptions such as schedule delays caused by the
digital product immaturity.
Learning from experience, Projects B and C managed to engage digital development
earlier and earlier in the subsequent front-end cycles, particularly to mitigate a large amount
of exceptions to the process schedule. Both projects successfully achieved their targets.
Project A experienced problems with poor quality in the digital-physical interface of one of
its products but otherwise met its targets, too.
The digital processes were the ones adapting the most by maturing more parts of the
digital experience much earlier to be able to accommodate the early binding of
interdependent design-decisions and deliverables in the first part of the development
process. These interdependent decisions included the packaging graphics and the bill of
materials (BoM). Locking the packaging graphics design required the digital partners to
supply screenshots from the digital experience long before they were ready. Partner A
explained how they had to “fake” a screenshot and feature list based on what they predicted
could be a great experience for the consumer, costing extra unplanned resources at the time.
 EJIM One project member stated, “The project had to stick with a less fun [software] experience,
because we had already put it on the box preventing us from developing a more fun
[experience].” Learning from this, Project C settled on preparing thematically similar
imagery without screenshots, which was sufficient to indicate the core idea of the digital
experience and resulting in a relatively less reduced digital solution space. The quality of the
resulting imagery was not considered to be as high as desired.
The digital partners also felt confined due to the user instructions and other marketing
materials that had to be locked early, such as a video demonstrating the product use, which
had to be shipped to the shops on relevant devices in good time before launch. New videos
had to be made after launch to reflect the final digital experience:
P3. With high interdependency between digital and physical product development,
imposing the early binding typical of physical development is limiting the
subsequent exploitation of new knowledge in digital development, hence, reducing
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

the digital adaptability and the potential value of the digital-physical product.
The many (inter)dependencies between digital and physical development took all three
projects by surprise in their first launch cycle, causing significant disruptions. The software
partners experienced how physical development perceived software change requests to
have very low costs. Partner C explained how these many changes took away time that
could have been used better. Partner A also explained how the project failed to create
transparent task management, resulting in COMP not being aware of which other tasks had
to be de-selected in favor of new tasks. The digital partners all compensated by adding
resources and reprioritizing work. However, the first product launch of Projects A and B
suffered from several quality problems. The software was updated after launch. However,
some problems, such as a large file size, could not be fixed easily after launch.
All projects experienced positive effects from digital mitigating physical errors. After a
launch from Project B, when approx. 40,000 products had been shipped to the shops, an
error was detected in an electronic component that was detrimental to the functioning of the
product. Fortunately, the software partner could update the software within a few days and
mitigate the physical problem. The physical marketing manager explained, “We saved huge
costs and a damaged brand by avoiding calling back the product. Fantastic!”
In all three projects, the physical design constraints dominated the overall design
decisions. Project B, for example, had to build physical products that were to be replicated
digitally and enhanced with various digital functionalities. After having passed the design
back and forth between digital and physical, the project team decided to have the physical
designers develop two-three options per product, leaving plenty of scope for creating
enough digital variety (the main digital constraint). Running out of time for the BoM, some
suboptimal physical design-decisions were made. Again, the software designers
compensated for the early binding by manipulating the objects in the digital realm.
The reduced digital solution space from the early binding and the physical design
constraints did not result in unacceptable product quality due to an adaptability-optimized
software development process and the intangible immateriality, leaving many design
options open late in the project:
P4. Digital development can adapt easier to the requirements of the physical
development process as well as absorb some of its undesired variability, without
causing a significant quality decrease of the product.
With P4 in mind, partner B successfully sharpened its agile processes by becoming more
cross-functional to increase its ability to react to unforeseen changes from the many
interdependencies between digital and physical. Physical product development also
supported the digital process adaptability by giving fast feedback for the digital
 development sprints. According to Partner C, COMP gave faster feedback than their regular Digital-
clients, i.e. within 24 h, which enabled the partner to incorporate this into the next sprint physical
and, thus, adapt fast: product
P5. Supporting and even accommodating more agility in digital development (if not development
already fully adaptability-optimized) helps software development to adapt to the
increased uncertainty from combining with physical product development.
Digital development continuously struggled to deliver the right quality at the right time for
the early milestones. Accordingly, all three projects experienced delays and changes to the
BoM, which challenged multiple dependencies between product, marketing and
manufacturing development. The pre-planned allocation of people’s time made it difficult
to re-schedule the many concurrent development projects worked upon by the largely
functional organization of COMP. This caused the projects to spend a lot of time on
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

stakeholder management, filling in forms, waiting for hierarchical decision making, while
causing stress and overtime. A project manager from Project C described, “We do need a
standard [product development] process, but we need more flexibility to customize it.
Today, it is very cumbersome and resource demanding to argue sufficiently to be allowed to
deviate from the plans.” The delays also caused budget overruns. For example, Project C
failed to deliver a digital file on time due to bugs, which incurred extra cost for an external
marketing production company. The delays also increased the risk of errors, as tasks had to
be “hand-carried” through otherwise standardized subprocesses. Due to the overall highly
complicated development process, this brought with it the risk of neglecting standardized
information flows or quality assurance processes. In other cases, the delayed digital
deliveries meant that design decisions had to be made without input from the digital
partner. The number of exceptions related to the changes to the budget, the schedule and the
BoM far exceeded the average project and resulted in much time spent on exception
management. One of the project managers admitted, “We should have categorized this
project as high risk and high complexity from the start.”
Project B stood out, as it adapted to radically new concept ideas from their digital
partner after the concept lock milestone, by successfully hacking the established
development process. One project member stated, “We forced this project into a standard
process that did not fit. Once we realized this, the project started to turn for the better.”
The project learned to change and fit the plan to new circumstances, negotiate new
timelines and resources with stakeholders and speed up development work via, e.g., more
digital-physical co-location. According to one of the leading project team members, the
team pulled off one of the biggest stunts in recent COMP history, however, with increased
risks, stress, overtime and budget.
All three projects agreed that more flexibility and adaptability should be planned into the
physical product development process using, for example, more schedule buffers and plan
for more adaptability after launch. One project member noted, “There are no [physical]
design resources available to ensure consistency between digital and physical later in the
downstream phase. They are now working on the next thing. Their resources are not
available anymore.” According to Project B it is still not clear how much the digital or the
physical product development processes should be adapted to each other. Nevertheless, all
projects proposed that COMP should put some effort into reducing lead times for critical
path processes, i.e. accommodating later binding, to better accommodate digital-physical
product development projects:
P6. In case of a high degree of interdependence, the physical product development
process is bound to experience high levels of exception management and can benefit
from building more flexibility into the process, including later binding and slack, to
accommodate the added uncertainty from the digital development process.
 EJIM So, the overall picture emerging is a combination of physical becoming more flexible and
digital becoming less adaptable. Had none of the partners adapted, a successful conduct of
the two processes with current technologies would have been unlikely:
P7. When combining a physical stability-optimized product development process and a
digital adaptability-optimized development process, either the former must become
more adaptable, the latter must become less adaptable or both need to change, which
may reduce the performance of one or both subprocesses but should lead to optimal
performance of the overall process.

4.3 Coordinating the digital and physical subprocesses

While the previous analysis focuses on changes that took place within the two subprocesses,
other actions were taken between the two processes through a range of coordination mechanisms.
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

Digital development had to adapt by increasingly engaging in an extensive front-end due

to many design interdependencies. Initially, the digital partners were commencing their
processes after concept lock, i.e. after the front-end work, leaving little digital solution space
and a short time until key milestones. Including the partners into the projects took a long
time and the digital-physical product concept was lacking the partners’ digital competences.
Partner C explained how the digital concept created by the largely physical designers did
not align with good digital practices in terms of language, graphics, colors and reuse options
that would enable an efficient development process. Partner C explained, “We did not feel
we had a lot of freedom to challenge […] Had we been involved earlier in the process or
allowed to change concepts the result would have been a better system.”
Engaging the digital partners earlier in the front-end processes was not easy due to a lack
of digital front-end capabilities at the partner companies’. Partner B explained, “Historically
we have not been involved in front-end processes. It was quite difficult for us to be involved in
the process, but not unenjoyable. The front-end guys are a bit mad as they are not considering
constraints. When we generate ideas, we always consider the constraints so it required a new
frame of mind for us.” One partner hired new people to meet the front-end needs. Another
struggled to allocate resources and found that hiring front-end resources conflicted with their
operating model, especially since their other clients did not have this need. Nevertheless, the
digital partners increasingly engaged more in the front-end work by delivering demonstration
videos and co-locating digital and physical designers for short periods of time with good
effects on performance. Still, the digital partners were not allocating enough time for front-end
from the perspective of the physical development teams.
Consequently, the physical development drove many of the design-decisions that should
have been digital-physical interdependent. Particularly in Project B, insufficient early digital
involvement in the first front-end cycle contributed to the project restarting concept development.
In contrast to COMP, the digital partners preferred not starting digital development
earlier on. Partner B explained that finishing well before launch and, in effect, aligning with
COMP’s development schedule before handing over to production, would remove much of
the motivating urgency for the software developers and make the digital solution outdated
before launch, causing extensive and expensive retrofitting of the product to the newest
technology, features, devices and digital platforms:
P8. If digital and physical collaborate toward early physical binding, digital
development is more likely to be able to successfully adapt to the early binding
and design constraint from the physical product development process.
Physical product development needed to learn about a new knowledge domain. Talking
about the digital and physical language differences, a physical graphic designer from
Project C explained: “It’s like speaking French and German” when both sides tried to
 understand each other’s design constraints. It was difficult for the physical teams to Digital-
understand the digital development process and the actual cost of change. Partner B physical
explained, “We were asked to add and change stuff quite often, with no consideration of product
what it costs, what not to do instead, etc. Being immature to software development,
[COMP] did not understand how complicated it is, especially supporting a large array of development
devices.” Similarly, it was hard for the digital teams to understand manufacturing
constraints, such as how readily COMP was willing to compromise digital product quality
to avoid the high cost of delay within the physical value chain by, e.g., missing out on a
shelf space reservation in shops. A digital producer from Partner C explained: “In digital
you don’t fight just as hard to reach a deadline, as you can typically release a patch later
that can fix the problem.”
To reduce the knowledge gap, internal digitally knowledgeable project members taught
the physical development teams about digital development methods and language.
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

The projects utilized such employees as liaison roles, helping to translate on behalf of both
sides. Unfortunately, these roles did not always understand the manufacturing processes
and the cost of delay. To further reduce the knowledge gaps the projects increasingly
focused on establishing digital-physical face-to-face time in the form of cross-functional
teamwork and multiple joint problem-solving sessions:
P9. To avoid faulty assumptions and, thus, ineffectiveness, digital-physical product
development can benefit significantly from ensuring cross-functionality, good
collaboration skills, a shared language and understanding of schedule and design
constraints, the cost of delay, the impact of uncertainty and the cost of change
associated with that.
Despite the sub-optimization and the digital partners compromising their adaptability to a
large extent, all projects were considered successful. Performance losses were largely
absorbed or hidden by other performance effects, buffers, inaccurate budgeting and
planning or digital development mitigating physical risks or errors after launch. This shows
the importance of making the right trade-offs throughout the digital-physical product
development process to carefully trade in a process sub-optimization effect for a more
significant performance gain. As one project manager from Project C put it: “If we only knew
the cost of delay [of milestones], we could better decide how to solve our problems.” Project
B, in which the physical development process was adapted the most, had several similar
learning points. According to the physical team: “We need to jointly design and plan our
processes up front to fit [the nature] of the project,” “We need to lock the specification
according to project risk and not when the schedule tells us to” and “Forcing our way of
working may be too risky.” Thus, making these trade-offs requires close cooperation and
effective communication to estimate the performance effects across both domains:
P10. In order to optimize a performance gain from combining the digital and the physical
product development processes, managing the trade-offs between these processes
requires close collaboration and effective communication.

4.4 New practices and contextual changes

The cases also show that new practices and contextual changes were adopted and
implemented to accommodate the new characteristics and materiality of the combined
digital-physical product development process. Table II presents an overview and also includes
practices recommended by project members that were not (sufficiently) implemented as well
as the perceived performance effects of not fully adapting to the new characteristics.
4.4.1 Uncertainty. COMP’s project teams did not have experience with digital product
development. Combined with the relative innovativeness of the new digital-physical products,
 EJIM Practices and context Performance effect

Uncertainty Team stress/

Implemented overtime
Much ad hoc problem-solving Higher process cost
Increased stakeholder management and network driven execution Lower product
Highly experienced project members comfortable working with high uncertainty quality
Ad hoc design of new support processes
More schedule buffers
More cross-functional co-location
Front-load work to avoid schedule delays
More front-end project members continuing to execution phase to avoid knowledge
loss
Establishment of some post launch operations
More time spent on understanding new market category
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

Suggested
Stronger collaboration/co-creation across digital and physical
Stronger risk focus
Shorten physical prototype development cycles
More and dedicated technical subject-matter experts in front-end to make “proof of
technology”
More T-shaped resources
More and earlier allocated IPR competences
Understand cost of delay for milestones
Faster project governance decision making
More risky business cases allowed
Fast reaction to post launch market feedback beyond digital, e.g., marketing or
pricing
New return policies
More reactive planning of processes, budgets and people
Shorten long lead-time processes to enable later binding
Digital platform development
More focus on marketing innovation
Mature and anchor post lunch operations
Diversity Lower product
Implemented quality
New digital competences Higher process cost
New digital technology to help develop digital prototypes
Suggested
Digital partner contract negotiation skills
New digital operating model competences and models
More digitally competent management
Interdependence Lower product
Implemented quality
Early and stronger focus on dependencies Higher process cost
Cadenced coordination meetings
More coordination competences
Cross-functional problem solving
Multi-domain liaison role
Some digital-physical co-location
Table II. New project organization
Overview of new Clarified roles
practices and Suggested
contextual changes Physical design should not be in the lead per default
with performance Improve partner collaboration skills
effects of not fully
adapting to the new
process characteristics (continued )
 Practices and context Performance effect
Digital-
physical
Complexity Lower product product
Implemented quality
More problem-solving loops development
The best subject-matter experts
New testing methods
Suggested
Policies for when to use internal or external competences
More T-shaped resources
Better online collaboration tools for non-co-located collaboration
Define digital quality acceptance criteria Table II.

this contributed to a higher uncertainty in the product requirements and schedules. This was
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

further aggravated in the case of Project B, which entered a new and much more volatile
market segment with new competitors, consumer needs and supply methods.
Due to the increased uncertainty, the flexibility needed within COMP often exceeded the
planned limits. This triggered the projects to front-load and enable more agility via, e.g.,
highly experienced project members, more cross-functionality and more schedule buffers
(see Table II). The project members suggested many additional changes to the ones
implemented, such as requesting more flexibility in budget and resource planning, T-shaped
resources and using shorter prototype development cycles.
The higher uncertainty and the digital immateriality encouraged COMP to spend more
time on understanding the market, return policies and business models. The existing
policies and models failed to answer key questions, e.g., regarding the digital distribution.
Other new practices involved innovative ways of communicating the new digital-physical
value propositions to the market place and ensuring that the products remained relevant to
the market after launch via product monitoring, maintenance, improvements, product data
collection, data analysis and decision-making to react with new marketing communication
or additional budget allocations for software updates.
Another challenge for the projects was the insufficient support from a slow moving and
largely physical product platform development process. Consequently, the projects had to
engage with uncertain technology development themselves. Finally, the rapidly developing
digital technologies required a more external orientation, not only to ensure access to
state-of-the-art digital technologies and information about a fast-moving market, but also to
expand the stakeholder management of the projects to include interfacing systems such as
operating systems, devices and app stores.
4.4.2 Diversity. COMP’s digital development resources were scarce, so several resources
were hired. The digital competences caused significantly larger and more diversely staffed
projects. The project management team initially doubled in size, to include the digital
competences needed to manage the new digital-physical interdependencies and requirements.
The projects also employed digital competences specific to marketing, software development,
software development management, data analysis and technology scouting. In addition to
these competences the projects requested digital competences for areas such as business model
development, purchasing (partner contract negotiations) and project governance. Lacking
digital insight at governance and advisory board level made it difficult to communicate
efficiently, get help to remove digital specific project impediments, and understand risks and
consequences. All these competences were needed in addition to the digital partners’
competences to enable successful front-end execution, project management, product
development and product quality assurance and, in effect, deliver the projects successfully.
The new roles were largely employed within various existing departments in COMP.
 EJIM 4.4.3 Interdependence. The projects experienced many digital-physical interdependencies.
Relative to Project A and B, Project C experienced a lower degree of interdependencies after the
concept lock in the second launch cycle, as clear interfaces between digital and physical product
components had been designed in a mature product architecture, allowing replacing
interdependencies from the first launch cycle with dependencies in the following cycles. Other
than suggesting better online collaboration tools, the responses to the higher degree of
interdependence included a new focus on dependency mapping, multiple cadenced coordination
meetings, integrating activities more on a task level via some co-location in the early project
phases and hiring digital coordination competences to also function as liaison roles.
4.4.4 Complexity. The perceived complexity of the digital-physical product development
process increased due to the integration of digital product components, which introduced
more requirements and made problems more multi-faceted. A physical product designer
from Project A explained how they not only had to work with and learn about additional
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

design requirements, requiring the best experts, but also how they spent more time on
problem solving as many problems could have several causes: physical, digital or the
integration between them. The projects also discovered that the mixed materiality of the
integrated digital-physical prototypes required new testing criteria and methods.
Furthermore, with bugs being a given to most software, all projects were challenged by a
lack of digital quality guidelines. Finally, the projects invested in digital tools and devices,
for example to make digital-physical prototypes in the front-end process:
P11. Compared to traditional physical product development, digital-physical
development is likely to involve a higher degree of uncertainty, interdependence,
complexity and diversity that, coupled with the digital materiality, requires
changes in product development practices, including rethinking marketing
strategies, becoming more externally oriented, developing new testing methods,
organizing the project to include digital competences at all levels and focusing on
product architecture and dependency mapping.
P12. Compared to traditional physical product development, digital-physical
development is likely to involve a higher degree of uncertainty, interdependence,
complexity and diversity that, coupled with the digital immateriality, requires
changes in the development context, including business model innovation,
rethinking marketing operations, establishing post launch operations and new
digital quality acceptance criteria, rethinking product platform(s), investing in
digital tools, and ensuring digital competences for project governance and support
functions including marketing, business model development and purchasing.

5. Discussion
This research contributes to conceptual theory building within the immature field of
digital-physical product development. It uses deductive reasoning by taking a starting point
in a literature-based research model, exemplifying the model and developing twelve case
study-based propositions on its constructs and relationships – see Figure 3.
The case studies confirm that digital and physical development processes differ in terms of
uncertainty, diversity (Boer and During, 2001) and materiality, which leads to dissimilar
organizational and managerial practices supporting these processes (e.g. Broy, 2005; Woodward
and Mosterman, 2007; Lwakatare et al., 2016; Porter and Heppelmann, 2015; Lwakatare et al.,
2016) and explains why the two processes are executed separately (P1 and P2).
Three supplementary strategies can be used to combine digital and physical
development effectively:
(1) align the two subprocesses, i.e. adapt them to each other;
 Physical product
Digital-
development process physical
characteristics product
development
Development
2 context

(Dissimilar) 11 Digital-physical
3–7, 12
development development
practices performance

Coordination
1 8–10
practices
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

Software development Figure 3.

process The research model
characteristics and propositions

(2) coordinate the two subprocesses; and

(3) create new development practices and a suitable development context.
Adapting the digital process to the physical development process does not necessarily
hamper quality (P4). However, imposing early binding, which is typical for physical
development, limits the exploitation of new knowledge in the digital process, and reduces
the digital adaptability and the potential value of the digital-physical product (P3).
Options to mitigate these problems include adapting the physical to the digital process
through a more agile development approach (e.g. Cooper, 2016) (P5), enabling the later
binding typical for digital development (Yoo et al., 2010), adding more slack (P6)
and adapting the two subprocesses to each other (Karlström and Runeson, 2006; Eklund
and Berger, 2017; Könnölä et al., 2016) (P7). Altogether, however, it seems that it is easier
for software to adapt to physical product development (e.g. Cordeiro et al., 2007; Eklund
and Bosch, 2012), rather than the other way around, and adopt a long-term, plan-driven,
stage-gate approach to dominate the combined process (Cordeiro et al., 2007; Eklund and
Bosch, 2012; Lwakatare et al., 2016).
In any case, the two subprocesses take place separately during the majority of their
execution (e.g. Evans, 2009; Huang et al., 2012; Cooper, 2016), especially if the digital
subprocess is outsourced. Complete alignment may not be possible or fail to create the
desired performance effects. In that case, additional mechanisms are needed, to coordinate
the two subprocesses. The case studies suggest various coordination mechanisms, including
early collaboration to reduce the early binding problem (P8), cross-functionality (Könnölä
et al., 2016; Eklund and Berger, 2017), collaboration skills and a shared language and
understanding of schedule and design constraints, the cost of delay, the impact of
uncertainty and the associated cost of change (P9). In addition, new development practices
are needed such as a digital-physical marketing strategy, external orientation, new testing
methods, including digital competences at all levels in the project organization and a focus
on product architecture and dependency mapping (P11). Finally, a context suiting the
combined digital-physical development process is needed, which better reflects the typical
software development environment (Porter and Heppelmann, 2015) and also includes
rethought marketing operations, post launch operations and new digital quality acceptance
 EJIM criteria, digital-physical product platform(s), investments in digital tools and digitally
competent project governance, marketing, purchasing and business development (P12).
Figure 3 positions the propositions in the research model depicted in Figure 1.

6. Conclusion
6.1 Contribution
This paper empirically explores the complex process of digital-physical product development.
Taking a process perspective and focusing on organizational and managerial practices and
the influence of context, organization theory is used as the theoretical lens.
The paper takes a starting point in a research model, investigates its constructs and
relationships, and offers tentative explanations in the form of twelve propositions for further
research of digital-physical product development. The propositions suggest that, when
combining a physical stability-optimized and a digital adaptability-optimized development
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

process, either the former must become more adaptable, the latter must become less adaptable
or both must change, which in all cases reduces the performance of either or both
subprocesses. A combination of actions remedies this problem and leads to optimal
performance of the combined process: adapting either or both of the subprocesses to the other,
effective coordination between the two subprocesses and implementation of new development
practices and a suitable development context. Trade-offs are inevitable, though, and require
effective coordination and communication between the stakeholders involved.
The practical implications of the research are embedded in the propositions and
summarized in Table II. They align with and add to observations made in previous literature
and offer valuable guidelines for manufacturing companies toward understanding important
trade-off, process and context design decisions supporting digitalizing manufacturers in an
increasingly fast-moving environment.

6.2 Limitations and further research

The research is based on a study of multiple projects in one large, mature and highly
successful company, which allows creating detailed insight but also presents some important
limitations related to method and context. Avenues for future research include exploring the
topic and the proposed propositions from the perspective of a smaller or less mature
manufacturer, a software company that is adding a physical product dimension, or a born
digital-physical product development company. Further exploration in the form of case
studies or action research is needed to understand which coordination practices or scenarios
are most effective and in which contexts, while larger scale studies are needed to move from
tentative theory and test the propositions.
The research took its starting point in the processes of digital and physical development
and used organization theory as the lens to investigate organizational, managerial and
contextual aspects. Other relevant lenses include a resource or knowledge-based perspective,
which would help study the role of IT systems and the ability of manufacturers to provide and
implement the right practices and context, i.e. development and managerial competences, as
well as dynamic capabilities and absorptive capacity.

References
Ardito, L., D’Adda, D. and Messeni Petruzzelli, A. (2017), “Mapping innovation dynamics in the internet
of things domain: evidence from patent analysis”, Technological Forecasting and Social Change.
Baines, T.S., Lightfoot, H.W., Evans, S., Neely, A., Greenough, R., Peppard, J., Roy, R., Shehab, E.,
Braganza, A., Tiwari, A., Alcock, J.R., Angus, J.P., Bastl, M., Cousens, A., Irving, P., Johnson, M.,
Kingston, J., Lockett, H., Martinez, V., Michele, P., Tranfield, D., Walton, I.M. and Wilson, H.
(2007), “State-of-the-art in product-service systems”, Proceedings of the Institution of Mechanical
Engineers, Part B: Journal of Engineering Manufacture, Vol. 221 No. 10, pp. 1543-1552.
 Bialasiewicz, J.T. (2017), “Interaction of concurrent processes in cyber-physical systems”, Proceedings Digital-
of the 14th International Conference on Industrial Informatics, INDIN 2016, IEEE, Poitiers, physical
July 19-21, pp. 516-521.
product
Boer, H. (1991), Organising Innovative Manufacturing Systems, Gower, Aldershot.
development
Boer, H. and During, W.E. (2001), “Innovation. What innovation? A comparison between product,
process and organisational innovation”, International Journal of Technology Management,
Vol. 22 Nos 1-3, pp. 83-107.
Broy, M. (2005), “Automotive software and systems engineering”, Proceedings of the 3rd ACM & IEEE
International Conference on Formal Methods and Models for Co-design, Verona, July 11-14,
pp. 143-149.
Caniato, F., Doran, D., Sousa, R. and Boer, H. (forthcoming), “Designing and developing OM research.
From concept to publication”, International Journal of Operations & Production Management.
Christensen, C.M. (2006), “The ongoing process of building a theory of disruption”, Journal of Product
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

Innovation Management, Vol. 23 No. 1, pp. 39-55.

Cohn, M. (2010), Succeeding with Agile, Pearson Education, Boston, MA.
Cooper, R.G. (1990), “Stage-gate systems: a new tool for managing new products”, Business Horizons,
Vol. 33 No. 3, pp. 44-54.
Cooper, R.G. (2016), “Agile-stage-gate hybrids: the next stage for product development”, Research
Technology Management, Vol. 59 No. 1, pp. 21-29.
Cordeiro, L., Barreto, R., Barcelos, R., Oliveira, M., Lucena, V. and Maciel, P. (2007), “Agile development
methodology for embedded systems: a platform-based design approach”, in Leaney, J., Rozenblit, J.W.
and Peng, J. (Eds), Proceedings of the 14th Annual IEEE International Conference and Workshops on
the Engineering of Computer-Based Systems, Institute of Electrical and Electronics Engineers Inc,
Tucson, AZ, March 26-29, pp. 195-202.
Diegel, O., Singamneni, S. and Withell, A. (2008), “A mechatronics approach to rapid product
development: a case study”, Proceedings of the 15th International Conference on Mechatronics and
Machine Vision in Practice (M2VIP), Massey University, Auckland, December 2-4, pp. 565-570.
Eklund, U. and Bosch, J. (2012), “Introducing software ecosystems for mass-produced embedded systems”,
in Cusumano, M.A., Iyer, B. and Venkatraman, N. (Eds), Proceedings of the 3rd International
Conference, ICSOB 2012, Springer International Publishing, Cambridge, MA, pp. 248-254.
Eklund, U. and Berger, C. (2017), “Scaling agile development in mechatronic organizations: a
comparative case study”, Proceedings of the 39th IEEE/ACM International Conference on
Software Engineering: Software Engineering in Practice Track, ICSE-SEIP 2017, Department of
Computer Science and Media Technology, Malmö University, Institute of Electrical and
Electronics Engineers, Malmö, pp. 173-182.
Eklund, U., Olsson, H.H. and Strom, N.J. (2014), “Industrial challenges of scaling agile in mass-produced
embedded systems”, in Dingsoyr, T., Moe, N.B., Tonelli, R., Counsell, S., Gencel, C. and Petersen, K.
(Eds), Proceedings of the 15th International Conference on Agile Software Development, Springer
International Publishing, Rome, May 26-30, pp. 30-42.
Evans, R. (2009), “Make functionality, not hardware”, Military & Aerospace Electronics, Vol. 20 No. 9,
pp. 16-17.
Haghighatkhah, A., Banijamali, A., Pakanen, O., Oivo, M. and Kuvaja, P. (2017), “Automotive software
engineering: a systematic mapping study”, Journal of Systems and Software, Vol. 128, June,
pp. 25-55.
Hendler, S. and Boer, H. ( forthcoming), “Digital-physical product developments: a review and research
agenda”, International Journal of Technology Management.
Holmström, J. (2018), “Recombination in digital innovation: challenges, opportunities, and the
importance of a theoretical framework”, Information and organization, Vol. 28 No. 2, pp. 107-110.
Huang, P.M., Darrin, A.G. and Knuth, A.A. (2012), “Agile hardware and software system engineering for
innovation”, Proceedings of the 2012 IEEE Aerospace Conference, Big Sky, MT, March 3-10, pp. 1-10.
 EJIM Joglekar, N.R. and Rosenthal, S.R. (2003), “Coordination of design supply chains for bundling physical
and software products”, Journal of Product Innovation Management, Vol. 20 No. 5, pp. 374-390.
Karlsson, C. and Lovén, E. (2005), “Managing new technology integration: integrating software in
manufactured products”, International Journal of Innovation Management, Vol. 9 No. 3, pp. 343-370.
Karlström, D. and Runeson, P. (2006), “Integrating agile software development into stage-gate
managed product development”, Empirical Software Engineering, Vol. 11 No. 2, pp. 203-225.
Könnölä, K., Suomi, S., Mäkilä, T., Jokela, T., Rantala, V. and Lehtonen, T. (2016), “Agile methods in
embedded system development: multiple-case study of three industrial cases”, Journal of
Systems and Software, Vol. 118, August, pp. 134-150.
Kuwashima, K. and Fujimoto, T. (2013), “Performance measurement in product development research:
A literature review”, Annals of Business Administrative Science, Vol. 12 No. 4, pp. 213-223.
Lee, J. and Berente, N. (2012), “Digital innovation and the division of innovative labor: Digital controls
in the automotive industry”, Organization Science, Vol. 23 No. 5, pp. 1428-1447.
Downloaded by University of Sunderland At 22:06 19 September 2018 (PT)

Lwakatare, L.E., Karvonen, T., Sauvola, T., Kuvaja, P., Olsson, H.H., Bosch, J. and Oivo, M. (2016),
“Towards DevOps in the embedded systems domain: why is it so hard?”, in Sprague, R.H. and
Bui, T.X. (Eds), Proceedings of the 49th Annual Hawaii International Conference on System
Sciences, HICSS 2016, University of Oulu, pp. 5437-5446.
Maleki, E., Belkadi, F. and Bernard, A. (2017), “Systems engineering as a foundation for PSS
development project: Motivations and perspectives”, in McAloone, T.C., Tseng, M.M.,
Mortensen, N.H., Shimomura, Y. and Pigosso, D.C.A. (Eds), Proceedings of the 50th CIRP
Conference on Manufacturing Systems 2017, Ecole Centrale de Nantes, Nantes, pp. 205-210.
Nambisan, S., Lyytinen, K., Majchrzak, A. and Song, M. (2017), “Digital innovation management:
Reinventing innovation management research in a digital world”, MIS Quarterly, Vol. 41 No. 1,
pp. 223-238.
Porter, M.E. and Heppelmann, J.E. (2015), “How smart, connected products are transforming
companies”, Harvard Business Review, Vol. 93 No. 10, pp. 96-114.
Siggelkow, N. (2007), “Persuasion with case studies”, Academy of Management Journal, Vol. 50 No. 1,
pp. 20-24.
Svahn, F. and Henfridsson, O. (2012), “The dual regimes of digital innovation management”, Proceedings
of the Annual Hawaii International Conference on System Sciences, Maui, HI, pp. 3347-3356.
Wacker, J. (1998), “A definition of theory: research guidelines for different theory-building research
methods in operations management”, Journal of Operations Management, Vol. 16 No. 4, pp. 361-385.
Wolfenstetter, T., Böhm, M., Krcmar, H. and Bründl, S. (2016), “Why product service systems
development is special”, in Framinan, J.M., Gonzalez, P. and Artiba, A. (Eds), Proceedings of the
International Conference on Industrial Engineering and Systems Management, IEEE IESM
2015, Institute of Electrical and Electronics Engineers Inc, Seville, October 21-23, pp. 1221-1228.
Woodward, M.V. and Mosterman, P.J. (2007), “Challenges for embedded software development”,
Proceedings of the 50th Midwest Symposium on Circuits and Systems, Montreal, Vol. 1-3, pp. 524-527.
Yin, R.K. (2009), Case Study Research: Design and Methods, 4th ed., Calif Sage Publications,
Thousand Oaks, CA.
Yoo, Y., Henfridsson, O. and Lyytinen, K. (2010), “The new organizing logic of digital innovation: an
agenda for information systems research”, Information Systems Research, Vol. 21 No. 4, pp. 724-735.
Yoo, Y., Boland, R.J., Lyytinen, K. and Majchrzak, A. (2012), “Organizing for innovation in the digitized
world”, Organization Science, Vol. 23 No. 5, pp. 1398-1408.

Corresponding author
Stine Hendler can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]
J PROD INNOV MANAG 2014;31(4):744–764
© 2013 Product Development & Management Association
DOI: 10.1111/jpim.12121

Collaborative Prototyping: Cross-Fertilization of Knowledge

in Prototype-Driven Problem Solving*
Marcel Bogers and Willem Horst

This paper presents an inductive study that shows how collaborative prototyping across functional, hierarchical, and
organizational boundaries can improve the overall prototyping process. Our combined action research and case study
approach provides new insights into how collaborative prototyping can provide a platform for prototype-driven
problem solving in early new product development (NPD). Our findings have important implications for how to
facilitate multistakeholder collaboration in prototyping and problem solving, and more generally for how to organize
collaborative and open innovation processes.
Our analysis reveals two levels of prototyping. Besides the more formal managerial level, we identify the informal
designer level, where the actual practice of prototyping takes place. On this level, collaborative prototyping transforms
the act of prototyping from an activity belonging exclusively to the domain of design engineers to an activity integral to
NPD, with participants from within the organization (different functions and managers) and from outside (consultants
and users). In effect, this collapses the discrete steps in the prototyping process (at the managerial level) to an essentially
continuous process of iterative problem solving (at the designer level) that is centered around the collaborative
prototype, which allows participants to see their suggestions implemented and exposing them to the design constraints.
The study, moreover, shows how, at various stages of the prototyping process, the actual prototype was used as a tool
for communication or development, thus serving as a platform for the cross-fertilization of knowledge. In this way,
collaborative prototyping leads to a better balance between functionality and usability; it translates usability problems
into design changes, and it detects emerging usability problems through active engagement and experimentation. As
such, the collaborative prototype acts as a boundary object to represent, understand, and transform knowledge across
functional, hierarchical, and organizational boundaries.
Our study also identifies some constraints in involving the appropriate stakeholders at the right time. The paper
specifically elaborates on the role of users in collaborative prototyping, which is important in order to cover all phases
of the problem-solving cycle but triggers an interesting challenge due to the “reverse empathy” that a user may develop
for the design constraints—parallel to the designer empathy for the user context. Finally, our study shows that despite
the continuous nature of the (designer) practice of prototyping, there are certain windows of opportunities (at the
managerial level) during which the collaborative prototyping approach actually leads to changes in the product design.

Introduction 2008; Ulrich, 2011; Verganti, 2009). In particular,

prototyping is considered an important design practice,

T
he importance of design research and design which thereby becomes a central element in corporate
thinking is increasingly being recognized within innovation processes (Leonard and Rayport, 1997;
the context of new product development (NPD) Mascitelli, 2000; Schrage, 2000), as exemplified by
due to the emphasis on the process of problem solving to several important advances in this area (e.g., Buchenau
address user needs and create new opportunities (Brown, and Fulton Suri, 2000; Hartmann, 2009; Terwiesch and
Loch, 2004; Thomke, 1998). There is, moreover, a rec-
Address correspondence to: Marcel Bogers, Mads Clausen Institute,
ognition that prototyping can improve NPD by perform-
University of Southern Denmark, Alsion 2, Sønderborg 6400, Denmark. ing it early in the process and by involving a number of
E-mail: [email protected]. Tel: +45 6550 1284. relevant stakeholders within and outside the design team
* The authors would like to thank the people at Danfoss Heating Solu-
tions who were involved in the project described in this paper for their and organization (Buchenau and Fulton Suri, 2000;
support during the project and their cooperation afterwards. Special thanks Mascitelli, 2000; Terwiesch and Loch, 2004).
go to Anders Østergaard Clausen for his support and engagement through-
out the process. The authors are also grateful for the editor’s guidance and Within NPD, the product specification is typically
the constructive reviewer comments, which helped improve this paper. made before prototyping starts, which has been
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 745
2014;31(4):744–764

questioned in both software and product development (cf. the early stages of the NPD process, typically before
Boehm, Gray, and Seewaldt, 1984; Rudd, Stern, and starting the design of the user interface (cf. Thomke,
Isensee, 1996; Thomke and Bell, 2001). In fact, in the 1998). However, how features are implemented and
early phase of the NPD process, prototyping can play an available in the product relates to their usability and use-
important role in informing the development of the fulness. Therefore, starting the design of the user inter-
product specification (Mascitelli, 2000; Thomke, 1998). face before finalizing the feature-set helps in defining an
For example, at IDEO, the development team starts with appropriate feature-set. Emphasizing the experiential
prototyping to develop the specification, and at later aspect in active engagement with prototypes, moreover,
stages develops prototypes based on this specification. enables coming up with new solutions in unknown
David Kelley, founder and chairman of IDEO, argues that design spaces (Buchenau and Fulton Suri, 2000). By
“organizations intending to be innovative need to move making interactive prototypes that simulate the function-
from specification-driven prototypes to prototype-driven ality at this early stage, it is possible to iteratively evalu-
specifications” (quoted in Schrage, 1996, p. 195). Hans- ate the prototype’s usability as features are added or
Christoph Haenlein, (former) Director of Prototyping at removed, or implemented in different ways.
IDEO, describes three prototyping phases to inspire, This paper presents an exploration of a collaborative
evolve, and validate the specification. In the first two prototyping process within a particular organization, in
phases, “project specifications are derived from the pro- which a prototype was used to help define the feature-set
totypes. Towards the end of a project, very complete by involving not only common stakeholders as interac-
prototypes are built to validate the design specification as tion designers and usability experts, but also other stake-
a whole” (quoted in Hartmann, 2009, p. 22). holders as marketers, directors, and users. This study is
Combining the prototyping and specifying ap- based on an NPD project in which the prototyping
proaches in this way can also help deal with the tension process was initiated before the features were fully
within NPD between the number of features and the defined, and it continued until the final software specifi-
usability of the product (Keijzers, den Ouden, and Lu, cation over a 1-year period. During this process, the
2008). This is especially true for electronic products with purpose of the prototype changed as both the feature-set
a user interface, since features can be added at relatively and the user interface became more defined. Depending
low cost, making it tempting to do so in a competitive on the purpose of the prototype, different stakeholders
market where the number of features is an important were involved in a collaborative prototyping process.
competitive factor (Rust, Thompson, and Hamilton, This paper is structured as follows. The next section
2006). Specifying the feature-set of a product happens in presents a review of the relevant literature on (colla-
borative) prototyping in NPD. The following section
describes the research methodology and introduces the
company case. Next, the empirical results of our study
BIOGRAPHICAL SKETCHES
are presented based on “reflection-in-action” within an
action research approach and “reflection-on-action”
Dr. Marcel Bogers is an associate professor of innovation and entrepre-
neurship at the University of Southern Denmark. He received his Ph.D. within three cases of collaborative prototyping processes
in management of technology from the Ecole Polytechnique Fédérale de with different stakeholders. Then, the findings section
Lausanne (Swiss Federal Institute of Technology). He also studied at presents an analysis and discussion of our results, which
the Eindhoven University of Technology, University of California at
Berkeley, and Chalmers University of Technology. His main interests
suggest that (1) collaborative prototyping in early NPD
center around the design, organization, and management of technology creates a prototype-driven approach to problem solving
and innovation. More specifically, he has studied areas such as business that collapses problem-solving cycles into a continuous
models, open innovation, users as innovators, collaborative prototyping, iteration centered around the prototype, (2) it improves
entrepreneurship, improvisation, and learning-by-doing.
the functionality–usability balance and focuses on actual
Dr. Willem Horst is a user experience software producer at LEGO and is design changes, (3) active engagement in collaborative
also a freelance interaction design consultant. He received his Ph.D. in
design and innovation from the University of Southern Denmark, and
prototyping enables collaboration across various bound-
has a background in industrial design (BSc) and IT product design aries, and (4) there are certain windows of opportunities
(MSc). His research focuses on the potential of prototypes as platforms for actual changes in the product design. Finally, the
for participation across different stakeholders in new product develop-
conclusion section summarizes our results, which is fol-
ment. He has been involved in new product development projects at
Philips, LEGO, and Danfoss to develop a collaborative approach to lowed by a discussion of the implications for manage-
prototyping interactive products. ment practice and theory in relation to collaborative
prototyping in NPD and beyond.
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
746 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

Literature Review describes the risk of exposing important stakeholders to

the prototype at a late stage in the project, as in such cases
Collaborative Prototyping in NPD “the prototype becomes a medium for persuasion, rather
than a vehicle to evoke discussion. It is used to prove a
In the management literature, the prototyping process is point, rather than to create a platform for a design dialog.”
conceptualized as an iterative trial-and-error learning This is especially true for top managers who are involved
process, following four steps: (1) design, (2) build, late in the design cycle and then “are being asked to
(3) run, and (4) analyze (Thomke, 1998; von Hippel, approve—rather than to review or assist—new-product
2005). Moreover, as a design practice, prototyping is a creation” (Schrage, 1996, p. 200).
central element in corporate innovation processes
(Leonard and Rayport, 1997; Mascitelli, 2000; Schrage, Toward Prototype-Driven Problem Solving
2000), thus making design (thinking) an important part
of problem solving within NPD (Brown, 2008; Ulrich, Boehm et al. (1984) make a distinction between taking a
2011; Verganti, 2009). An important development within specifying and a prototyping approach in software devel-
the literature related to prototyping (and NPD and inno- opment, concluding that a prototyping approach produces
vation at large) is a consideration of how designers can comparable prototypes but with significantly less code
better engage other stakeholders in the prototyping and effort than the specifying approach, although the
process. latter approach produced code that was more coherent
In the literature, various reasons are discussed for and easier to integrate.
involving others in the prototyping process. Terwiesch Moving from specifying to prototyping may enable a
and Loch (2004) describe the collaborative prototyping better understanding of both technical and customer
process as a search process for the ideal product specifi- need-related problems, for example due to the improved
cation for custom-made products involving the producer fidelity of the problem-solving process (Foray, 2004;
and user of the product. In participatory design, the idea Thomke and Bell, 2001; von Hippel, 2005). In the context
of cooperative prototyping, which involves the end users of rapid prototyping, for example of a user interface,
of the interface that is being designed rather than dem- high-fidelity prototypes are particularly beneficial to
onstrating it to them, is used to learn more about the provide a holistic perspective based on interactive explo-
actual use context and shape the technology (Bødker and ration and testing, although possible drawbacks may
Grønbæk, 1991). Toolkits for innovation give users the include cost and time considerations (Dey, Abowd, and
possibility to design, simulate, or prototype a custom Salber, 2001; Rudd et al., 1996).
product or service (e.g., Franke and Piller, 2004; The transition from specification-driven prototypes to
Jeppesen, 2005). An important characteristic of such tool- prototype-driven specifications raises the question who
kits is that they “enable users to carry out complete cycles should be involved in the prototyping process (cf.
of trial-and-error learning” (von Hippel and Katz, 2002, Schrage, 1996). Usually, the people developing the speci-
p. 825), which is an important aspect of the prototyping fication (e.g., marketers or top management) are not the
process. Although this shifts the locus of innovation to same people who make the prototypes (e.g., design engi-
users, the design of the toolkit is a collaborative process, neers). In a prototype-driven process, it will thus be
where the manufacturer can optimize its design based on important to involve these various stakeholders, who may
feedback from users (Bogers, Afuah, and Bastian, 2010). be spread across various functions and hierarchical levels
Many such approaches relate to involving end users in (Adler, 1995; Buur and Matthews, 2008; Song et al.,
prototyping, and as such relate to cross-organizational 1998).
NPD processes. In addition, intra-organizational bound-
aries across the different functions involved, such as Experience Prototyping
research and development (R&D), production, and
management, also play an important role in NPD Collaborative prototyping can be seen as a practice that
(Atuahene-Gima and Wei, 2011; Song, Montoya-Weiss, relies on the involvement of various stakeholders. Engag-
and Schmidt, 1997; Song, Thieme, and Xie, 1998). In ing in such activities involves an exploration of the rel-
such context of internal NPD processes, Henderson evant design space as in line with the more general
(1995) considers the politics of prototypes, as they discussion of how to best use prototypes within (partici-
embody particular points of view, and can be used as patory) design (Hartmann, 2009; Lim, Stolterman, and
arguments in an organization. Schrage (1996, p. 200) Tenenberg, 2008; Schuler and Namioka, 1993). For
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 747
2014;31(4):744–764

example, human embodied engagement has been Table 1. Overview of Data Sources
advocated within interaction design as a way to shape
Data Source Amount of Data
experience, understanding, and interactions in design
practices (Klemmer, Hartmann, and Takayama, 2006). Prototype iterations (total) (70)
Moreover, the concept of learning by doing highlights the Shared with larger team 25
Developed in workshops 13
important role of the use experience in identifying and
E-mails (total) (60)
solving problems within innovation activities (Macher Design line specialist 23
and Mowery, 2003; von Hippel and Tyre, 1995). Interaction design consultant 23
Buchenau and Fulton Suri (2000, p. 424) propose Intern 9
“experience prototyping” as “a form of prototyping that R&D project manager 3
Other 2
enables design team members, users and clients to gain
Documentation (total) (22)
first-hand appreciation of existing or future conditions Click-thru scenarios 13
through active engagement with prototypes.” The aim of Design change proposals 4
an experience prototype would be to design an integrated Usability reports 3
experience to better understand, explore, or communicate Organization structure 2
what it might be like to engage with the product that is Video material (total) (24 hours)
Workshops 9 hours
being designed (Buchenau and Fulton Suri, 2000). The Usability tests 15 hours
high fidelity of actively engaging in a prototype will Interviews (total number/time) (5/3 hours)
improve the overall performance of the problem-solving Design line specialist 1 hour
process (Foray, 2004; Lim et al., 2008). R&D project manager 30 minutes
R&D senior director 30 minutes
Global webmaster 30 minutes
Method and Research Context Product marketer 30 minutes
Most data are available in Danish, some in English, and some fractions in
Study Setting and Data Sources Dutch (depending on the people involved). Audio and video material was
transcribed in the original language and translated into English for the
analysis. Thus, all quotations used in this paper were translated into English
The study presented in this paper adopts multiple (if not in English already).
methods to report and analyze a series of experiences in
the development of the interface of the living eco® radia-
tor thermostat at the Danish company Danfoss Heating behavior and interaction, and ties together all assets.
Solutions—a division of the larger Danfoss corporation. Second, 20 text files describe the structure of the inter-
The research presented here was developed in two stages. face (e.g., which icons are placed on each screen, or what
In the first stage, an action research approach was happens next when an icon is selected), and another 5
taken in which the second author was involved as an text files describing various default values and settings.
external interaction design consultant (IDC1) at Danfoss Finally, there are 80 graphics that simulate each of the
Heating Solutions over a period of a year (in 2009 and segments of the final segment display. Because each
2010). During this time, the collaborative prototyping iteration was stored in a different folder, it was possible to
method was developed and deployed in several work- trace back the development of the prototype, and pinpoint
shops, to develop the interactive prototype with when design decisions were taken by looking at the date
multistakeholder involvement. Data collected during this and time these files were last modified. Using a custom-
period include video material from those workshops, made analysis tool allowed stepping through the project
e-mail conversations with the design line specialist day by day, which enabled us to see, for example, what
(DLS), project documentation, and the 70 iterations of kind of design decisions were made in the various work-
the interactive prototype. See Table 1 for an overview of shops, when specifications were updated, and what the
the different data sources that are used in this paper. e-mail response was to the different prototype versions.
The prototype was run and developed on a PC. It was In the second stage, an exploratory case study
designed to be interactive and flexible. For this reason, it approach was followed using the case of Danfoss Heating
was broken up into three different components. First, Solutions. More specifically, the case study compared
there is the programming code, which defines the general three types of prototyping within Danfoss Heating Solu-
tions. The analyzed information about the prototyping
1
A complete list of acronyms of involved stakeholders can be found in activities came from different sources, such as documen-
Appendix A. tation about specifications, scenarios and strategies,
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
748 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

e-mail correspondence related to the prototyping activi- laborative prototyping approach—cf. “thick description”
ties, and video material from various workshops. In addi- (Geertz, 1973)—and will describe how it was used in an
tion, five interviews were conducted with the DLS, the NPD project by organizing workshops that facilitate the
R&D senior director (RDSD), the R&D project manager involvement of various stakeholders. The presented evi-
(RDPM), the global webmaster, and a product marketer dence is based on the observations and reflections during
after the prototype had been used within the organization relevant events, as well as from video recordings (which
for over a year. The interviews focused on the role the also allow us to present evidence of conversations and
prototype had played in the work of the different disci- interactions). Thereby, illustrative examples of relevant
plines involved. Each interview was recorded and tran- activities, events, and discussions can be presented,
scribed, and lasted between 30 minutes and one hour. complemented by later reflections (cf. reflection-on-
(See Table 1 for an overview of data sources.) action below) and related literature to put our reflection-
in-action into a broader perspective. Especially the IDC
Data Analysis (i.e., the researcher–designer) and the DLS will be central
in this reflective narrative given their roles in this project
In order to combine the two stages within our research and research study.
design, namely action research and case study, our The second part of this study entails reflection-on-
analysis is organized around “reflection-in-action” and action, in which the researchers, after the first stage of the
“reflection-on-action,” respectively (Schön, 1983). research was concluded, “think back on a project they
Essentially reflecting the main principles of action have undertaken . . . and they explore the understandings
research, reflection-in-action entails that the researcher as they have brought to their handling of the case” (Schön,
practitioner “reflects on the phenomenon before him, and 1983, p. 61). As such, this exploratory study follows the
on the prior understandings which have been implicit in principles of the qualitative case study approach
his behaviour. He carries out an experiment which serves (Eisenhardt, 1989; Eisenhardt and Graebner, 2007; Yin,
to generate both a new understanding of the phenomena 2003). In order to ensure construct validity, a general
and a change in the situation” (Schön, 1983, p. 68). structure of questions and framework was used to inves-
Action research works within the realm of practical tigate and analyze the different prototyping activities,
knowing while emphasizing cycles of action and while a time line of activities was also used to construct a
reflection, with activities comprising planning, action, chain of evidence (i.e., how certain events caused particu-
and fact-finding (Coghlan, 2011; Lewin, 1946). The lar effects). Triangulation was established by using mul-
reflection-in-action presented in this paper builds on an tiple sources of information and by relying on multiple
action research project that extended over a 1-year period. informants (e.g., interviews). Even though the case study
The project also built on design research, in which this is conducted in a single organization, our study is based
approach is commonly taken when an important aspect of on three separate cases—the unit of analysis being the
the research is to find out how new design methods work prototyping activities—which enable the prediction of
in practice (Binder and Redström, 2006). Instead of similar results across the cases (literal replication) and
observing what happens as an objective researcher, the predict contrasting results for predictable reasons (theo-
researcher takes a dual role of researcher–designer who retical replication), which thereby increase external valid-
takes an active part in the design activity (i.e., the design ity of the study (Yin, 2003). The analysis of the
of the user interface in the NPD project). This type of prototyping cases was based on the identification of cat-
research is future-oriented and action-driven—through egories of findings from within the cases, while compar-
doing interventions and analyzing the outcome, it is pos- ing the finding across cases as an analytic technique.
sible to develop “field-tested and grounded technological Given the inductive nature of this research, the principles
rules to be used as design exemplars of managerial of grounded theory were applied to construct categories
problem solving” (van Aken, 2004, p. 221). The of findings by developing categories of information
researcher–designer works on a practice problem and (open coding), interconnecting the categories (selective
designs a solution concept. This is the basis for the design coding), and building a story that connects the categories
of the intervention, which is a context-specific solution. (axial coding), upon which the final findings are based
Through evaluating and reflecting on the outcome, the (Dougherty, 2002; Strauss and Corbin, 1990). As such,
researcher can revise the solution concept and redesign a the construction of categories can be seen as an iterative
specific solution (Andriessen, 2007). The findings section process that establishes common meaning across mul-
will provide an overview of the development of the col- tiple observations (Locke, 2001).
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 749
2014;31(4):744–764

About the Company and NPD Project user interface would have to be revised multiple times.
Therefore, the interface and the exact features were not
This paper is based on the development of the Danfoss frozen until these tests were done, although some deci-
living eco® radiator thermostat at Danfoss Heating Solu- sions were made on aspects that related to the product
tions. The eco® is a programmable radiator thermostat hardware. These hardware decisions provided the frame-
containing electronics and a user interface, which can be work for choosing an appropriate prototyping approach
mounted onto any radiator and has a similar form factor and medium. It was decided that the product would have
as a conventional radiator thermostat. Based on the three buttons (up, down, and enter) and a circular segment
schedule set by the user and the temperature measured by display with a diameter of 25 mm. The choice for a
the temperature sensor, a small motor controls the radia- segment display, as opposed to for example a matrix
tor valve to regulate the temperature. The products offer display, was an important constraint. With a matrix
customers a convenient way of saving energy, by for display, the exact icons can be changed at a later stage of
example automatically lowering the temperature at night the project at low cost because it is possible to make them
and/or working hours. in code. With a segment display, all segments (icons,
The department responsible for the development of the digits, etc.) have to be specified and “frozen” during the
eco®, normally develops mechanical products, such as electronics development. The cost of the chip required to
conventional radiator thermostats, and the eco® is the drive the display depends on the number of segments it
first of its kind for this department. Since the department has to control. Moreover, segment displays are tailor-
did not have all the necessary expertise in-house, the made, and once such a display is made it is very costly to
internal development team had to collaborate with differ- change it. Therefore, defining the (minimum) number of
ent internal and external partners. Examples of external segments required to make up all the possible screens and
partners in this project are users, usability consultants, finding the right layout with appropriate icons on the right
and IDCs, and examples of internal partners are other scale were an important objective. For this reason, the
departments in the wider Danfoss Heating Solutions interactive prototype had to be very detailed with regard
organization with expertise in software or electronics. to graphics and be on the right scale, without the high cost
Each partner, both internal and external, was located at a of changing the segments. To do this, a touchscreen PC
different city, and this was the biggest managerial chal- running a virtual prototype of the interface scaled to the
lenge according to the RDPM: real dimensions was used, which enabled valid tests on
The biggest challenge is to manage a project where the the legibility of the icons in usability tests and the ability
resources are on three different geographical places: in to change the virtual segments if necessary at low cost
the UK, and in Denmark, a couple in Vejle. That was the (see Figure 1). This is a different type of prototype than
biggest one, the space between us. usually used during development, as the DLS explains:

And although it started as a regular project, it rapidly Usually when we talk about prototypes, then we are much
expanded as the project gained momentum, and grew to further in pure hardware terms before we can call it a
be the biggest project in the organization as the RDSD prototype. So it is perhaps the final display we sit and
explains: play with, which then gives us a lot of limitations,
because now we have this display and we cannot go
When we started the project, we did not think it would be back. So that is where the value really kicks in, that we
any more important than any other projects, but right have something that resembles reality early on.
now it is by far the biggest and most important project in
the whole Heating Solutions organization, also including
electronics. It is by far the biggest project. Looking at the Findings
cost and investment in the project, but also looking at the
market scope and money we will spend on communica- Reflection-in-Action: Developing a Collaborative
tion in the market and the interest from top management. Prototyping Approach
I think that is one of the first projects where they are
linked very close to our timeline and looking into all Designing a collaborative prototype. An important
details. challenge in this project was to design a user interface on
a small product surface, which is easy to use, yet offers a
Very early in the project, it was planned to do several high level of functionality to the end user, such as the
usability tests of the interface, and it was clear that the ability to fully customize a week schedule. The design of
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
750 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

Figure 1. The First Virtual and the Final Physical Prototype

the user interface consisted of three stages that are Moreover, an export tool was built into the virtual proto-
described in this paper, each with particular challenges type to export a picture of the current screen with a single
and stakeholders the development team had to relate to. key press, to support effective communication as most of
Finding the limits of how many features can be offered the IDC’s work was done remotely (cf. Horst, 2011).
without making the interface too complicated and clut- Thus, it was the active engagement of various stake-
tered was important in the early stage when the exact holders with the prototype—in line with the principles of
feature-set was specified. Important stakeholders at this “collaborative prototyping” (Terwiesch and Loch, 2004)
stage were PMs, the product portfolio director (PPD), and and “experience prototyping” (Buchenau and Fulton
the RDSD. When the feature-set was defined, the chal- Suri, 2000)—that enabled an efficient overall prototyping
lenge was to optimize the interface to ensure the usability process.
of the product. The external usability consultants—the
senior user experience manager (SUEM) and user expe- Balancing functionality and usability in a
rience consultant (UEC)—who conducted several usabil- collaborative prototyping workshop. Because of the
ity tests were key stakeholders in this stage. Furthermore, importance of the user interface within the collaborative
fine-tuning the interface details after a useable interface prototyping of this product, the tension between the
was developed was important in the last stage of the number of features and the usability soon became appar-
interface design to focus on the use experience beyond ent (Keijzers et al., 2008; Rust et al., 2006). Building on
ease of use. In this stage, end users were the main the main features that were developed in several work-
stakeholders. shops before the IDC joined the project, an initial version
In terms of developing an active involvement of the of the prototype was made based on 13 click-thru sce-
relevant stakeholders in this project, it turned out that narios describing those features. Participants in the first
flexibility and communicability were two important collaborative prototyping workshop were the IDC, the
issues in developing the specific prototyping approach. DLS, an innovation intern, two design engineers (DE1
That is, the IDC designed the prototype with flexibility in and DE2), the communication technology manager
mind, so changes could be made rapidly, allowing for (CTM) and marketing engineer (ME) from the develop-
many design iterations—cf. Verganti (1999) on planned ment team, the R&D director (RDD), the RDSD, the
flexibility, and Thomke (1997) on flexibility in NPD. In PPD, and two external usability consultants (SUEM and
fact, it was so flexible that the IDC could change it on the UEC). Using the prototype triggered a discussion about
fly during collaborative prototyping sessions, going whether there were enough features to distinguish the
through the four steps of the informal prototyping product from competitors, or too many features that made
process; participants could try (run), evaluate (analyze), the interface overly complicated.
give input, and make suggestions (design), which the IDC By seeing the user interface on the real scale, the
implemented (build) in an iterative process until the par- trade-off became very concrete, especially for more
ticipants were satisfied. Second, the IDC designed sup- advanced features, such as full customization of the
portive tools to be used together with the prototype, by schedule:
both the IDC and others. Two of these tools were designed
to make changes to the prototype without coding: the first DE1 Now we have the possibility to have several
to edit basic parameters, such as blink frequencies and different periods, but we aren’t able to have dif-
time-outs, and the second to edit the virtual segments. ferent temperatures in those periods?
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 751
2014;31(4):744–764

DLS No, that is something we decided not to do. CTM I think that the user interface as it is now, with
DE1 Can our competitors do that? all the functions that we desire, yes then it gets
PPD Yes. Yes, that is something we have to include. complicated. But if we could reduce the func-
DLS So we have to be able to, on the size of a 10 tion list, that would help a tremendous amount.
kroner [coin with diameter of ca. 25 mm], to
choose a flexible temperature change for every The important role of the prototype in finding a
single period? balance between functionality and usability is confirmed
RDSD So if I understand what DE1 is saying, you want by a later reflection of the DLS:
to be able to put it down to 17 degrees on
Monday evening and to 18 on Tuesday evening,
This [prototype] has also been used to set the boundaries
and . . .
for the basic specification during its development. . . . It
DE1 To have it on 23 degrees during one day, and 22
has also been used like some sort of evidence for when
on the next, and so on.
we have to stop to put features into this product. Because,
DLS What adds value? Does it add value to have a
again, we can see at an early stage: “Well, we will never
completely clogged user interface? We just have
ever get this particular feature up and running with this
a set-point [normal temperature] and then a
screen size and this number of buttons.” So there it has
set-back temperature [energy saving tempera-
been really strong. . . . There are many funny and wacky
ture] that you can change.
ideas for such an interface. “Why can’t we do this, and
UEC A competitive parameter is also user-
why can’t we also do that too?” And so on. But again, it
friendliness, not just the amount of features.
is enormously powerful to be able to prove why we do it
ME I think it is much more important to have the
like this, relatively easy. . . . With the pressure and the
freedom to set the period from this time to this
[feature] requests that have come from marketing I feel
time where we should lower the temperature.
we have struck, or found, the balance fairly well. [The
DLS Of course it is possible, and we have also played
balance] between what can and what cannot be done. It
with it, but there is a significant difference in the
could easily have gone to the wrong side without this
user interface, if we include the temperature
tool. There have been really many feature requests, both
parameter in programming the schedule.
from the market and from marketing.
DE1 So when they are setting the schedule they think:
“Why can’t I do this?” Our competitors can.
DLS Yes, but does it add value? Not only the function list but also how these functions
SUEM In any case there is the problem with this trade- should be available on the product was discussed. One
off on the user-friendliness. It means that if you particular example is the “mounting” function. To put the
have to compromise the user-friendliness, you eco® on the radiator or to take it off, the motor has to
could say, that you can’t even reach that func- rewind the regulating pin to be able to screw it on/off.
tionality. You could say it has a price. Everybody agreed that this function should be in the
DLS It doesn’t need to be worse, just because it is product, but who should access this function was a point
different [from our competitors]. We can argue of discussion:
for why we are doing it the way we do.
DE2 On the other side, we always hear from market-
This led to some reflection on the amount of functions ing that the end-user shouldn’t be able to take it
the product should have: off so easily. For that they have to use a tool.
That was the idea: they shouldn’t actually do
CTM We have really many functions, and I seriously that.
doubt to what extent these advanced functions CTM Why should they use a tool if they want to take
will be used in really many installations, it off to program it?
because it’s simply too difficult to figure out. We SUEM I think many will feel insecure about taking it
can of course write that we can do these fine off, I mean. If I would be at home . . .
things on the box, which our competitors can UEC I wouldn’t do it either. I’m not a plumber.
also do, so we have a product that matches that. CTM But we have to somehow, no matter what, be
ME We still have to turn the decision when it is out able to put it in that state.
in a shop. Then people have to say: “ok, this UEC But the point is, it doesn’t have to be that intui-
one can do this, and Danfoss can’t do nearly as tive if it is anyway a service [and installation is
much, but it costs the same. Then I will choose done by a trained service mechanic].
the competitor.” DLS It shouldn’t be intuitive.
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
752 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

CTM No, because if it’s intuitive then the end-user SUEM If you’re going through the time, and you make
will also do it. “Oh, what is it doing now? Ok, some mistake, then that’s just too bad? Or can
the radiator is getting really hot . . .” [with the you go back?
regulating pin pulled back the heating valve IDC No, then you have to go to the advanced menu.
will be completely open] (laughing) But it SUEM Then you will have to get your little user
should be there in some way. manual.
DLS We could of course have the ok icon . . .
In this early stage of the development of the product IDC . . . and the back icon there.
and related prototyping approach, there was a clear rec- SUEM You could easily imagine since this is the very
ognition and discussion of the tension between function- first thing they do when they get this one that
ality in terms of features and usability, given the strategic something can go wrong. That they do what I
importance of such issues (Keijzers et al., 2008; Rust just did, that they press next too quickly. And
et al., 2006). The prototype essentially offered a platform then you’re sitting there like: “Oh . . . already
now the first user makes a mistake . . .”
to explain and relate various perspectives on the interface
(laughing)
that enabled cross-fertilization of knowledge (Atuahene-
Gima and Wei, 2011; Carlile, 2002; Nicolini, Mengis, At this point, the involved stakeholders realized that
and Swan, 2012). for the second usability test, starting with setting the date
and time instead of the normal home screen would be
Translating usability problems into design changes better to thereby simulate the user experience after insert-
within collaborative prototyping. In the next stage, the ing the batteries into the product for the first time. But
first usability test was done by the external consultancy, rather than seeing it as something that could go wrong, it
and the DLS and IDC worked on the interface to was more seen as a warm-up exercise—how hard could it
address the usability problems described in the usability be to set the date and time? However, having conducted
report. Decoding the usability report and translating many usability tests, she knew that the first task is impor-
usability problems into design changes were challeng- tant since it can either give the user confidence or insecu-
ing since the development team had not participated in rity for the rest of the test. Without this intervention and
the usability test. Then, the developers organized a the fast update of the screens, this problem would not have
prototyping workshop in which the SUEM participated. been detected until the next usability report. Thus, the
Having conducted the usability test, she knew a lot active engagement of various stakeholders, in this case to
more about the usability problems and how they arose better understand user needs within the context of usabil-
than was (and could be) described in the usability ity, continued to enable the possibility to bring together
report. different perspectives and shortcut some parts of the
An example of an instance where she drew on her prototyping process (Buchenau and Fulton Suri, 2000;
deep understanding of a problem was when she asked if Nicolini et al., 2012; Terwiesch and Loch, 2004).
it was possible to make a certain icon smaller since it Finally, when the end of the workshop was approach-
was “overshadowing” nearby icons. This was easily ing, she also raised the point that although she felt a lot of
done with the graphic editor, but it was difficult to the usability problems were going to be solved, the
evaluate if it was small enough. She then directed the changes might have undesired side effects resulting in
developers to go to the particular screen where the other usability problems:
problem had been most prominent in the test. This
DLS I’m looking forward to seeing the next round.
enabled all participants to see that the icon needed to be
SUEM Me too, definitely. It is hard to anticipate how
even smaller, and a few seconds later the next problem
much will still be needed. We can sit here and
could be tackled.
change this and that, but . . . The things we have
Another example illustrates how she applied her changed now, I’m thinking . . . We are of course
expertise in usability testing during the session. After working towards eliminating problems, but it is
setting relatively complex things like a schedule, the user hard to exclude that what we are changing now
has to go through an extra step to confirm, cancel, or go will introduce new problems. That would be
back. For setting the date and time, it had been decided unfortunate, but that is how it will be, and in the
this was not necessary. When she tried the function, she end it will help to prioritize some things, so they
asked what would happen if she made a mistake, which become more clear. Those are the trade-offs,
was the start of the following discussion. right.
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 753
2014;31(4):744–764

IDC Then we also know what the focus should be in EU1 If I make a mistake, for example on Thursday.
the user manual, what they can do without and That I set it too early, and on a weird time
what they need the manual for and show it period, so it isn’t quite right, and I want to do it
clearly and step-by-step how to. differently, that I then come back to Thursday.
That is nicer. (comes back to “Thursday” and
hits enter) Ah . . . this is confusing . . .
The role of the prototype in this setting is referred to as IDC Why?
a “sketching tool” by the DLS when discussing this par- EU1 Because now, look . . . So Thursday, Thursday,
ticular workshop in a later reflection: see, this one isn’t right. So if I click here, I want
to return to Thursday. Yes. But then I have to do
It is always exciting to get things tested. It is a difficult this again (press down), and then I can set it
process, but also incredibly healthy. It has really given so again. So just before I clicked too quickly, and
much, it has. And one could say that, with the way in then I jumped back too many steps.
which we have used the tool as a sketching tool in the IDC The reason you can’t stay on that day is because
beginning, it has also been an easy and fast process in otherwise you wouldn’t be able to come back.
the adaptation towards the tests, and we could also easily So you need an extra step.
correct things between the different tests. We have done EU1 Yes, but this is already nicer . . . Oh no, wait . . .
multiple sessions of these end-user tests, and so the three See, if I for example make a mistake here, then I
of us could sit together and update it quickly in between have to go back to this one. Yes. But actually it’s
the tests. So from the one test to the next relatively little not so much easier [than before].
time passed. IDC So would it be easier . . . So, would it be more
logical if you return here . . . (swaps two lines of
It is, therefore, clear that even though the prototype code) That then the schedule is the default?
helps create a climate of problem solving and collabora- EU1 Yes.
tion (cf. Brown, 2008; Ulrich, 2011), there are also con- IDC I’ll quickly go through this. Can you try it now?
straints in terms of creating a full understanding of the If it is more logical?
user experience. One of the lessons here is that the col- EU1 Thursday, change it, because I want to go to bed
laborative prototyping approach, which at this particular early and I make a mistake. Then if I want to
part of the development process only used a usability come back, yes, that I’m directly on the clock, so
report and expert to represent the users’ perspective, can I can adjust it again. Yes, I like this better.
mostly shortcut the front end of the prototyping process, IDC Okay.
namely design and build, but the run and analyze steps EU1 I think this makes a big difference.
remain the domain of actual use (cf. Thomke, 1998).
The user and prototyper together make sense of why
Detecting emerging usability problems through active the change confused him, whether the change is an actual
engagement and experimentation. After the second improvement, and how the negative side effect can be
usability test, a number of problems were identified that addressed. In the course of a few minutes, these problems
were undesired side effects of some of the changes—as are effectively solved. What is noteworthy is that in the
inherent to any design process. In order to explore final design of the interface, there are several (minor)
whether collaborative prototyping during the run step elements that can be traced back to these workshops—
(usability testing) could help in detecting these side that is, users made contributions to product details.
effects immediately, end users were involved in Another aspect of doing prototyping in the run phase is
prototyping workshops. The IDC ran four such work- that the revised prototype embodies the input from users
shops with potential end users, and the DLS also ran 10 on the usability. The DLS compared receiving an updated
workshops, using the supportive tools to make modifica- prototype with receiving a usability report:
tions to the prototype.
During the first workshop by the IDC, there was an In that case [of getting a usability report] I can also
instance of an emerging usability problem as a side effect misunderstand the words, and I have to get in and inter-
of a change suggested by the end user (EU1). The change pret these words in such a report myself and develop my
related to which icon should be the default one to be own impression of this user interface. But when I am
selected when coming back from the schedule function provided with this piece of software, for example from
after making a mistake. The IDC made an update and [the IDC]; [the IDC has] done such a test, and then this
EU1 explains what he is doing: modified piece of software comes back. Well then I can
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
754 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

directly relate to it and see the result of this test. In that organization. After further analyzing the three types of
way you save a lot of, firstly time, but also misunder- prototyping as already described above, the remainder of
standings, which can happen with a report. this section presents our general analysis in order to
further conceptualize the main elements and issues within
The DLS, who had no background in interactive our collaborative prototyping approach.
prototyping, used the supportive tools, and especially the
tool to edit dynamic settings, in collaboration with dif- Three cases of collaborative prototyping. Our study
ferent internal users. In a later interview: embeds three cases of the use of the collaborative
prototyping approach within Danfoss Heating Solutions.
The good thing about this tool is that you can try things In each case, the prototype was brought in for a different
out immediately. Especially with these “soft-coded” purpose, based on the practice problem at hand, and was
things, where you can very easily change a parameter staged in a different way to involve the relevant stake-
and test it right away. . . . Things we have played around
holders, which may be within or outside the design team
with a lot are the default values in it: blinking frequen-
and organization. Table 2 gives an overview of the cases,
cies, time-outs and such things. You call it participation
describing the role of the prototype, the practice problem,
workshops, where you test directly with a user and
correct immediately, until you reach a satisfying result.
research problem, solution concept, specific solution, and
. . . As a developer you lose the feeling for those param- outcome of each of the interventions (cf. Andriessen,
eters, time-outs and frequencies and things like that. So 2007). As such, it gives an overview of the types of
there it was very easy to go in and find the values, by problems for which a collaborative prototyping approach
being able to adjust them until the end-user says: “Now can be used, and who to collaborate with.
it is good.” . . . So it is of course about being able to In the first case, the prototype played an instrumental
correct things on-site, directly, instead of having to write role in negotiating the features by making the trade-off
something down, and tomorrow you have forgotten what between functionality and usability very concrete. This
it was you had to correct and what it was he said, and enabled the participants to prioritize the features that
these kinds of things. So to get it adapted to the test were most important to have and leave out the features
person you are sitting with, immediately, has also sped that added a lot of complexity but little value for the
up the process tremendously, and moreover you get consumer. Exposing the prototype to a broad range of
everything. stakeholders from the different departments involved in
the project at this early stage was important not only to
It, thus, became clear that the involvement of (in this communicate (as in Mascitelli, 2000) but also for them to
case) a possible end user was essential in extending the get ownership over the prototype as they saw their sug-
collaborative prototyping approach to all phases of gestions implemented (Amabile, Conti, Coon, Lazenby,
the problem-solving process (Thomke, 1998). It was the and Herron, 1996; Siegel and Kaemmerer, 1978). The
active engagement of the user that led to a better under- ability to make suggestions that could be implemented
standing of the relevant experience and related design directly in the prototype also played an important role for
parameters (Buchenau and Fulton Suri, 2000; Klemmer the part of the specification that was not yet detailed
et al., 2006), which was enabled by the direct interaction (Boehm et al., 1984), in line with the principles of expe-
with the prototype and the instant changes by the designer rience prototyping (Buchenau and Fulton Suri, 2000).
(cf. Mascitelli, 2000; Terwiesch and Loch, 2004). The Essentially, it created shortcuts in the design and proto-
face-to-face interaction provided an opportunity for type cycles, which speeded up the process because mis-
mutual improvisation to unstick and share relevant understandings were discovered immediately.
knowledge, as well as to increase NPD performance In the second case, the prototype was important as a
(Bogers and Larsen, 2012; Vera and Crossan, 2005). way to gain a deeper understanding of the usability prob-
lems described in the usability report. Involving the
Reflection-on-Action: Exploring usability consultant who conducted the usability test was
Collaborative Prototyping very important because she knew a lot more about the
problems than were described in the usability report. By
Building on the above-presented action research as going through the different screens and making changes,
reflection-in-action, a next step is to build on those find- the developers were able to address the problems
ings and further explore the notion of collaborative (Buchenau and Fulton Suri, 2000). Through the work-
prototyping as it has been developed in this particular shop, she could bring in her tacit knowledge and expertise
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 755
2014;31(4):744–764

Table 2. Overview of the Different Collaborative Prototyping Cases

Characteristic Case 1 Case 2 Case 3

Role of prototype Communication tool Sketching tool Testing tool

Practice problem When specifying the features, the The usability report, which described After the second usability tests, a number
company was afraid that they a number of usability problems, did of usability problems were detected that
would add too much complexity, not provide concrete directions on were side effects of “improvements” that
which would have a big negative design improvements were made to the prototype
impact on usability
Research problem How can the impact of features on How can the usability report be How can undesired side effects and
usability be evaluated when the translated into design emerging usability problems be detected
feature-set is still flexible? improvements? when “improvements” are implemented?
Solution concept Start prototyping before the Involve the usability tester who wrote Involve end users in prototyping during the
feature-set is finalized to talk about the report in prototyping during the run (test) step
concrete features and their impact design step
on usability
Specific solution Prototyping workshop with Prototyping workshop with the Prototyping workshops with end users,
development team, usability developer, interaction design implementing and evaluating design
consultants, and management consultant, and usability consultant suggestions during the workshops
represented
Outcome A revised prototype (and revised A revised prototype addressing the A revised version of the prototype
specification), reflecting the input usability problems detected to be embodying the feedback from the users,
from all stakeholders, which could used in the next usability test which could be shared with the
be used in a usability test development team

(Mascitelli, 2000). However, design changes have side empathy” of the user for the design context. However, it
effects, and it was difficult for the internal developers should also be noted that not all user involvement is
to assess if a change would introduce new usability equally productive. More generally, it became clear that
problems, thus calling for a need to implement the there may be varying capabilities, and thus roles for dif-
collaborative prototyping approach across a larger set ferent users in collaborative prototyping (cf. Terwiesch
of stakeholders and within all phases of the general and Loch, 2004). For example, one end user (EU2), who
problem-solving process (Atuahene-Gima and Wei, is familiar with traditional prototyping, noted that it
2011; Thomke, 1998). requires another attitude and more skills of the user to be
In the third case, the prototype was evaluated with end involved in such a collaborative prototyping approach—
users, and usability problems were addressed as they which could thus be labeled as turning users from
were discovered. By implementing design changes and problem finders to problem solvers (cf. von Hippel, 1994,
evaluating them, emerging usability problems were 2005; von Hippel and Katz, 2002).
quickly identified and solved. The resulting modified pro-
totype proved to be an effective way for the other devel- Conceptualizing the prototyping process. Building on
opers to relate to since they can immediately see the the above cases, the next step is to further analyze how
changes. This is more difficult with usability reports, but our collaborative prototyping approach compares with
it should be stressed that collaborative prototyping with the general notion of prototyping. In the management
users cannot replace regular usability tests, as it is impor- literature, the prototyping process is conceptualized as an
tant to confirm how the changes improve usability for a iterative trial-and-error learning process, following four
broader range of users than the particular user who sug- steps: (1) design, (2) build, (3) run, and (4) analyze
gested it. Moreover, the end user learns more about the (Thomke, 1998; von Hippel, 2005). The results show
design rationale behind the interface through discussion that, based on the interactive and experiential charac-
with the prototyper, which has an influence on their teristics of the collaborative prototyping approach
ability to understand it and the perceived usability of the (Buchenau and Fulton Suri, 2000; Mascitelli, 2000;
system. Therefore, formal usability tests are complemen- Terwiesch and Loch, 2004), the general steps within the
tary to collaborative prototyping sessions with users, and prototyping process happened on two levels within the
the collaborative prototyping approach not only gives the NPD project. First, there is a formal or “managerial”
designer empathy for the user context (Brown, 2008; prototyping process, in which the four steps are separate
Leonard and Rayport, 1997) but also created “reverse activities that can be planned on the project time line.
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
756 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

Figure 2. Managerial Collaborative Prototyping Process (Adapted from Thomke, 1998)

Second, there is an informal or “designer” prototyping liseconds how fast an icon should blink. The prototyper
process, which is more fluid, and the four steps are more then has to go through another, much faster and informal,
difficult to distinguish. prototyping process. The four steps could be to (1) make
In the context of developing a user interface for a an educated guess about an appropriate blink frequency,
product, the managerial prototyping process could consist (2) implement it in the prototype, (3) run the prototype,
of the following steps: (1) designing the user interface, and (4) evaluate if the frequency indeed is appropriate,
(2) making a prototype of the interface, (3) doing a usabil- and repeat from (1) if necessary. In this particular
ity test, and (4) analyzing the results from the test, and example, these steps can take a few seconds each, and the
repeat from (1) if necessary (see Figure 2). What could be prototyper is likely not aware of following a four-step
referred to as the designer prototyping process is what process, but engages in “reflection-in-action” (Schön,
happens at the operational level during prototyping when 1983), or what Klemmer et al. (2006) call “thinking
the involved stakeholders may not be aware of which step through prototyping.”
of the prototyping cycle they are active in. Table 3 provides an overview of the different kinds of
For example, imagine that a prototyper gets a specifi- learning that happen in the managerial and designer
cation of the user interface as the outcome of the design prototyping processes, and indicates who is the learner,
step and has to implement it (normally in the build step). who is a stakeholder with an interest in what is learned,
However, often the specification does not contain all the and how long each trial-and-error learning cycle typically
details necessary for straightforward implementation. An takes in the context of the design of a product interface.
early specification may, for instance, not specify in mil- Moreover, Table 4 gives an example of how the

Table 3. The Managerial and Designer Processes in Collaborative Prototyping

Characteristic Managerial Prototyping Process Designer Prototyping Process

Learning Usability and performance of the product interface Familiarity with the design space, constraints, and implications of
decisions on usability
Knowledge Typically codified (e.g., usability report) Typically tacit or sticky (costly to transfer)
Learner Development team Prototyper(s)
Stakeholders Management Development team and management
Full learning cycle Few weeks per iteration Few seconds/minutes per iteration

Table 4. Exemplifying the Prototyping Process and Practice

Prototyping Process Prototyping Practice

Phase Build Phase Run Phase

Design Develop the interface specification Can we make this icon smaller? Can we change the order of these screens?
Make a document with suggested changes
and improvements
Build Implement the specification or changes into Make the icon smaller Change order of screens
a simulation of the interface
Run Conduct a usability test with end users Run the simulation, go to a specific Run the simulation, go through scenario
Run an internal test screen
Analyze Identify usability problems and document Assess if the icon is the right size Assess if the change is an improvement/
them in a usability report eliminates the problem
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 757
2014;31(4):744–764

Design and interview—one finding is that our collaborative

prototyping approach especially enables communication
and collaboration across different kinds of boundaries. An
important element in this is that the prototype serves as a
communication tool in which direct interaction is possible,
which facilitates the informal practice of prototyping.
In line with Carlile’s (2002) pragmatic view of knowl-
Collaborative edge and boundaries, the prototype, as a communication
Analyze prototyping Build
tool, acts as a boundary object that can be used to repre-
sent, understand, and transform knowledge across bound-
aries. In our case, these boundaries can be between
functional areas, and between the organization and exter-
nal stakeholders, which are all important to facilitate
NPD and innovation performance (Atuahene-Gima and
Wei, 2011; Bogers and Lhuillery, 2011; Leonard-Barton,
Run
1995). In some cases, the prototype also enabled commu-
Figure 3. Designer Collaborative Prototyping Process as Col- nication and collaboration across hierarchical levels. As a
lapsed Problem-Solving Cycle
platform for collaboration, collaborative prototyping thus
develops and integrates knowledge that can serve as a
prototyping steps become fuzzy when going down to the source of innovation within NPD, although the actual
more informal designer level, which deals with the actual design will determine the facilitating or hampering nature
prototyping practice. of the platform (Brown, 2008; Carlile, 2002; Ulrich,
In essence, the collaborative prototyping approach 2011). As such, it not only directly triggers cross-
developed in this project activates the informal disciplinary collaboration and facilitates work across
prototyping process, which then brings together a wide different boundaries, but moreover provides the basic
variety of stakeholders who interact and collaborate in a infrastructural support of collaboration, thereby address-
way that shortcuts the general (managerial) prototyping ing the different roles of objects in cross-disciplinary
cycles (Atuahene-Gima and Wei, 2011; Mascitelli, 2000; collaboration (Nicolini et al., 2012).
Terwiesch and Loch, 2004). Thus, also in line with the Given the general difficulties to transfer and develop
principles of experience prototyping (Buchenau and innovative knowledge across boundaries, our approach
Fulton Suri, 2000), our collaborative prototyping to collaborative prototyping provides an explicit way
approach creates a direct and active involvement of to overcome interpretive barriers in NPD (Dougherty,
various stakeholders in which the respective (discrete) 1992), for example by direct feedback and face-to-face
prototyping activities are collapsed into an ongoing and interaction in relation to proposed solutions (Mascitelli,
almost continuous prototyping process. Figure 3 shows a 2000). In particular, it was the active engagement of the
representation of how collaborative prototyping serves as various stakeholders that led to a better understanding of
a hub among the design, build, run, and analyze activities. the relevant experience and related design parameters, as
The traditional or managerial prototyping cycles are in line with experience prototyping (Buchenau and
hereby made less discrete as there is an (almost) continu- Fulton Suri, 2000), while the high fidelity enables the use
ous iteration among these activities through the collab- of the prototype as an interactive tool for development,
orative prototyping approach. marketing, and sales (Rudd et al., 1996). In a sense, the
collaborative prototyping platform created a local com-
Collaborative prototyping as a platform for munity of practice (Brown and Duguid, 1991) centered
collaboration. The actual practice of collaborative around the prototyping activity, and the interface devel-
prototyping at the informal level of the problem-solving opment in particular, which in turn created a collective
process, as described above, involves an approach in transformation of understanding through a process of
which the actual prototype acts as a platform for collabo- shared meaning and sense-making (Bechky, 2003;
ration. As a prototype-driven approach to problem solving Weick, Sutcliffe, and Obstfeld, 2005).
within NPD, the results show a number of themes
that become relevant in this context. As shown in Windows of opportunities within collaborative
Table 5—which provides evidence from the workshops prototyping. The practice of collaborative prototyping
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
758 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

Table 5. Collaborative Prototyping as Platform for Communication and Collaboration

Element Evidence from Interviews Evidence from Workshops

Prototype as DLS At a very early stage . . . it has been a really SUEM If you’re going through the time, and you
communication strong tool, where we were able to exploit make some mistake, then that’s just too
tool the flexibility of it, precisely by using it bad? Or can you go back?
early on. In addition it was used in the IDC No, then you have to go to the advanced
whole communication part, which turned menu.
out to be very important. And we have SUEM Then you will have to get your little user
used it a lot as an internal facilitator to get manual.
people to get a thorough understanding of DLS We could of course have the OK icon . . .
what concept this is, and what kind of IDC . . . and the back icon there.
product it is, also at a very early stage. SUEM You could easily imagine since this is the very
RDPM It is always good to use tools and first thing they do when they get this one
information that are as easy to understand that something can go wrong. That they do
as possible. If I had written this in what I just did, that they press next too
words—you know, one picture tells more quickly. And then you’re sitting there like:
than a thousand words. That was very “Oh . . . already now the first user makes a
good to use that as communication mistake . . .” (laughing)
material.
Cross-functional RDSD It has also been used as a communication DE1 Now we have the possibility to have several
communication tool towards our sales people, sales and different periods, but we aren’t able to have
marketing. So I am sure that we have different temperatures in those periods?
saved both money and time in this project DLS No, that is something we decided not to do.
using this tool. It has been involved in so DE1 Can our competitors do that?
many different parts of the project. So it is PPD Yes. Yes, that is something we have to
not only to settle the Man-Machine include.
Interface, but also as documentation in DLS ...
different ways. RDSD So if I understand what DE1 is saying, you
DLS We presented the ideas, what kind of ideas want to be able to put it down to 17
we are going for now, to practically all degrees on Monday evening and to 18 on
departments that are involved in this Tuesday evening, and . . .
project. So they got an understanding of it
DE1 ...
and had the opportunity to provide
feedback early on. DLS What adds value? . . .
UEC A competitive parameter is also user-
friendliness, not just the amount of features.
ME I think it is much more important to have the
freedom to set the period from this time to
this time where we should lower the
temperature.
DLS/DE1 ...
SUEM In any case there is the problem with this
trade-off on the user-friendliness. . . .
Collaboration across RDPM In the beginning, the purpose of using the EU1 . . . if I for example make a mistake here, then
organizational prototype was to develop a user-friendly I have to go back to this one. . . .
boundaries user interface. That was in the beginning IDC . . . So, would it be more logical if you return
of this prototype, but because this here. . . [swaps two lines of code] . . .
prototype was so good and excellent we EU1 Yes.
used it in different ways. Then we used it IDC I’ll quickly go through this. Can you try it
also for communication towards now? If it is more logical?
customers; possible, potential future EU1 Tries the updated prototype Yes, I like this
customers. And we also used it for how to better.
do the instructions, user guides.
DLS Things we have played around with a lot are
the default values in it: blinking
frequencies, time-outs and such things.
You call it participation workshops, where
you test directly with a user and correct
immediately, until you reach a satisfying
result.
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 759
2014;31(4):744–764

Figure 4. Time Line of Prototyping Iterations at the Managerial Level (6 Months)

—at the designer level—which thus serves as a platform invitations for change, typically preceding the iteration,
for prototype-driven problem solving, can be described as while in some cases the iterations are done leading up to
a process in which the traditional phases within problem a certain event.
solving (see Figure 2) are collapsed into an almost con- What is then interesting to observe is that the iterations
tinuous iteration that blurs the boundaries between those within a short period of time; that is, within one
phases (see Figure 3). In this process, the collaborative of the windows of opportunities, become much less
prototyping approach brings together a wide variety of discontinuous—as in line with our proposed process of
stakeholders who interact and collaborate to solve rel- collaborative prototyping (see Figure 3). Accordingly,
evant problems in the early stage of NPD (cf. Atuahene- Figure 5 shows the cumulative iterations within a specific
Gima and Wei, 2011; Mascitelli, 2000; Terwiesch 2-hour period (on August 25, 2009) as an indication of
and Loch, 2004). However, the highly interactive and the more continuous or stepwise process, as enabled by
collapsed/continuous nature of this prototyping process the collaborative prototyping approach in which the rel-
does not mean that all stakeholders are solving problems evant scale is minutes or even seconds rather than days or
and adjusting the prototypes all the time. In fact, by weeks (cf. Dey et al., 2001; Rudd et al., 1996). Thus,
relying on several sources—including prototype while the designer level of prototyping essentially entails
iterations, event descriptions, e-mails, and other a continuous problem-solving cycle, the managerial level
documentation—it was possible to construct a history of of the prototyping project is in fact more discontinuous.
interactions and iterations related to the actual prototype,
which shows that the overall process, at the managerial
level, is highly discontinuous. Conclusion
Figure 4 shows a time line with an overview of all
iterations that were made in the course of a 6-month This paper presented a particular approach to collabora-
period within the project. These iterations comprise tive prototyping, which involves the active engagement of
changes in programming code, structure of the interface, various stakeholders across functional, hierarchical, and
and graphics that are part of segment display. As the organizational boundaries. Our combined action research
figure shows, the iterations tend to be performed in and case study of the development of a radiator thermo-
narrow time windows, separated by longer periods of stat at Danfoss Heating Solutions provides a rich descrip-
limited or no iterations—much in line with Tyre and tion and analysis of how collaboration across various
Orlikowski’s (1994) finding of “windows of opportuni- boundaries was needed in order to successfully develop
ties” within the process of technological adaptation. the product. The action research approach was embedded
Most of these “windows” were triggered by certain in the challenge to design a user interface on a small
events or interactions. The top of Figure 4 (above the product surface, which is easy to use but also offers a high
chart) shows the main events and interactions—such as level of functionality to the end user. The design of the
workshops, tests, and intensive e-mail exchange—that user interface consisted of three stages, which serve as
are often linked to a chain of iterations as triggers or three cases of collaborative prototyping practices, each
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
760 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

Figure 5. Cumulative Prototyping Iterations at the Designer Level (2 Hours)

with particular challenges and involved stakeholders, fol- problem-solving cycle is collapsed into an ongoing,
lowed by a further analysis of the overall case. almost continuous process. This approach of colla-
In terms of developing an active involvement of the borative prototyping was used in the NPD project by
relevant stakeholders in this project, it turned out that organizing workshops that facilitate involving various
flexibility and communicability were two important stakeholders across functional and organizational bound-
issues in developing the specific prototyping approach. aries. In particular, it was the active engagement of
Therefore, a prototype was developed that allowed for various stakeholders with the prototype that enabled an
rapid changes and many design iterations (cf. Dey et al., efficient overall prototyping process, in line with the prin-
2001; Thomke, 1997; Verganti, 1999). Combining per- ciples of “collaborative prototyping” (Terwiesch and
spectives within design and management research led Loch, 2004) and “experience prototyping” (Buchenau
to the conceptualization that breaking up the formal and Fulton Suri, 2000). As such, our collaborative
prototyping process into discrete steps, which are done by prototyping approach enabled a shift from specification-
different experts (e.g., a prototype that is designed by driven prototypes to prototype-driven specification.
interaction designers, built by design engineers, and The prototype essentially offered a platform to explain
tested by usability experts), is problematic because it is a and relate various perspectives on the interface that
learning process in which not all the learning can be enabled cross-fertilization of knowledge (Atuahene-
easily coded and subsequently decoded from the outcome Gima and Wei, 2011), and thus effectively acted as a
of each step (e.g., a specification, prototype, or usability boundary object for cross-disciplinary collaboration
test). For example, separating making from evaluation in (Carlile, 2002; Nicolini et al., 2012). As such, the collab-
the prototyping process is problematic because the design orative prototyping approach also enabled the translation
rationale of the prototyper may be hard to decode by the of usability problem into actual design changes. The
evaluators, and the feedback that the prototyper receives active engagement of various stakeholders, in this case to
through evaluation may be hard to implement since the better understand user needs within the context of usabil-
evaluators are not aware of the concrete constraints the ity, continued to enable the possibility to bring together
prototyper has to deal with (cf. Leonard-Barton, 1988; different perspectives and shortcut some parts of the
von Hippel and Tyre, 1995). prototyping process (Buchenau and Fulton Suri, 2000;
Our findings suggest that involving these experts in Klemmer et al., 2006; Nicolini et al., 2012; Terwiesch
the other steps of the problem-solving process and intro- and Loch, 2004).
ducing an element of prototyping, which allows going More generally, the prototype helps create a climate of
through full trial-and-error learning cycles, is a way to problem solving and collaboration (cf. Brown, 2008;
share this tacit knowledge and integrate it into the devel- Ulrich, 2011), although some constraints in terms of cre-
oping prototype (Mascitelli, 2000; Thomke, 1998). By ating a full understanding of the user experience were
focusing on the actual practice of prototyping, another also identified, while the knowledge and skills of the user
finding that emerged was that the ordinary prototyping to act as a problem solver should also be considered (cf.
process mostly refers to a managerial level, whereas our Danneels, 2003; von Hippel, 1994, 2005). For example,
approach activates a designer level in which the entire the opportunities and constraints of users developing
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 761
2014;31(4):744–764

“reverse empathy” for the design constraints—parallel to is relatively easy to involve, for example, a large team in
the designer developing empathy for the user context— a brainstorming session in a build step, or do an internal
should be considered. One of the lessons is that the col- evaluation of the product with top management in a run
laborative prototyping approach can mostly shortcut the step. This is more difficult in the designer prototyping
front end of the prototyping process (design and build), process because it happens fast and requires technical
but that the later run and analyze steps largely remain expertise to be involved. However, in the designer
the domain of actual use (Terwiesch and Loch, 2004; process, many small decisions are made on the exact
Thomke, 1998). Therefore, end user involvement is an implementation, and the prototyper learns what the impli-
important element in extending the collaborative prototy- cations are of design decisions, and becomes familiar
ping approach to all phases of the problem-solving pro- with the design space and constraints. This knowledge is
cess (cf. Bogers et al., 2010). The case, thus, shows that “sticky” in that it is not easily transferred by exposing
the active engagement of the user leads to a better under- others to the prototype that comes out of this process (von
standing of the relevant experience and related design Hippel, 1994). In line with Mascitelli (2000), face-to-face
parameters (Buchenau and Fulton Suri, 2000; Klemmer interaction between relevant stakeholders in NPD can
et al., 2006), which was enabled by the direct interaction enable creative improvisation and real-time knowledge
with the prototype and the instant changes by the designer sharing. Involving others in the informal prototyping then
(Mascitelli, 2000; Terwiesch and Loch, 2004). would be a way for them to learn about the design space,
The continued analysis, moreover, shows that despite constraints, and implications. This could be done either
the continuous nature of the (designer) practice of by involving them in this process during the build step, or
prototyping, there are certain windows of opportunities by introducing a degree of prototyping in the design or
(at a more managerial level) during which the collabora- run steps since the informal process consists of the same
tive prototyping approach actually leads to changes in the four steps. Prototyping could be a part of, for example,
product design (cf. Tyre and Orlikowski, 1994). Thus, team brainstorming sessions or evaluation sessions.
while the designer level of prototyping essentially entails The early phase of the NPD project in our study shows
a continuous problem-solving cycle, the managerial level that there is always a balance between the specification
of the prototype and NPD project is in fact more discon- and the prototype in the sense that there were a number of
tinuous, triggered by certain events of interactions that features that were set at the outset, but otherwise many
act as invitations for change. design parameters were left open (cf. Boehm et al.,
1984). Starting prototyping before and during the devel-
Implications opment of the specification is especially relevant for
products with user interfaces. Prototype-driven problem
Our results have several implications for prototyping solving allows the evaluation of the usability of an inter-
theory and practice. Most generally, our conceptualiza- face, and makes the trade-off between adding features
tion of the prototyping process at different levels, com- and adding complexity very concrete. This can help in
prising a managerial and designer level, brings together specifying a feature-set, which does not only look good
literature from both design and management to work on the box, but also adds most value for the customer in
toward an integrative model of the prototyping process. use. Another aspect of product interfaces is that they often
In particular, the prototype-driven problem-solving include dynamic aspects, such as how fast icons blink or
process as an informal practice implies an important role how long a backlight stays on after the last interaction.
of the active engagement of various stakeholders, thereby Although it is possible to estimate appropriate times for a
integrating the principles of “collaborative prototyping” blink frequency or a backlight time-out, finding a good
(Terwiesch and Loch, 2004) and “experience proto- value is often a trial-and-error process. When the proto-
typing” (Buchenau and Fulton Suri, 2000), as well as the type is used to help define the feature-set, it will be
more general principles of prototyping (Schrage, 1996; important to not only involve stakeholders, such as
Thomke, 1998; Ulrich and Eppinger, 2008) and (partici- interaction designers and usability experts, but also for
patory) design (Hartmann, 2009; Leonard and Rayport, example marketers and the PPD.
1997; Schuler and Namioka, 1993; Ulrich, 2011). Our study, moreover, has important implications for
Making a distinction between the designer and mana- managers and academics alike who are interested in col-
gerial prototyping process is important because there is a laborative prototyping as a way to effectively involve not
difference in how internal stakeholders can be brought only external stakeholders such as users (Bogers et al.,
into the process. In the managerial prototyping process, it 2010; Poetz and Schreier, 2012; Terwiesch and Loch,
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
762 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

2004) but also internal stakeholders across hierarchical and Bogers, 2013). As such, the focus on external stake-
and functional boundaries (Adler, 1995; Buur and holders could be extended beyond a user-centric perspec-
Matthews, 2008; Song et al., 1998). Furthermore, it is tive (e.g., Baldwin and von Hippel, 2011; von Hippel,
important to both separate and link the formal managerial 2005) to include the broader value network or ecosystem
level of prototyping and the informal practice of (Adner and Kapoor, 2010; Bogers and West, 2012;
prototyping, where the actual prototype acts as an impor- Laursen and Salter, 2006). Our multistakeholder
tant boundary object for communication (Carlile, 2002; approach is embedded within the concept of experience
Nicolini et al., 2012). Collaborating with various experts prototyping, which enables involved stakeholders to gain
in different steps of the (formal) prototyping process in first-hand appreciation of existing or future conditions
prototyping workshops enables the different participants through active engagement with prototypes (Buchenau
to unstick their knowledge, which is directly imple- and Fulton Suri, 2000; Klemmer et al., 2006). This
mented in the evolving prototype. However, our study approach to prototyping, as a way to design an integrated
also shows some constraints in prototype-driven problem experience to better understand, explore, or communicate
solving, while our finding of windows of opportunities at what it might be like to engage with the product that is
the level of the prototype project emphasizes the triggers being designed, could be further explored in future
or invitation for change, as well as that the periods of research. Moreover, a particular finding in this context is
actual change are characterized by more continuous that the collaborative prototype integrates different roles
adaptation—thus reinforcing the need to consider differ- of boundary objects (Carlile, 2002; Nicolini et al., 2012),
ent levels of the problem-solving process. which offers a basis for further exploration of the mutual
This approach is most relevant for managers in NPD benefit of collaborative prototypes and boundary objects.
projects where prototyping starts at an early stage, and On the other hand, there may be value in applying
who are managing a cross-functional team (Song et al., some of our finding to a number of established or
1997; Ulrich and Eppinger, 2008). Our study provides growing research streams. For example, different areas of
problem classes that this approach successfully addressed design may be affected by a better understanding of col-
in this particular case, although the results from this case laborative prototyping, including product design and
study are not yet “field-tested and grounded technological organization design (Tushman and Nadler, 1978; Ulrich,
rules” (van Aken, 2004). In the terms of van Aken (2004), 2011), also considering the notion of modularity
this project is at the stage of α-testing where the solution (Baldwin and Clark, 2000; Sanchez and Mahoney, 1996).
concept has been tried in multiple prototyping cases Moreover, it may be useful to integrate our conceptual-
within a single company context. In subsequent research, ization of collaborative prototyping into a complementar-
the technological rules should be applied to practice ity perspective on the interface between intra- and
problems in different cases and by third parties to interorganizational sources and processes of innovation
do β-testing until theoretical saturation is reached (Bogers and Lhuillery, 2011; Hillebrand and Biemans,
(Eisenhardt, 1989) to establish field-tested and grounded 2004; Neyer, Bullinger, and Moeslein, 2009). Finally, our
technological rules. perspective on collaborative prototyping could, in some
More broadly, out inductively derived framework way, inform how organizations build innovation capabili-
could be further developed and applied within the context ties and routines at various interfaces, in line with the
of a number of theoretical frameworks and perspectives. recent calls for more research into the microfoundations
On the one hand, there are a number of perspectives that of such processes (Barney and Felin, 2013; Felin and
are the basis for our analysis, which could be further Foss, 2005; Lewin, Massini, and Peeters, 2011).
elaborated in future research. For example, our study
extends Terwiesch and Loch’s (2004) definition of col-
laborative prototyping by not only focusing on the user–
References
producer interface but also developing a more general Adler, P. S. 1995. Interdepartmental interdependence and coordination: The
multistakeholder view on collaborative prototyping, case of the design/manufacturing interface. Organization Science 6 (2):
147–67.
which can include any stakeholder from within and
Adner, R., and R. Kapoor. 2010. Value creation in innovation ecosystems:
outside the organization. As such, it would be useful to How the structure of technological interdependence affects firm per-
link our collaborative prototyping approach to perspec- formance in new technology generations. Strategic Management
Journal 31 (3): 306–33.
tives that deal with the involvement of a variety of stake-
Amabile, T. M., R. Conti, H. Coon, J. Lazenby, and M. Herron. 1996.
holders in NPD within more open innovation models Assessing the work environment for creativity. Academy of Manage-
(Chesbrough, 2003; Dahlander and Gann, 2010; West ment Journal 39 (5): 1154–84.
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
COLLABORATIVE PROTOTYPING J PROD INNOV MANAG 763
2014;31(4):744–764

Andriessen, D. 2007. Combining design-based research and action research Dougherty, D. 1992. Interpretive barriers to successful product innovation
to test management solutions. 7th World Congress Action Learning, in large firms. Organization Science 3 (2): 179–202.
Action Research and Process Management, Netherlands, Groningen, Dougherty, D. 2002. Grounded theory research methods. In The Blackwell
August 22–24, 2007. Companion to Organizations, J.A.C. Baum (Ed.). Oxford: Blackwell.
Atuahene-Gima, K., and Y. Wei. 2011. The vital role of problem-solving Eisenhardt, K. M. 1989. Building theories from case study research.
competence in new product success. The Journal of Product Innovation Academy of Management Review 14 (4): 532–50.
Management 28 (1): 81–98.
Eisenhardt, K. M., and M. E. Graebner. 2007. Theory building from cases:
Baldwin, C. Y., and K. B. Clark. 2000. Design rules: The power of modu- Opportunities and challenges. Academy of Management Journal 50 (1):
larity. Cambridge, MA: MIT Press. 25–32.
Baldwin, C. Y., and E. von Hippel. 2011. Modeling a paradigm shift: From Felin, T., and N. J. Foss. 2005. Strategic organization: A field in search of
producer innovation to user and open collaborative innovation. Orga- micro-foundations. Strategic Organization 3 (4): 441–55.
nization Science 22 (6): 1399–417.
Foray, D. 2004. The economics of knowledge. Cambridge, MA: MIT Press.
Barney, J. B., and T. Felin. 2013. What are microfoundations? Academy of
Management Perspectives 27 (2): 138–55. Franke, N., and F. T. Piller. 2004. Value creation by toolkits for user
innovation and design: The case of the watch market. The Journal of
Bechky, B. A. 2003. Sharing meaning across occupational communities: Product Innovation Management 21 (6): 401–15.
The transformation of understanding on a production floor. Organiza-
tion Science 14 (3): 312–30. Geertz, C. 1973. The interpretation of culture. New York, NY: Basic Books.
Hartmann, B. 2009. Gaining design insight through interaction prototyping
Binder, T., and J. Redström. 2006. Exemplary design research. Design
tools. Unpublished doctoral thesis, Stanford University, Stanford, CA.
Research Society Wonderground International Conference, Lisbon,
Portugal, November 1–4, 1–13. Henderson, K. 1995. The political career of a prototype: Visual represen-
tation in design engineering. Social Problems 42 (2): 274–99.
Bødker, S., and K. Grønbæk. 1991. Design in action: From prototyping by
demonstration to cooperative prototyping. In Design at work: Coop- Hillebrand, B., and W. G. Biemans. 2004. Links between internal and
erative design of computer systems, ed. J. Greenbaum and M. Kyng, external cooperation in product development: An exploratory study.
197–218. Hillsdale, NJ: Lawrence Erlbaum Associates. Journal of Product Innovation Management 21 (2): 110–22.
Boehm, B. W., T. E. Gray, and T. Seewaldt. 1984. Prototyping versus Horst, W. 2011. Supportive tools for collaborative prototyping. Proceedings
specifying: A multiproject experiment. IEEE Transactions on Software of the Nordic Design Research Conference (NORDES), Helsinki,
Engineering 10 (3): 290–303. Finland, May 29–31.
Bogers, M., A. Afuah, and B. Bastian. 2010. Users as innovators: A review, Jeppesen, L. B. 2005. User toolkits for innovation: Consumers support each
critique, and future research directions. Journal of Management 36 (4): other. Journal of Product Innovation Management 22 (4): 347–63.
857–75. Keijzers, J., E. den Ouden, and Y. Lu. 2008. The “double-edged sword” of
Bogers, M., and H. Larsen. 2012. The role of improvisation in processes of high-feature products: An explorative study of the business impact,
innovation. Proceedings of the Participatory Innovation Conference, Proceedings of the 32nd Annual Product Development and Manage-
Melbourne, Australia, January 12–14, 2012. ment Association (PDMA) International Research Conference.
Bogers, M., and S. Lhuillery. 2011. A functional perspective on learning Klemmer, S. R., B. Hartmann, and L. Takayama. 2006. How bodies matter:
and innovation: Investigating the organization of absorptive capacity. Five themes for interaction design. Proceedings of the 6th Conference
Industry and Innovation 18 (6): 581–610. on Designing Interactive Systems, ACM, University Park, PA, 140–
149.
Bogers, M., and J. West. 2012. Managing distributed innovation: Strategic
utilization of open and user innovation. Creativity and Innovation Man- Laursen, K., and A. Salter. 2006. Open for innovation: The role of openness
agement 21 (1): 61–75. in explaining innovation performance among U.K. manufacturing
firms. Strategic Management Journal 27 (2): 131–50.
Brown, J. S., and P. Duguid. 1991. Organizational learning and
communities-of-practice: Toward a unified view of working, learning, Leonard, D., and J. Rayport. 1997. Spark innovation through empathic
and innovation. Organization Science 2 (1): 40–57. design. Harvard Business Review 75 (6): 102–13.
Brown, T. 2008. Design thinking. Harvard Business Review 86 (6): 84–92. Leonard-Barton, D. 1988. Implementation as mutual adaptation of technol-
ogy and organization. Research Policy 17 (5): 251–67.
Buchenau, M., and J. Fulton Suri. 2000. Experience prototyping. Proceed-
ings of the Conference on Designing Interactive Systems 2000 (DIS Leonard-Barton, D. 1995. Wellsprings of knowledge: Building and sustain-
‘00), ACM Press, New York, NY, 424–433. ing the sources of innovation. Boston, MA: Harvard Business School
Press.
Buur, J., and B. Matthews. 2008. Participatory innovation. International
Journal of Innovation Management 12 (3): 255–73. Lewin, A. Y., S. Massini, and C. Peeters. 2011. Microfoundations of internal
and external absorptive capacity routines. Organization Science 22 (1):
Carlile, P. R. 2002. A pragmatic view of knowledge and boundaries: Bound- 81–98.
ary objects in new product development. Organization Science 13 (4):
Lewin, K. 1946. Action research and minority problems. Journal of Social
442–55.
Issues 2 (4): 34–46.
Chesbrough, H. W. 2003. Open innovation: The new imperative for creat-
Lim, Y.-K., E. Stolterman, and J. Tenenberg. 2008. The anatomy of proto-
ing and profiting from technology. Boston, MA: Harvard Business
types: Prototypes as filters, prototypes as manifestations of design
School Press.
ideas. ACM Transactions on Computer-Human Interaction 15 (2):
Coghlan, D. 2011. Action research: Exploring perspectives on a philosophy 1–27.
of practical knowing. Academy of Management Annals 5 (1): 53–87.
Locke, K. 2001. Grounded theory in management research. Thousand
Dahlander, L., and D. M. Gann. 2010. How open is innovation? Research Oaks, CA: Sage.
Policy 39 (6): 699–709.
Macher, J. T., and D. C. Mowery. 2003. “Managing” learning by doing: An
Danneels, E. 2003. Tight–loose coupling with customers: The enactment of empirical study in semiconductor manufacturing. The Journal of
customer orientation. Strategic Management Journal 24 (6): 559–76. Product Innovation Management 20 (5): 391–410.
Dey, A. K., G. D. Abowd, and D. Salber. 2001. A conceptual framework Mascitelli, R. 2000. From experience: Harnessing tacit knowledge to
and a toolkit for supporting the rapid prototyping of context-aware achieve breakthrough innovation. The Journal of Product Innovation
applications. Human–Computer Interaction 16 (2–4): 97–166. Management 17 (3): 179–93.
 15405885, 2014, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/jpim.12121 by Hong Kong Metropolitan, Wiley Online Library on [20/08/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
764 J PROD INNOV MANAG M. BOGERS AND W. HORST
2014;31(4):744–764

Neyer, A.-K., A. C. Bullinger, and K. M. Moeslein. 2009. Integrating inside Ulrich, K. T., and S. D. Eppinger. 2008. Product design and development
and outside innovators: A sociotechnical systems perspective. R&D (4th ed.). New York, NY: McGraw-Hill.
Management 39 (4): 410–19. van Aken, J. E. 2004. Management research based on the paradigm of the
Nicolini, D., J. Mengis, and J. Swan. 2012. Understanding the role of design sciences: The quest for field-tested and grounded technological
objects in cross-disciplinary collaboration. Organization Science 23 rules. Journal of Management Studies 41 (2): 219–46.
(3): 612–29.
Vera, D., and M. Crossan. 2005. Improvisation and innovative performance
Poetz, M. K., and M. Schreier. 2012. The value of crowdsourcing: Can in teams. Organization Science 16 (3): 203–24.
users really compete with professionals in generating new product
ideas? The Journal of Product Innovation Management 29: 245–56. Verganti, R. 1999. Planned flexibility: Linking anticipation and reaction in
product development projects. Journal of Product Innovation Manage-
Rudd, J., K. Stern, and S. Isensee. 1996. Low vs. high-fidelity prototyping ment 16 (4): 363–76.
debate. Interactions 3 (1): 76–85.
Verganti, R. 2009. Design-driven innovation: How to compete by radically
Rust, R. T., D. V. Thompson, and R. W. Hamilton. 2006. Defeating feature innovating the meaning of products. Boston, MA: Harvard Business
fatigue. Harvard Business Review 84 (2): 98–107. School Press.
Sanchez, R., and J. T. Mahoney. 1996. Modularity, flexibility, and knowl- von Hippel, E. 1994. “Sticky information” and the locus of problem
edge management in product and organization design. Strategic Man- solving: Implications for innovation. Management Science 40 (4): 429–
agement Journal 17 (Winter Special Issue): 63–76. 39.
Schön, D. A. 1983. The reflective practitioner: How professionals think in
von Hippel, E. 2005. Democratizing innovation. Cambridge, MA: MIT
action. New York, NY: Basic Books.
Press.
Schrage, M. 1996. Cultures of prototyping. In Bringing design to software,
ed. T. Winograd, 191–205. New York, NY: ACM Press. von Hippel, E., and R. Katz. 2002. Shifting innovation to users via toolkits.
Management Science 48 (7): 821–33.
Schrage, M. 2000. Serious play: How the world’s best companies simulate
to innovate. Boston, MA: Harvard Business School Press. von Hippel, E., and M. J. Tyre. 1995. How learning by doing is done:
Problem identification in novel process equipment. Research Policy 24
Schuler, D., and A. Namioka. 1993. Participatory design: Principles and (1): 1–12.
practices. Hillsdale, NJ: Lawrence Erlbaum Associates.
Weick, K. E., K. M. Sutcliffe, and D. Obstfeld. 2005. Organizing and the
Siegel, S. M., and W. F. Kaemmerer. 1978. Measuring the perceived process of sensemaking. Organization Science 16 (4): 409–21.
support for innovation in organizations. Journal of Applied Psychology
63 (5): 553–62. West, J., and M. Bogers. 2013. Leveraging external sources of innovation:
A review of research on open innovation. Journal of Product Innova-
Song, X. M., M. M. Montoya-Weiss, and J. B. Schmidt. 1997. Antecedents
tion Management, doi:10.1111/jpim.12125.
and consequences of cross-functional cooperation: A comparison of
R&D, manufacturing, and marketing perspectives. The Journal of Yin, R. K. 2003. Case study research: Design and methods (3rd ed.).
Product Innovation Management 14 (1): 35–47. Thousand Oaks, CA: Sage Publications.
Song, X. M., R. J. Thieme, and J. Xie. 1998. The impact of cross-functional
joint involvement across product development stages: An exploratory
study. The Journal of Product Innovation Management 15 (4): 289– Appendix A. Acronyms of Involved
303. Stakeholders and Participants
Strauss, A. L., and J. Corbin. 1990. Basics of qualitative research:
Grounded theory procedures and techniques. Newbury Park, CA: Sage.
Terwiesch, C., and C. H. Loch. 2004. Collaborative prototyping and the CTM Communication technology manager
pricing of custom-designed products. Management Science 50 (2): DE Design engineers (DE1 and DE2)
145–58. DLS Design line specialist
Thomke, S. H. 1997. The role of flexibility in the development of new EU External user (EU1 and EU2)a
products: An empirical study. Research Policy 26 (1): 105–19. GW Global webmaster
Thomke, S. H. 1998. Managing experimentation in the design of new IDC Interaction design consultantb
products. Management Science 44 (6): 743–62. INT Innovation intern
Thomke, S. H., and D. E. Bell. 2001. Sequential testing in product devel- ME Marketing engineer
opment. Management Science 47 (2): 308–23. PPD Product portfolio director
Tushman, M. L., and D. A. Nadler. 1978. Information processing as an RDD R&D director
integrating concept in organizational design. Academy of Management RDPM R&D project manager
Review 3 (3): 613–24.
RDSD R&D senior director
Tyre, M. J., and W. J. Orlikowski. 1994. Windows of opportunities: Tem- SUEM Senior user experience managera
poral patterns of technological adaptation in organizations. Organiza-
tion Science 5 (1): 98–118.
UEC User experience consultanta
a
Ulrich, K. T. 2011. Design is everything? The Journal of Product Innova- External stakeholder.
b
tion Management 28 (3): 394–98. Researcher–designer.
 COMPTES RENDUS - BOOK REVIEWS 475
of literacy. Knowledge without correct political discipline was seen- accurately no doubt- to
be dangerous to the bourgeois order.

Gardner's book is an invaluable contribution to the literature of worl<ing-class education and

it speaks directly to anyone interested in the methodology of historical inquiry as well.

Bruce CURTIS
Wilfrid Laurier University

* * *

THoMAs P. HUGHES- Networks of Power: Electrification in We stem Society, 1880-1930. Baltimore:

Johns Hopkins University Press, 1983. pp. xi, 474.

This lucid and wide-ranging study of electrification cannot be recommended too highly for
historians of technology in general and of electric power systems in particular. Thomas P. Hughes
describes and analyzes the changing configuration of electrical generation, transmission, distribution
and consumption from the small central station systems of the 1880s to the large regional power
netwmi<s of the 1920s. His approach is both comparative and theoretical. On the one hand, Hughes
examines the evolution of electrical supply in the United States, Britain and Germany (only passing
reference is made to Canada) . On the other hand, he rests his analysis of particular individuals, in-
ventions and institutions on the assumption that the history of all large-scale technology is a history
of "systems" .

According to Hughes, electric power systems evolvP-d through five phases: invention and
development; technology transfer, during which site-specific technologies were adapted to different
environments; system growth, involving not only increases in size but correction of technical im-
balances that obstructed fulfillment of "system goals"; the achievement of "substantial momentum"
and the emergence of various organizations and political arrangements that formed what Hughes calls
"the system 's culture"; and a last phase when , in response to problems oflarge-scale planning and
funding, system control carne into the hands of financiers, consulting engineers and, in some cases,
govemrnent agency entrepreneurs. This model enables Hughes to order a vast array of technical detail
while incorporating non-technical elements into his story of systems development. It suggests a
universal pattern of evolution while allowing for the isolation of those characteristics which defined
a system's distinctive "style". Hughes argues that, as an explanatory device, the model is at once
peculiarly appropriate to the systems outlook of electrical men such as Thomas Edison and generally
applicable to the history of other modem technological systems. He suggests, indeed, that the study
of technology in light of his model is revealing of the systematizing nature of modem societies.

This book, then , is ambitious in scope and purpose. While close attention is devoted to specific
problems, such as the contribution of Lucien Gaulard and his English partner, John D . Gibbs , to
transformer development in the 1880s, or the metamorphosis of the American firm of Stone and
Webster from a partnership of consulting engineers into a large utility holding company in the 1920s,
these are always discussed in the context of broader issues . In the case of Gaulard and Gibbs, their
work is related to the general significance of transmission and distribution in electrical supply systems,
to the phenomenon of simultaneity in the invention process and to the problem of technical imbalances
in systems growth. In the case of Stone and Webster, their transformation is related to early twentieth-
century structural trends in the electrical industry and to the regionalization of power systems that
occurred after the World War I.

In addition to constructing a model of system formation and growth, Hughes offers certain
concepts for understanding the internal dynamic of technological change, notably the military met-
476 HISTOIRE SOCIALE- SOCIAL HISTORY

aphor of "reverse salients" , which he uses to describe technical imbalances in the third phase of
the model. Inventors and engineers viewed technology as a goal-seeking system and were therefore
concerned , like military strategists, to identify components that hampered the system as a whole.
These were then defined as "critical problems" and corrected. When such corrections failed to
harmonize with existing components, new systems were called forth, a process illustrated by the
resolution of the conflict between direct current and alternating current systems.
The order and coherence that distinguish this study are achieved at some cost. In adopting
the systems approach of his leading individual and institutional characters, Hughes substantially
accepts as his own framework their basic assumptions of technological rationality. The result is an
engineer's analysis of electrification from the inside out, and the social dimensions of the process,
despite the book's sub-title, receive comparatively little consideration. Still Hughes is not unaware
of the Whiggish tendencies of his model . He insists that technology is man-made, that it is not neutral
and that the capitalist entrepreneurship of Edison , Emil Rathenau, Samuellnsull and Charles Merz
was integral to their system building. And if systems have an internal drive and a universal character,
Hughes also argues that they are influenced and differentiated by culture, geography and historical
contingencies. Certainly, readers with any awareness of Adam Beck and Robert Bourassa will ac-
knowledge the force of his contention that technological systems acquire a conservative momentum.

Kenneth C. DEwAR
Mount Saint Vincent University

* * *

VERNON L. LIDTKE- The Alternative Culture: Socialist LAbor in Imperial Gemumy. Oxford: Oxford
University Press, 1985 . pp. x, 299.

This volume represents a rather unique attempt by Professor Lidtke to see German Social
Democracy prior to 1914 from a radically different perspective. Instead of explaining this movement
from the point of view of the Social Democratic Party or the leaders of the powerful pre-war trade
unions, he has literally gone into the popular working-class associations of the time to determine more
closely the feelings and ideas of Germany's emerging laboring class. Making use of analytical
techniques that have become common to social history over the past 20 years, Lidtke progressively
examines working-class clubs, the festivals they held, the songs they sang, the poetry and drama
that they listened to and the lectures and courses they attended. This prolonged look at working-class
life from the bottom up is a major contribution to German historiography. Dr. Lidtke is at the very
same time an extraordinary writer. There is real power in the clarity and simplicity of his writing
style. His meanings are never in doubt. But this strength turns against itself as the volume progresses,
primarily because Lidtke' s crystal-clear interpretations unfortunately, on occasion, carry him beyond
the evidence he has himself accumulated.
The media by which the workers themselves, in the words of the author, created their own
"social [and] cultural milieu" were the voluntary associations, or Vereine, of this era. They steadily
sprang into existence before the anti-Socialist laws of 1878 and especially after those laws were struck
down as of 1890. The two oldest and, as it turned out, the largest groups here were the singing and
gymnastic clubs. Other voluntary organizations encouraged such physical activities as swimming,
boxing and rowing as well as the exploration of nature. Because Lidtke is not strong when it comes
to numbers, it is difficult to tell from his narrative just what percentage of the German working class
belonged to these voluntary associations. According to the contemporary German historian, Michael
Schneider, the free trade unions had enrolled by 1913 slightly more than two million members, of
whom Lidtke says some 186,958 were organized into gymnastic societies. But the actual number
of those involved in singing and other societies is never definitely stated. As a result, it is impossible
This article was downloaded by: [128.197.26.12] On: 15 September 2016, At: 01:48
Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA

Information Systems Research

Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org

Toward a Theory of Remixing in Online Innovation

Communities
Michael A. Stanko

To cite this article:

Michael A. Stanko (2016) Toward a Theory of Remixing in Online Innovation Communities. Information Systems Research

Published online in Articles in Advance 13 Sep 2016

. http://dx.doi.org/10.1287/isre.2016.0650

Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact [email protected].

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.

Copyright © 2016, INFORMS

Please scroll down for article—it is on subsequent pages

INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
 Information Systems Research
Articles in Advance, pp. 1–19
ISSN 1047-7047 (print) — ISSN 1526-5536 (online) http://dx.doi.org/10.1287/isre.2016.0650
© 2016 INFORMS

Toward a Theory of Remixing in

Online Innovation Communities
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Michael A. Stanko
Poole College of Management, North Carolina State University, Raleigh, North Carolina 27695, [email protected]

W ithin online innovation communities, remixing (i.e., the community’s use of an existing innovation as
source material or inspiration to aid in the development of further innovations) is an interesting form of
knowledge collaboration. This study investigates an open theoretical question: Why are particular innovations
remixed by online innovation communities? Some innovations languish, while others are widely remixed. Com-
munity members (even those unknown to the innovation’s creator) may remix, taking the source innovation in
directions the original innovator may have never imagined. Within online innovation communities, remixing is
not bound by some of the constraints to knowledge collaboration found in more traditional environments. To
address our research question, we begin with variables constituent to innovation diffusion theory, essentially
remixing this long-established theory to predict cumulative remixing in online innovation communities, using
arguments grounded in the user innovation and learning literatures. We also consider two forms of communica-
tion that are relevant to knowledge sharing in online communities (online community interaction and front page
presence). Regression analysis (using data pertaining to 498 3D printable innovations) shows that community
interaction is highly influential in determining which innovations are remixed by the community. Conversely,
the innovation having a presence on the community’s front page does not have a significant effect on remixing.
Observability has an inverse-U-shaped relationship with remixing; this suggests the value placed on experien-
tial learning. Fuzzy set qualitative comparative analysis (fsQCA) is used as a supplementary analysis technique
(with robustness testing), and the results are largely consistent with regression findings but offer interesting
insight into innovation configurations that consistently result in remixing from the community. Within specific
configurations, fsQCA results indicate a contingent effect of observability and complexity; that is, under certain
conditions, complex innovations are more likely to be remixed by the community.
Keywords: online innovation communities; remixing; user-generated content; innovation diffusion theory;
3D printing
History: Samer Faraj, Georg von Krogh, Karim Lakhani, Eric Monteiro, Senior Editors; Lars Frederiksen,
Associate Editor. This paper was received on October 30, 2014, and was with the author 7 months for 3
revisions. Published online in Articles in Advance September 13, 2016.

1. Introduction innovations (Lenhart et al. 2010, Lessig 2008). Remixes

This research endeavors to understand an online inno- leverage “meaning created by the [source] reference
vation community’s propensity to remix particular to build something new” (Lessig 2008, p. 76). The
innovations. Online innovation communities devoted shift toward user-generated content (UGC) and away
to, for instance, software (e.g., sourceforge.net), 3D from professionalism (i.e., the “cult of the amateur”)
printing (e.g., thingiverse.com), and multimedia (e.g., has drawn criticism (e.g., Keen 2008), though it is
scratch.mit.edu), facilitate read/write culture (see clear that read/write culture is gathering momentum.
Improvements in information technology (IT) tools
Lessig 2008), in which members of online communities
and their widespread availability (such as design soft-
can consume innovations developed by other com-
ware and tools easily learned online, infrastructure for
munity members, but are also empowered to remix
widespread sharing of UGC, and the availability of
others’ ideas to develop new concepts. The term remix lower cost hardware) fuel this momentum, enabling a
was originally applied narrowly to music and later shift from the traditional producer-to-consumer inno-
video, though in recent years remix has become com- vation model toward an ecosystem of user inno-
monplace in contexts such as literature and pho- vator collaboration (Rifkin 2014, Yoo et al. 2010).
tography, as well as within innovation communities Against this backdrop, an important research question
(Hill and Monroy-Hernandez 2012, Navas 2012). In becomes, Why are particular innovations remixed by
online innovation communities, remixing is the com- an online innovation community? Some innovations
munity’s use of an existing innovation as source mate- are remixed many times by the innovation community,
rial or inspiration to aid in the development of further taking the source material in directions the original
1
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
2 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

innovator may not have dreamed of. Online commu- concepts (Kolb 1984). Remixing also results in a ref-
nities enable knowledge collaboration without direct erence point (i.e., the remixed innovation) to test the
dialogue, even between members who do not know validity of thinking during the learning process. The
each other (Faraj et al. 2011). At the same time, many importance of learning and the context of remix-
other innovations are ignored by the community, fail- ing within online innovation communities necessitate
ing to generate remixes. We set out to explain this extending, adapting, and updating Rogers’ (2003) the-
disparity. orization to address our research question. Online
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Examining the determinants of remixing is neces- communities are flexible, complex settings that differ
sary to developing a more complete understanding drastically from traditional organizational forms, as
of online innovation communities, which is of inter- members (and their interests, identities, and efforts)
est to both managers and researchers. There is an constantly shift within and flow into or out of the com-
acknowledged need to better theoretically understand munity (Faraj et al. 2011). Community interaction and
knowledge collaboration in online communities (Faraj knowledge sharing are defining elements of online
et al. 2011), as well as users’ roles as innovators and innovation communities (Jeppesen and Frederiksen
shapers of others’ contributions (Majchrzak et al. 2013, 2006, von Krogh et al. 2003), typically with publicly
Nambisan et al. 1999). Importantly, there are benefits to visible discourse with regard to specific innovations.
harnessing the potential of user innovators by devel- Community interactions may include criticism or sug-
oping an innovation that is remixed. Online innova- gestions to improve innovations. We argue that online
tion communities generate tremendous value through community interactions are important to stimulat-
remixing: improving innovations’ shortcomings, com- ing remixing. Also, given that learning is a priority
bining aspects of multiple innovations, or optimiz- within these communities, we theorize that excessive
ing innovations for particular applications (Hill and observability can affect the community’s tendency to
Monroy-Hernandez 2012). Innovating firms are aware remix innovations. Given that much can be learned
of this; organizations such as Audi, Starbucks, Harley by observing these innovations, the value created for
Davidson, and Nike are increasingly looking to har- community members through remixing may be less-
ness the power of user innovators (Füller et al. 2008). ened in these cases.
Innovation tactics drawing on experts, customers, The remainder of this article is organized as follows.
and critics, such as crowdsourcing and hackathons, The next section introduces our theoretical founda-
have become much more mainstream (Dockser Mar- tion. In Section 3, hypotheses are formed, extend-
cus 2014, Poetz and Schreier 2012). Despite remix- ing innovation diffusion theory to explain remixing
ing becoming commonplace across contexts and the in online innovation communities. The data used for
potential value of understanding its determinants, lit- this research (relating to 498 3D printable innova-
tle is known about the factors making the remixing tions) are described in Section 4. Two distinct analysis
of an innovation likely (Hill and Monroy-Hernandez approaches are taken: First, hypotheses are formally
2012). Even in cases where remixing is not sought after, tested using regression analysis. After this, fuzzy set
an understanding of the determinants of remixing may qualitative comparative analysis (fsQCA) is used to
help discourage the unwanted behavior. detect complex contextual effects. The article con-
Why particular innovations are remixed by an cludes with a discussion of the study’s contributions
online innovation community is an open theoreti- as well as limitations and future research directions.
cal question. To address this, we begin with Rogers’
(2003) diffusion factors as explanatory variables, alter- 2. Theoretical Background
ing Rogers’ arguments, and extending innovation dif- The idea that users might customize a product is not
fusion theory to explain remixing. Given our interest a new one. Rogers (2003, p. 17) discusses the notion
in understanding remixing in innovation communi- of reinvention: “the degree to which an innovation is
ties, it is perhaps natural that we, in essence, remix changed or modified by a user in the process of its
a long established theory for this purpose. Tradition- adoption and implementation.” This notion of rein-
ally, the diffusion of innovations has been thought to vention includes users who opt to use several, but not
center on the reduction of potential consumers’ uncer- all, components of an off-the-shelf software offering,
tainty (Rogers 2003, Tornatzky and Klein 1982). Here, or car enthusiasts who add after-market accessories
we argue that the community’s remixing of an existing after purchasing a car. The distinction for IT-enabled
source innovation can involve a different considera- UGC is important: within online innovation commu-
tion: community members’ desire to learn. Learning is nities, prior consumption is not required to modify an
a primary incentive for user innovation (Lakhani and innovation. The ability to change the invention is not
Wolf 2005, Ståhlbröst and Bergvall-Kåreborn 2011). restricted to those who first choose to consume. Inter-
Remixing is one way to gain experiential learning. Per- ested members of the community can remix an inno-
sonal experience adds meaning to previously abstract vation (a form of knowledge collaboration between
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 3

the creator of the original innovation and the com- have examined the effect of Rogers’ (2003) five fac-
munity of remixers), regardless of consumption and tors on consumption (e.g., Arts et al. 2011, Tornatzky
even despite a lack of social relationship with the inno- and Klein 1982). These meta-analyses show that rel-
vation’s creator or those who have consumed. This ative advantage (the degree to which an innovation
demonstrates the unprecedented fluidity of collabora- is perceived as better than the idea it supersedes)
tion within online communities (see Faraj et al. 2011) and compatibility (whether an innovation is congruent
not possible in traditional contexts, nor in the time of with existing understandings of members of a social
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Rogers’ (2003) theorizing. system) are influential determinants of consumption.

Given our research question (why are particular Empirical support related to the influence of trialabil-
innovations remixed by an online innovation commu- ity (the degree to which an innovation may be exper-
nity?), we employ constituent constructs from inno- imented with), observability (the degree to which an
vation diffusion theory (Rogers 2003). Innovation innovation is available and visible), and complexity (the
diffusion theory’s identification of multiple innovation degree to which an innovation is perceived as rela-
attributes makes it well suited to adaptation to pre- tively difficult to understand and to use by members
dict community remixing. This is as opposed to other of a social system) on consumption has been mixed,
theoretical explanations of the propagation of technol- with many nonsignificant findings reported for these
ogy not incorporating wide ranging sets of product theorized diffusion factors (Arts et al. 2011, Jeyaraj
attributes, and instead focusing on characteristics, atti- et al. 2006, Tornatzky and Klein 1982). In Section 3, we
tudes, and intentions of users, such as the technology consider the peculiarities of remixing, such as the pri-
acceptance model (Davis 1989, Venkatesh 2000) and oritization of learning, as well as the fluid, interactive,
the unified theory of acceptance and use of technology publicly visible nature of online communities in devel-
(Venkatesh et al. 2003).
oping theoretical arguments extending Rogers’ (2003)
Innovation diffusion theory has been widely applied
theory to predict remixing by an online innovation
to organizational IT innovations (such as laptops,
community.
computer-aided drafting, email, business productiv-
Communication channels also play a central role
ity software, electronic record keeping; Brancheau and
in innovation diffusion theory (Agarwal and Prasad
Wetherbe 1990, Oborn et al. 2011) as well as to behav-
1998, Rogers 2003). The contemporary communica-
iors such as shopping online (Pavlou and Fygenson
tion mechanisms that characterize online communities
2006) or browsing the mobile Internet (Hsu et al.
2007). Throughout this paper, we use the term inno- are critical in enabling collaboration and knowledge
vation in a broad sense, consistent with Rogers’ (2003) sharing, and have shaped the culture of these com-
definition: an idea, practice, or object perceived as munities (Goh et al. 2016, von Krogh et al. 2012).
new by the individual considering it. In prior diffu- Accordingly, we consider the level of online community
sion research, the typical assumption has been that interaction (the degree to which an innovation is exten-
adoption of a product is binary: either the product is sively discussed online by members of the innovation
adopted or not (Mahajan et al. 1995), with a general community). We also consider front page presence (pro-
emphasis on total market response (i.e., sales volume). motion of the innovation on the online community’s
Note that adoption refers to an individual user’s deci- home page). Importantly, front page presence is an
sion, whereas diffusion refers to the collective response effort to promote the innovation, while observability
of all users. There has been some investigation into the is a trait of the innovation itself. To illustrate, Rogers
diffusion of UGC; for instance, researchers have exam- (2003, p. 259) uses the examples of hardware innova-
ined the diffusion of open source software (Bonaccorsi tions being more observable and software being less
and Rossi 2003) and beta releases (Makinen et al. 2014), observable due to the nature of these products. Of
as well as the diffusion of YouTube videos (e.g., Liu- course, it is possible for either type of innovation to be
Thompkins and Rogerson 2012, Susarla et al. 2012) widely communicated through front page presence.
and multimedia content (Hill and Monroy-Hernandez
2012). Here, we are concerned with functional innova- 2.1. Remixing and Motivations for
tions (i.e., not art forms). The notion that aspects of a User Innovation
functional innovation may impact its propensity to be Though the notion of remixing has not yet received
remixed has yet to be examined. much research attention from innovation scholars (for
Since innovation diffusion theory is well estab- an exception in the context of multimedia production,
lished (excellent reviews include Greenhalgh et al. see Hill and Monroy-Hernandez 2012), the growing
2004, Rogers 2003), our focus is on considering the literature on user innovation, particularly in innova-
challenges to applying it to explaining remixing tion communities, is relevant in developing hypothe-
within online innovation communities and extend- ses with regard to remixing. Note that we do not
ing it accordingly, rather than summarizing its theo- draw a strong distinction between various research
rized effects on consumption. Several meta-analyses streams related to user innovators in developing our
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
4 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

hypotheses (e.g., open source software development, employment opportunities) motivations for participa-
peer production communities, customer communities) tion (Fleming and Waguespack 2007, Hertel et al. 2003,
given the commonalities that these groups share, such Jeppesen and Frederiksen 2006, Roberts et al. 2006).
as (1) the importance community members place on The desire to feel creative during the coding process
learning and their need to make use of the innova- as well as the need for intellectual stimulation have
tions developed; (2) the centrality of remixing, known also been shown to be very important motivators for
as code reuse, splintering, or tweaking in other con- contributions to open source software development
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

texts; and (3) members’ range of involvement in these (Lakhani and Wolf 2005). It is possible that motives for
communities, from an active core to more periph- participation in an innovation community evolve over
eral, passive members (Haefliger et al. 2008, Hill and time (Shah 2006); community members may initially
Monroy-Hernandez 2012, Jeppesen and Frederiksen join to improve an innovation for their own use, but
2006, Rullani and Haefliger 2013, Ståhlbröst and stay involved due to learning or enjoyment.
Bergvall-Kåreborn 2011).
Research related to innovation communities iden-
tifies numerous motivations for community involve- 3. Hypotheses
ment (Shah 2006, von Krogh et al. 2012). Two of the Online innovation communities will tend to remix
most critical motivators for involvement in innovation innovations with a high degree of relative advantage
communities are the desires to (1) learn (Ståhlbröst for several reasons. One important motivation in inno-
and Bergvall-Kåreborn 2011) and (2) make use of the vation communities is own use; that is, communi-
innovation being developed (Lakhani and Wolf 2005). ties of user innovators are pragmatically motivated to
Both these incentives are thought to be relevant to pre- develop the best possible innovations to meet their
dicting the community’s collective remixing behavior own needs (Hertel et al. 2003, Shah 2006, von Hippel
and will inform our hypothesis development. First, 1988). Given this, efforts to produce innovations of
interest-based learning is a substantial part of the superior quality (for community members’ use) will be
answer to “why would community members invest aided by starting with excellent source material. Thus,
time and energy to remix an existing innovation?,” just the community is incented to locate relatively advan-
as learning has been found to be an important moti- tageous source innovations when remixing. Relatively
vator for other types of contributions to innovation advantageous source innovations offer a “credible
communities (Ståhlbröst and Bergvall-Kåreborn 2011, promise” as to the viability of future related devel-
von Krogh et al. 2012). Where traditional learning is opments (Haefliger et al. 2008), attracting community
hierarchical (teacher to student), remixing represents members to promising source innovations. The higher
democratic learning by doing through building on the ratio of benefits to costs of a relatively advantageous
ideas of other community members. Because of the innovation encourages potential remixes, just as it is
importance of learning in determining an online inno- traditionally thought to foster consumption (Rogers
vation community’s remixing behavior (Lessig 2008), 2003). Given the importance of reputation in innova-
we will draw on the theory of experiential learning tion communities (Lerner and Tirole 2002), commu-
(Kolb 1984, Straub 2009). The remix process begins nity members will prefer to associate themselves with
with some uncertainty regarding the obstacles com- a source innovation that is relatively advantageous,
munity members may encounter along the way, but and conversely, may distance themselves from source
with assurance that they will develop their skills, prob- innovations lacking relative advantage.
lem solve, and reflect on choices made once the remix Learning is an important outcome from remixing
is complete; this experiential learning process is cen- (Franke and Hader 2014, Hill and Monroy-Hernandez
tral to remixing. 2012). Experiential learning theory stipulates that
Own use is also thought to be an important moti- learning is a tension- and conflict-filled dialectic (i.e.,
vation for remixing, just as it has been shown to be truth finding) process of observation, action, and
elsewhere in innovation communities (Lakhani and reflection (Kolb 1984). Real life experiences validate
Wolf 2005, von Krogh et al. 2012). Community mem- abstract concepts; personal experience is viewed as the
bers self-servingly wish to develop the best possible focal point for learning within experiential learning
innovations for their own use, which drives them to theory. In the case of remixing, selecting source inno-
pursue remixing, just as the desire for superior innova- vations with a high degree of relative advantage will
tions motivates members to contribute effort in other result in better learning outcomes since the observa-
ways within innovation communities (Hertel et al. tion stage will include assessing and understanding
2003, Shah 2006). Beyond learning and making use of superior source material. Furthermore, for remixers,
the innovation, innovation community members have the reflection phase will often include a learning com-
numerous other intrinsic (e.g., enjoyment, exposure parison between the source innovation and the remix
to new innovations) and extrinsic (e.g., reputation, developed (this can be thought of as a competitive
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 5

comparison to understand each innovation’s relative the community’s propensity to remix a source inno-
merits); comparison against high-quality source mate- vation. Recall that observation is the first stage in the
rial again improves learning outcomes. Remixers col- experiential learning process (Kolb 1984). Community
lectively select source innovations with high levels of members are able to engage in a more participative
relative advantage as a means to direct their remix observation phase for trialable innovations (without
experience toward more optimal learning outcomes; the concern of encountering restrictions), encouraging
the community to begin and make continued progress
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

purposefully directing learning is commonplace in

self-regulated learning environments (Dabbagh and on remixes (see Buijs 2003). Trialability allows experi-
Kitsantas 2012). Thus, we argue the following: mentation in a low-risk setting, letting remixes gather
momentum without the concern of complying with
Hypothesis 1 (H1). Relative advantage has a positive restrictions. Decreasing community members’ fear of
relationship with remixing. risk taking is important to promote creative behaviors
The online innovation community will not respond (Dewett 2006), such as remixing.
to an innovation (through remixing or otherwise) Trialability also allows for a greater spectrum of
unless the innovation in question can be placed within allowable outcomes through experimentation. Com-
the context of existing community understandings munities of user innovators are self-servingly moti-
(Hargadon and Douglas 2001). Source innovations that vated to develop innovations that are valuable to
are more compatible with community expectations themselves (Hertel et al. 2003, von Hippel 1988). Thus,
allow for the innovation to be mapped more eas- limitations on possible outcomes lessen the potential
ily to members’ existing cognitive schemas (Moreau value derived from remixing, acting as a disincen-
tive. In this context, limitations include restrictions
et al. 2001). Generally, it is easier for innovations to
on the type of remixes that can be developed from
generate a market response if they do not challenge
a particular source innovation. Limits imposed on
consumers’ expectations of how a product should be
potential outcomes will reduce community motiva-
used (Gourville 2006). In the same way, members of
tion and engagement (Zhang and Bartol 2010), lessen-
the innovation community will collectively be more
ing the community’s propensity to remix the source
likely to conceive of improvements and personaliza-
innovation.
tions for source innovations that are compatible with
their expectations. Also, innovations that are compat- Hypothesis 3 (H3). Trialability has a positive relation-
ible with previously introduced ideas are more likely ship with remixing.
to be well received by the innovation community
Observability is posited here to have an inverse-
as they tap into established interest areas within the
U-shaped relationship with remixing. In cases where
community.
observability is very low, the community will be
On the other hand, source innovations that are
unlikely to remix the source innovation. Very low lev-
not compatible with the expectations of the innova-
els of observability may present a practical barrier to
tion community will tend to languish, failing to be
remixing—community members may not be able to
remixed. In the words of Hargadon and Douglas (2001, remix an innovation without thoroughly observing it.
p. 478), “Purely novel actions and ideas cannot reg- Furthermore, creative activities (such as remixing) are
ister because no established logics exist to describe fostered by seeing and understanding the ideas of oth-
them.” Community members will not anticipate ben- ers (Amabile 1996). In the context of the theory of expe-
efits from the use of a remixed innovation when they riential learning, the observation phase that begins the
are unable to comprehend the source innovation itself. cycle of experiential learning (Kolb 1984) cannot be fer-
For community members to become entangled (in tile when the source innovation is not easily observed.
the parlance of sociomateriality; see Orlikowski 2007) Unobservable innovations will not capture the collec-
with a source innovation through remixing, a basic tive imagination of community members and will fail
understanding is first required. Put simply, commu- to demonstrate that the source innovation is of enough
nity members will likely not be inspired to remix substance to warrant remixing. Thus, innovation com-
noncompatible innovations that they may not even munities will be unlikely to remix innovations with
understand how to use, just as consumers generally very low observability.
are slow to consume incompatible innovations (Arts Innovation diffusion theory holds that observabil-
et al. 2011, Straub 2009). ity makes the benefits of a product more apparent
Hypothesis 2 (H2). Compatibility has a positive rela- and thus increases consumption (Arts et al. 2011). How-
tionship with remixing. ever, with respect to remixing, since part of the benefit
in question is learning, high levels of observability
Trialability (the degree to which an innovation may reduce the incremental learning possible for commu-
be experimented with) is thought to positively impact nity members and, thus, part of the value generated
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
6 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

through remixing. While learning is a primary motive source innovations will result in community mem-
within innovation communities (Lakhani and Wolf bers collectively remixing these innovations more
2005, Ståhlbröst and Bergvall-Kåreborn 2011), commu- frequently.
nity members may have little left to learn from remix-
Hypothesis 5 (H5). Complexity has a negative rela-
ing highly observable innovations, since all aspects
tionship with remixing.
of the innovation are apparent through observation.
The final two hypotheses concern contemporary
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Therefore, we expect that those innovations with very

high levels of observability will discourage remixing. forms of communication within online communities
With less incremental learning available, interested (front page presence and online community inter-
members of the innovation community will collec- action). Members of online communities who are
tively tend not to remix these highly observable inno- actively engaged in producing original innovations
vations. Very high observability acts to shift some or remixing others’ innovations can be thought of
community learning regarding the source innovation as core members. Conversely, those that only con-
from remixing (a form of active learning) to simply sume others’ innovations (i.e., passive users) can be
examining (passive learning; see James et al. 2002). thought of as being on the periphery of the inno-
In short, innovations with enough observability to vation community (Dahlander and Frederiksen 2012,
demonstrate the innovation’s substance and spark the Faraj et al. 2011). Core members typically produce the
collective imagination of the online innovation com- vast majority of contributions within innovation com-
munity, but not so much observability as to diminish munities (Dahlander and Frederiksen 2012). Periph-
the incremental learning available through remixing, eral members contribute to the innovation community
are the most likely to be remixed extensively by an only sporadically, tending to lurk in the background
innovation community. as invisible observers (Rullani and Haefliger 2013).
While peripheral members tend to not develop their
Hypothesis 4 (H4). Observability has an inverse- own contributions, they are important as consumers
U-shaped relationship with remixing. of others’ innovations within the community (Setia
et al. 2012). This is akin to the “onion-like” hierar-
Innovations that are simple (i.e., not complex) are chy observed in open source software development
more likely to generate a remix response from an inno- (Crowston and Howison 2006).
vation community for several reasons. First, simple Core members of online communities prefer the
innovations are, by nature, easy for the community to opinions of others within the community, which gen-
understand. Simplicity allows a larger base of commu- erally comply with community norms, over commu-
nity members to envision ways that they could benefit nications tactics intended to reach a broad audience
from remixing the source innovation. Given that com- (Kozinets et al. 2010). Core members actively take con-
munity members must be able to comprehend (at least trol over their information consumption and resist
generally) an innovation before they can remix it, some efforts to direct their attention, such as promotion on
community members will find very complex innova- the community’s home page (see Mangold and Faulds
tions confusing and effortful to understand initially 2009). Because of the substantial investment in time
(Rogers 2003). These members may avoid overly com- that core members make within the community and
plex innovations entirely. In short, complexity acts as a the ease of online communication, core members tend
barrier to community engagement with an innovation to actively resist this type of online promotion (Cox
(Arts et al. 2011). 1998, Crowston and Howison 2006) as an “inevitable
Second, simple innovations are more straightfor- defensive position in the circumstances” (Cox 1998).
ward to build on through remixing (and improve- This is one of the processes that shapes innovation
ments made via remixing may be more apparent). communities into having the core/peripheral structure
Adding features to or modifying the initial design is (Rullani and Haefliger 2013). Since the core members
less complicated when drawing on a straightforward producing the vast majority of remixes do not respond
source innovation (Snodgrass and Vanderwart 1980). to online promotional tactics such as an innovation
Less complex source innovations provide commu- having a presence on the community’s home page, this
nity members with fewer constraints in reimagining negates any potential impact of front page presence
the innovation, offering more open-ended possibilities on the community’s remixing behavior.
for future engagement (Hill and Monroy-Hernandez
Hypothesis 6 (H6). Front page presence has no rela-
2012). Because of this lack of constraints, simpler
tionship with remixing.
innovations may be more enjoyable to remix; enjoy-
ment has been shown to be an important motivator While we argue in H6 that front page presence
in innovation communities (Ståhlbröst and Bergvall- does not relate with remixing, H7 posits that online
Kåreborn 2011). The ease and joy of remixing simple community interaction has a positive relationship
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 7

with remixing, just as traditional word of mouth 2012). The first portion of the data we use to test
is typically thought to have a positive relationship our hypotheses was drawn from a publicly available
with consumption (Agarwal and Prasad 1998, Rogers “snapshot” Thingiverse data set, made available by
2003). Interactive knowledge sharing is germane to peerproduction.net. These data were webscraped (i.e.,
online innovation communities; commentary visible data were systematically harvested from the Internet)
to all community members can quickly and eas- on November 5, 2013. Since a one year time lag is
ily be generated by any member of the community desired, we use only the objects publicly posted in
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

(von Krogh et al. 2003). This characteristic of online October 2012. Of the 79,498 designs that had been pub-
innovation communities may render electronic word licly posted since the inception of Thingiverse in 2008,
of mouth even more impactful in determining com- 1,150 of these were posted during October 2012. For
munity remixing than traditional word of mouth com- these 1,150 objects, time-invariant variables (such as
munication is in a typical innovation diffusion context. the category of the posting and the number of files
Since interactions between members of online inno- necessary to print the design) needed for our analysis
vation communities often include discussion of how to were added to the original data via manual webscrap-
achieve better or different solutions (Shah 2006), online ing. During this process, the original publicly available
community interaction regarding a source innovation data were checked for accuracy and consistently found
is thought to be one of the most important stimula- to be reliable.
tors of community remixing of that source innovation. Thingiverse is an ideal setting to test our hypothe-
Even where discussion among community members ses, given this community’s focus on individual ob-
points out limitations or needs that are not addressed jects. Several aspects of this context and the time
by a design, this communication may spur the com- period examined ensure that potential user-level
munity to address these concerns through remixing. determinants of remixing are not likely confounds.
Discourse regarding a new innovation can also be First, Thingiverse is a meritocratic space (Klein 2013),
important in establishing the innovation’s legitimacy where objects (rather than designers) are evaluated.
to the community (Barrett et al. 2013), particularly The site architecture is organized around objects, with
given the public visibility of community comments. comments only organized by object. During the time
Community members may view community interac- span investigated, Thingiverse users likely knew little
tion pertaining to a source innovation as an indica- about the designer of an object they were examining.
tor that a remix of that source innovation might also Says Eric Mortensen, Thingiverse’s community and
generate conversation within the community, which audience development manager, “Thingiverse isn’t a
is likely important to community members concerned social network” (Mortensen 2015). Also, both 3D print-
with their reputations (see Lerner and Tirole 2002). ing and remixing represent substantial commitments.
Each print typically requires several decisions to be
Hypothesis 7 (H7). Online community interaction has
made (such as infill—the proportion of a solid object
a positive relationship with remixing.
that should be filled in with material during print-
ing), multiple hours of printing time, plus the cost of
4. Empirical Study materials. Remixing also involves a substantial time
To examine an online innovation community’s remix- commitment. This ensures that users are printing or
ing behavior, we use data from Thingiverse (http:// remixing an object due to the merits of that object.
www.thingiverse.com), thought to be the largest During the time span investigated, user-level fea-
online community dedicated to 3D printable objects. tures now common on UGC sites had not yet been
Thingiverse allows users to post their objects (i.e., established within Thingiverse. Currently, like many
innovations) for other members of the community to UGC communities, Thingiverse has a “follow” feature,
print, comment on, and remix. Designers post one allowing community members to easily monitor other
or more images of their objects as well as one or members to keep up with their work. This feature was
more files necessary to print or remix the object on launched by Thingiverse in October 2012 (the same
the object’s Thingiverse page. Thingiverse designers month objects in our sample were posted; Mortensen
generally choose from Creative Commons licenses 2015), but was not widely used until much later: “It
(McCarthy 2012), which allow more or less restriction has taken a very long time for people to pick up ‘fol-
on the use and remixing of their objects. When com- lowing’ on Thingiverse” (Mortensen 2015). Since the
munity members print or remix an object, they are follower network was not established on Thingiverse
directed to report this activity. during the time span we investigate, relational differ-
Since time is required for community remixing ences in users’ network placement cannot confound
once innovations are made public, we allow one year our results. Also, a “group” feature (letting Thingi-
between the posting of objects and measurement of the verse members participate in discussions with oth-
frequency of remixing (Hill and Monroy-Hernandez ers focused on a particular subject, software tool, or
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
8 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

printer) was launched in 2014 (Mortensen 2015). Prior object without going through the object’s page. Con-
to this, designers’ individual interests were not eas- sistent with the total product concept (Levitt 1980),
ily discernible (other than through their designs), and the page is considered an extension of the physical
the possibility of a shared group relationship between object. One preferred method researchers have used
members did not exist.1 to measure relative advantage is to rely on potential
Several types of postings inappropriate for our anal- adopters’ evaluations of the innovation (Tornatzky
ysis were screened out prior to analysis: and Klein 1982). Consistent with this approach, we
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

1. We removed postings that were incomprehensi- use the number of likes from members of the inno-
ble or marked as a user testing his ability to post to vation community normed by the number of page
the site. views. Liking is a way to quickly and conspicuously
2. We removed postings that contained no files to show positivity toward an online object (Gerlitz and
download or had no images of the object. Helmond 2013), collectively providing an indicator of
3. We removed postings by Makerbot Industries whether or not this meritocratic community (Klein
(the parent company of Thingiverse), since these are 2013) sees a particular design as having advantages
not UGC. over existing innovations. Norming the number of
4. Since we are concerned with functional inno- likes by page views avoids bias in the measure due
vations (i.e., not art or purely aesthetic work), we to some objects generating more awareness within
removed several categories of objects (e.g., “art,” the community. Compatibility (whether an innovation
“sculptures,” “scans and replicas”). These categories is congruent with existing understandings of mem-
include objects that are inconsistent with our focus on bers of a social system) is measured using an indica-
functional innovations and our application of innova- tor variable indicating whether the source object was
tion diffusion theory. Uncategorized posts were also itself developed by remixing a previous design (1) or
removed. not (0). Being remixed from a previous design specif-
5. Categories with fewer than 10 postings during ically addresses one aspect of compatibility identified
the focal time span and those categories with no by Rogers (2003): compatibility with previously intro-
reported makes or remixes were removed. This allows duced ideas. Designs that are remixed from a prior
us to control for category effects without encountering design are, by nature, more consistent with previ-
estimation concerns in our analyses. ously introduced ideas within the community than
6. Postings that did not use one of the standard Cre- those that are not. The measure employed is con-
ative Commons licenses typically used on Thingiverse sistent with primary measures of compatibility that,
and postings with licenses that explicitly did not allow for instance, rate innovations on “how much differ-
for remixing were removed. ent the innovation is from older ways of doing the
After this screening process, our data set contained job” (Tornatzky and Klein 1982, p. 34), and with sec-
498 designs from 22 categories, including 3D printer ondary studies using patents’ citations to measure
accessories (n = 87), 3D printer parts (n = 76), kitchen technological compatibility (for instance, the compat-
and dining (n = 36), 3D printing (n = 34), mobile phone ibility index developed by Maurseth and Verspagen
(n = 32), electronics (n = 26), toy and game accessories 2002). Importantly, this measure does not capture all
(n = 21), and organization (n = 20). dimensions of compatibility; examining other dimen-
sions of compatibility is discussed as a direction for
4.1. Measurement future research. Trialability (the degree to which an
The measures (see Appendix A) for Rogers’ (2003) innovation may be experimented with) is measured
five diffusion factors relate to the object itself, or using the number of accommodations granted by the
the object’s Thingiverse page (created by the object’s license chosen by the designer. The objects in our
designer). We do not distinguish between the object data set make use of four different Creative Com-
and its page; it is not possible to view or access the mons licenses, which govern the following activities:
(1) the ability to use the object commercially and
(2) the ability to license a remix under different terms
1
A user-level examination of the designers and remixers of the 100 than the original object. Thus, our scale for triala-
most recently posted objects in our sample indicates that follower
status impacted zero remixes in this subsample. Furthermore, while bility ranges from 0 to 2, indicating the number of
the group feature was not launched until 2014, we examine users’ allowances made for the particular innovation. Inno-
later entry into group membership as a possible indicator of latent vations with more accommodating licenses can be
user-level differences. While, as of June 10, 2015, these groups are experimented with more widely and freely, and thus
somewhat seldom used, in this same subsample, zero objects were are more trialable. Both license restrictions measured
remixed by another user who later joined one of the same groups
as the creator of the original design. Neither examination reveals represent meaningful nontechnical obstacles to exper-
the presence of user-level determinants of remixing within our imentation for members of online innovation commu-
context. nities. “I think they [license restrictions] absolutely
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 9

limit behavior—certainly within the Thingiverse com- Front page presence (promotion of the innovation on
munity” (Mortensen 2015). Consider that commercial the online community’s home page) is measured as an
use of 3D printed objects is very common (for instance, indicator variable: whether or not the design was fea-
there were over 18,000 3D printed objects for sale on tured on Thingiverse’s home page (www.thingiverse
Etsy.com as of February 24, 2016). The first restriction .com). “The front page obviously has a much broader
limits commercial experimentation with the object. In audience than the rest of Thingiverse” (Mortensen
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

cases where this license restriction is absent, commer- 2015). In terms of driving community awareness, fea-
cial use of the object is unconstrained. The second turing leads to an innovation getting “a massive boost,
restriction limits experimentation with future remixes, probably 0 0 0 several orders of magnitude” (Mortensen
which is a particularly relevant form of experimen- 2015). This increase in traffic is essentially the goal of
tation given our research question. We do not claim online promotional tactics such as display ads. Once a
that this measure exhaustively captures all forms of design is featured (generally within 48 hours of being
experimentation with an innovation (i.e., all forms of publicly posted), a badge indicating the design’s fea-
trialability); other possibilities are discussed as future tured status is permanently displayed on the design’s
research directions. Also, recall that we have removed Thingiverse page. Note the distinction in measure-
objects from our sample that use licenses explicitly ment between front page presence and observability.
prohibiting remixing. (Since they cannot be remixed, While front page presence impacts how often the inno-
these objects are not informative in addressing our vation is seen, observability refers to how much of the
research question.) Accordingly, it can be fairly said innovation can be seen—a trait of the underlying inno-
that all objects in the sample have at least some degree vation (including the page it resides on). Online com-
of trialability. Observability (the degree to which an munity interaction (the degree to which an innovation is
innovation is available and visible) is measured as extensively discussed online by members of the inno-
the number of images posted by the creator on the vation community) is measured using the number of
innovation’s Thingiverse page. An innovation that has online comments posted pertaining to the innovation
many images posted is considered more observable by community members. All community members
than an innovation with a single image. Innovations (including those who have or are considering printing
with only one posted image may not allow community or remixing the object as well as the original designer)
members to easily view all aspects of the innovation may comment. It is possible for back and forth dis-
(which may provoke uncertainty; Rogers 2003). Other cussion to occur between the object’s creator and a
objects have 10 or more images posted; we consider community member interested in the specific object,
these to be more available and visible to the com- or between two or more community members with-
munity. Complexity (the degree to which an innova- out the creator participating. Discussion often relates
tion is perceived as relatively difficult to understand to requests for help printing the object and settings
and to use by members of a social system) is mea- for various 3D printers, and also includes suggestions,
sured using the number of downloadable files posted praise, and criticism. A count of the comments on
within the design’s Thingiverse page that are neces- innovations’ discussion boards indicates which objects
sary to print the design. Files provided in alternate are being discussed within the community (and con-
formats, resolutions, or sizes are not counted in this versely, which are not).
measure (as they are not strictly necessary to print Remixing (the innovation community’s use of an
the design). While an object requiring only one file existing innovation as source material or inspiration
tends to be simpler, more complex designs can require to aid in the development of further innovations) is
a large number of files to be printed for consumption measured using the cumulative number of members of
or to be manipulated for a remix. Multiple files are the community reporting that they have “remixed” the
commonly used for designs that contain subcompo- source object into another design (Hill and Monroy-
nents (each of which needs to be printed separately). Hernandez 2012).2 Remixing is captured based on
The requirement of multiple time-consuming prints community members reporting their remixing of an
followed by assembly represents additional difficulty object in response to prompts from Thingiverse. The
in use (part of the definition of complexity). Consistent
with this study’s focus on individual objects, this mea- 2
There is no way to directly like the source object from a remixed
sure focuses on the “detail complexity” of the object, object’s page, and thus remixing is not likely to affect the rela-
rather than other forms of complexity (such as inter- tive advantage measure employed here. Thingiverse’s community
dependence or dynamic complexity), similar to oper- and audience development manager was asked whether it was
possible that a source object would gain likes simply by virtue
ations management researchers who have used the of being remixed. He responded, “I can’t see how the original
number of components as an indicator of detail com- would gain anything concrete just by virtue of it being the source”
plexity (e.g., Bozarth et al. 2009). (Mortensen 2015).
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
10 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

structure of the online community as well as cul-

1
tural norms within the community appear to sup-
port the practice of self-reporting. Remixing correlates

00371∗∗∗
closely with downloads (p < 000001), indicating that

1
self-reporting is working effectively. Consistent with
our earlier discussion of the possibility of remixing

00207∗∗∗
00204∗∗∗
without initial consumption in online innovation com-
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

1
munities, we note that 35.2% of the objects in our
sample that are remixed by the community have zero
reported makes (i.e., 3D prints). Community members

00190∗∗∗
00139∗∗∗
00166∗∗∗
are remixing without consuming in at least this por-

1
tion of cases.3
Several control variables are used in our regression
analyses. Consumption (the innovation community’s

00574∗∗∗
00182∗∗∗
00294∗∗∗
00187∗∗∗
use of an existing innovation in its current form) is

1
measured using the cumulative number of members of
the online innovation community reporting that they
have “made” (i.e., 3D printed) the object (a form of

−00120∗∗∗
−00054
00061
00031
−00038
digital consumption; see Belk and Llamas 2013). Note

1
that consumption is used as a control variable, since
consumption of an innovation may prompt remixing
(for instance, if users are dissatisfied with consump-

00122∗∗∗

00116∗∗∗
00140∗∗∗
tion of the source innovation; Rogers 2003). Date posted

00066
00071
00051
3

1
allows us to control for the time that has elapsed since
the design was posted, using the day of the month
(1 to 31); recall that all innovations in our sample were

00219∗∗∗
00199∗∗∗
00182∗∗∗
00372∗∗∗
00468∗∗∗
00083∗
−00025
publicly posted in the same month. Twenty-one cat- 2

1
egory indicator variables are used, with 3D printing
accessories as the default, to control for potential cat-
egory effects.
00259∗∗∗

00183∗∗∗
00116∗∗∗
00139∗∗∗
00296∗∗∗
00390∗∗∗
00091∗∗
00015
1

4.2. Analyses
Our dependent variable—remixing—is a count vari-
able, which is overdispersed in that it contains a dis-
0.0249
Max.

25

1
2
36
24
1
213
27
Descriptive Statistics and Spearman Correlations Among Variables

proportionate number of zero entries. A substantial

portion (82.3%) of innovations in our sample have not
been remixed; similarly, 69.7% of the innovations in
Min.

0
0
0
0
1
1
0
0
0

our sample have no reported consumption. As noted

by Susarla et al. (2012) and Nambisan et al. (1999),
most UGC languishes in obscurity. Because of this pre-
0000459
00446
00562

00201

ponderance of zero values, we report Spearman cor-

SD

1035

3072
1089

10051
2054

relations (with summary statistics, in Table 1), which

can accommodate nonnormality (Neter et al. 1996).
Furthermore, to test our hypotheses, we use negative
0000581

000422
Mean

binomial (NB) regression, a generalization of Poisson

00325

00273

00920
1017
4014
1073

3003

regression suited to this overdispersion (Hilbe 2011,

Sutanto et al. 2013).4 Note that observability is mean
p < 0001.

centered prior to calculating the squared term.

Online community interaction

∗∗∗

3
The practice of remixing without first printing an object could be
Front page presence

p < 0005;
Relative advantage

substantially more common than this. From our data, we cannot

identify other cases in which one community member prints an
Consumption
Compatibility

Observability

object while, independently, another community member remixes

Complexity
Trialability

∗∗
Remixing

the object without printing it.

p < 0010;

4
The Vuong test can be used to detect whether a zero inflation
Table 1

(ZI) model is preferable to NB regression. Unfortunately, ZI mod-

1.
2.
3.
4.
5.
6.
7.
8.
9.

els that otherwise mirror those in Table 2 fail to converge using

∗
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 11

Table 2 Regression Analyses—Dependent Variable: Remixing

Model 1 2 1b 2b

Poisson with Poisson with

Negative Negative endogenous endogenous
Estimation binomial binomial regressors regressors
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Relative advantage 106049∗∗∗ 99080∗∗∗ 96092∗∗∗ 93012∗∗∗

Compatibility 00167 00175 00195 00235
Trialability 00114 00183 000950 00186
Observability 000216 00114∗∗ 000173 00112∗
Observability squared — −000150∗∗ — −000151∗∗
Complexity 000843 00131 000732 00124
Front page presence −00504 −00293 −00439 −00153
Online community interaction 000648∗∗∗ 000546∗∗ 000507∗∗∗ 000500∗∗∗
Consumption 00124∗∗∗ 00133∗∗∗ 00122∗∗ 00113∗∗
Date posted −000198∗ −000173 −000193∗ −000172
Category controls Included∗∗ Included∗∗ Included∗∗∗ Included∗∗
n 498 498 498 498
Log likelihood -269.75 -266.37 — —
LR• 2 135.84 (30)∗∗∗ 142.60 (31)∗∗∗ — —
Pseudo-R2 00201 00211 — —
Hansen’s J — — 000473 0000160
∗ ∗∗ ∗∗∗
p < 0010; p < 0005; p < 0001.

4.2.1. Regression Analysis. Model 1 includes lin- the mean number of images posted for our sample
ear terms only, while model 2 represents a full test of (x̄ = 4014).
the proposed determinants of remixing (see Table 2). Note that if consumption is instead specified as
These models show strong, positive effects of rela- the dependent variable in models otherwise mirroring
tive advantage (p < 0001), supporting H1, and online models 1 and 2, front page presence has a positive
community interaction (p < 0001 in model 1, p < 0005 (p < 0001) effect on consumption, contrasting starkly
in model 2), supporting H7. The inverse-U-shaped with its nonsignificant effect on remixing. Online com-
relationship hypothesized between observability and munity interaction, relative advantage, and compati-
remixing is significant (p < 0005), supporting H4. There bility also positively (p < 0001) impact consumption,
is a nonsignificant relationship between front page with nonsignificant findings for the other variables of
presence and remixing (p > 0010). While this is con- interest (including the linear and nonlinear effects of
sistent with the logic of H6, the nature of hypothesis observability). These results are consistent with exam-
testing makes it impossible to support a conclusion inations of Rogers’ (2003) diffusion factors in tradi-
of a zero effect of front page presence on remixing. tional settings; for instance, Tornatzky and Klein’s
The hypothesized effects of compatibility (H2), trial- (1982) meta-analysis also finds that relative advantage
ability (H3), and complexity (H5) are not supported. and compatibility are significantly related to consump-
Additional investigation of the inverse-U-shaped rela- tion, with nonsignificant effects of trialability, observ-
tionship between observability and remixing shows ability, and complexity.
that the point at which observability has the most pos- In models 1 and 2, consumption is shown to be
itive effect on remixing corresponds with an innova- a significant determinant of remixing. The use of an
explanatory variable that can itself be predicted by
tion having 3.80 posted images, slightly lower than
other predictor variables raises concerns for potential
bias from endogeneity (Baum et al. 2003, Neter et al.
Stata 14.0. This is a common concern with ZI models (e.g., Jacobs 1996). To address this, we reestimate models 1 and
2013). If the category control variables are removed, a ZINB model 2, fitting models accounting for endogenous regres-
will converge. While removing these indicator variables is not
ideal, this approach allows for the Vuong test to be performed. We
sors (similar to Han and Mithas 2013). Here, Pois-
use the parameter-corrected Vuong test as outlined by Desmarais son regression with endogenous regressors (two-step
and Harden (2013) to examine a model that otherwise mirrors generalized method of moments) is used; the results
model 2, without category control variables. The Akaike informa- are shown in Table 2 (models 1b and 2b). Two instru-
tion criterion (AIC)-corrected Z is nonsignificant (1.08, p > 0010); a mental variables (variables that are not in the explana-
nonsignificant finding indicates that the ZI model is not superior,
tory equation but are related with the endogenous
and thus the more parsimonious NB model should be selected.
The Bayesian information criterion (BIC)-corrected Z significantly explanatory variable) are specified: description length
favors the NB model (−3027, p = 000005). Thus, a ZI model is not (the character count of the description of the object
preferable to the NB model employed here. written by the designer) and the total number of
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
12 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

page views for the focal object. These are appropriate the absence of the same condition is consistently asso-
instrumental variables since they only affect remixing ciated with the outcome. fsQCA is an exploratory
indirectly (i.e., through consumption). First, since the technique—it investigates the relationships between
object’s description is typically written by the designer the specified outcome and all possible combinations of
to convince community members of the value of pos- the conditions specified. The ability to uncover combi-
sessing the physical object, we expect that the length of nations of causal conditions that lead to the outcome
the description will only directly influence consump- can produce findings that offer valuable nuance to
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

tion. Next, page views are an indicator of commu- complement regression-based analysis (e.g., Amenta
nity members’ awareness of the object. Many page et al. 2005, Ordanini et al. 2014, Stanko and Olleros
views are generated by peripheral community mem- 2013). fsQCA should not be seen as a test of the
bers, who are only consuming ideas of others and hypotheses put forward here, but rather, as a way to
not adding their own contributions (Crowston and uncover additional configural nuances that cannot be
Howison 2006). While a small number of community inferred from the regression analyses. Note that fsQCA
members typically generate the vast majority of con- has no counterpart to regression’s notion of nonlinear
tributions to innovation communities (Dahlander and effects.
Frederiksen 2012), it is the large number of periph- To begin fsQCA, the data used for the regression
eral community members that are able to generate analysis must be transformed into fuzzy set scores,
substantial page views. The awareness indicated by representing the level of membership in each set (from
page views (alongside the set of explanatory variables 0 to 1). To calculate fuzzy set scores (using fsQCA
hypothesized) is not sufficient to impact remixing 2.5), three anchors must be specified for determin-
other than through its effect on consumption. Instru- ing membership in each condition: the thresholds for
mental variables must also correlate with the endoge- (1) full membership, (2) nonmembership, and (3) the
nous variable and be independent from the error crossover point—the point at which there is maximum
process (Baum et al. 2003). Both instrumental vari- ambiguity about a case’s membership status (Kent
ables significantly correlate with consumption (both 2008, Ragin and Rubinson 2011). For instance, a score
p < 00001). A small Hansen’s J statistic indicates that of 0.75 indicates that an innovation is mostly in the
instruments are uncorrelated with the error term. For set under consideration, whereas a score of 0.25 indi-
both models 1b and 2b, the J statistics (Table 2) are cates that the innovation is mostly outside the set.
very small (both p < 0010), indicating the validity of the Anchors should be set using meaningful theoretical or
instruments. Since this estimation approach does not interpretive rationale, not through the use of sample
substantively change our results (i.e., no coefficient is statistics (Ragin 2007). Appendix B reports the calibra-
significantly different using this estimation based on tion points used in this study, with detailed results
the 95% confidence intervals of the negative binomial and robustness tests. Since there is no way to multi-
coefficients), we can conclude that endogeneity is not plicatively control for the effect of the date posted in
a serious threat in interpreting these results. fsQCA, this control variable is not used in our fsQCA.
4.2.2. Fuzzy Set Qualitative Comparative Anal- Since use of the category control variables specified
ysis. To supplement our regression analyses and in the regression analyses would lead to an uninter-
uncover contextual effects that may be impactful pretable number of configurations (there are 22 cate-
in determining remixing by an online innovation gories), these are also not used in our fsQCA. Next, we
community, we conduct fsQCA using the data set inspect all possible configurations to observe the dis-
described above. fsQCA’s emphasis is on identifying tribution of innovations in our sample. To ensure that
sets of conditions that lead to the outcome in ques- identified solutions are adequately populated, elim-
tion (El Sawy et al. 2010, Fiss 2011, Ordanini et al. inating unpopulated and sparsely populated config-
2014). Though it is outside the scope of this research urations is necessary (Ragin 2008b). In our case, we
to thoroughly introduce fsQCA, several outstanding remove all configurations where n ≤ 1, which results
reviews are available (e.g., Ragin 2008a, Ragin and in maintaining over 80% of our cumulative sample,
Rubinson 2011). fsQCA and regression are dissimi- consistent with Ragin’s (2008a) benchmark. Since rela-
lar but complementary techniques—while regression’s tively few innovations in our sample garner front page
emphasis is on the discrete effect from each indepen- presence (4.2%), removing sparsely populated config-
dent variable, fsQCA uses Boolean algebra to iden- urations results in no configurations that include front
tify configurations of conditions that consistently lead page presence. Accordingly, we remove the front page
to the outcome (Vis 2012). This focus on configura- presence variable from the fsQCA.
tions allows for the detection of complex causal mech- Table B.2 reports fsQCA results. In fsQCA, consis-
anisms. For instance, fsQCA may detect that within tency is the degree to which cases sharing a com-
one configuration, the presence of a condition leads bination of conditions demonstrate the outcome in
to an outcome, but within a second configuration, question (Kent 2008, Ragin 2008b). For all analyses,
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 13

a consistency cutoff of 0.75 is used (Ragin 2008a). includes the interaction of these two terms (observabil-
In fsQCA, coverage is a measure of the proportion ity × complexity). The interaction term is nonsignificant
of cases of interest explained by the causal expres- (p < 0010), with other terms materially unchanged. This
sion. It is analogous to regression’s R2 , the degree to indicates that the contingent effect fsQCA identifies is
which the solution predicts the outcome (Ragin 2008b). only impactful within these specific configurations—
Two explanatory configurations consistently lead to there is no significant interaction of observability and
community remixing. Consistency scores for the two complexity for the sample as a whole. fsQCA’s abil-
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

configurations are 0.79 and 0.78; these configurations ity to identify this contextual, contingent effect that is
usually, but not always, result in remixing. First, inno- not detectable by regression analysis across the entire
vations that are relatively advantageous, compatible, sample illustrates the complementary nature of fsQCA
trialable, nonobservable, noncomplex, characterized and regression (Ragin 2008a).
by high levels of online community interaction, and
consumed tend to be remixed by the community. Sec-
ond, innovations that are characterized by all explana- 5. Discussion
tory conditions specified (see Table B.2) are also likely This study develops and tests arguments theorizing
to be remixed. In fsQCA notation, the midlevel dot an online innovation community’s propensity to remix
represents “and,” with the symbol ∼ representing innovations (i.e., the community’s use of an exist-
“not.” Thus, the configurations associated with remix- ing innovation as source material or inspiration to
ing can be denoted as follows: aid in the development of further innovations), draw-
1. Relative advantage œ compatibility œ trialability œ ∼ob- ing on variables constituent to innovation diffusion
servability œ ∼complexity œ online community interaction œ theory. Theorizing and empirically demonstrating the
consumption; effects of these “remix factors” with a large sample
2. Relative advantage œ compatibility œ trialability œ of innovations is a needed step toward understanding
observability œ complexity œ online community interaction œ this type of knowledge collaboration in online inno-
consumption. vation communities, which is not bound by some of
Many of the findings from this analysis are con- the constraints typical in traditional environments. For
sistent with our regression findings. Relative advan- instance, knowledge collaboration between the creator
tage, online community interaction, and consumption of a source innovation and the community of remix-
are consistently associated with remixing, just as they ers can happen without members knowing each other
were found to be in the regression analyses. There are and without consumption of the source innovation.
also several novel insights generated. Within config- Within online innovation communities, the nature of
uration 2, complexity promotes remixing, indicating what it means for an innovation to diffuse is chang-
that high levels of complexity may prompt remixing ing (Lessig 2008). In addition to passively consuming
under specific circumstances. Within configuration 1, innovations developed by others, community mem-
a low level of complexity is associated with remixing; bers expect to actively engage with innovations, remix-
it is noteworthy that this configuration also is charac- ing to further innovate and make them their own.
terized by nonobservability. Interestingly, within these By understanding these remix factors, our findings
configurations, nonobservable, noncomplex innova- empower innovators to put forward work that is most
tions are likely to be remixed, just as observable and likely to generate an energetic remix response from an
complex innovations are. Simple, nonobservable inno- innovation community.
vations may allow for some learning through remix- Online innovation communities are interactive in
ing that is not otherwise possible (due to the lack of nature. Members can easily take part in publicly vis-
observability) within this configuration. fsQCA refines ible discourse regarding an innovation; these online
our thinking and shows this relationship to be com- communities give user innovators a platform for
plexly contingent: for innovations that are members knowledge sharing. We find that this community inter-
of the other sets indicated, an appropriate alignment action is very influential in determining whether an
of observability and complexity promotes community innovation will be remixed extensively by an online
remixing. Finally, while our regression analyses did community. Publicly visible online discussion per-
not detect significant effects of compatibility or triala- taining to an innovation may contain criticism of
bility, fsQCA indicates that within these two configu- the innovation, which encourages community mem-
rations, both are important to ensuring remixing. bers to address shortcomings of the innovation by
Comparison of the two fsQCA solutions (see remixing it. Discussion among community members
Table B.2) shows that the observability and complexity may also pertain to alternative uses or customizations
conditions change in tandem, with no other differ- for the innovation, which again prompts community
ences between configurations. Accordingly, we ran a remixing. In reputation-conscious online communities
regression model (otherwise identical to model 2) that (Lerner and Tirole 2002), community interaction serves
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
14 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

to bring attention to particular interest areas, which of all skill and knowledge levels while allowing for
likely encourages further related developments (such learning (even from relatively simple innovations that
as remixes). While community interaction is influen- require less observability to be understood; Labay and
tial in determining an innovation’s propensity to be Kinnear 1981). For observable, complex innovations,
remixed by the community, front page presence is not it seems plausible that learning may still be possi-
an effective tool to stimulate remixing. The core users ble through remixing due to the source’s complex-
generating the vast majority of remixes within inno- ity, while observability allows community members’
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

vation communities actively resist this type of pro- experiential learning to begin. Consider also the two
motional effort (Crowston and Howison 2006). Our alignments of observability and complexity that are
findings suggest that untargeted online advertising not associated with remixing within these configura-
will not be efficient in instances where encouraging tions: complex, nonobservable innovations may prove
remixing is the goal. It appears that front page pres- too difficult for community members to understand
ence allows an innovation to achieve a level of aware- and creatively remix. Finally, for noncomplex innova-
ness within the community (which is necessary for tions that are highly observable, remixers’ learning is
consumption), but remixing requires greater impetus limited.
(potentially need- or learning-based), which can be
fostered by extensive interaction between community 5.1. Managerial Implications
members. This research has meaningful implications for inno-
vators and managers concerned with the remixing
The inverse-U-shaped relationship between observ-
of their innovations. First, in instances where remix-
ability and remixing points to the value placed on
ing is sought, promoting the innovation on an online
learning by community members. While some observ-
community’s home page will not be efficient. Where
ability is necessary for innovations to spark interest
remixing is desired, promotional efforts should be
and for community members to begin experiential
focused on tactics generating interaction between com-
learning, too much observability dampens the com-
munity members, such as blogger outreach or oth-
munity’s tendency to remix an innovation. Overly
erwise encouraging community engagement (Füller
observable innovations may act to push some poten-
et al. 2008). Innovators wishing to foster remixing
tial remixers toward passive learning (i.e., observa-
would also be well served to avoid very high levels of
tion) instead of active learning (i.e., remixing).
observability. It is conceivable that a staged approach,
Consumption’s role in determining remixing within
in which community members could see some source
online innovation communities has not yet been thor-
innovation detail immediately with more information
oughly examined. Here, consumption is shown to
becoming available as learning occurs (for instance,
be an important determinant of community remix-
in the form of a user toolkit; Franke and Hader 2014,
ing, but is also shown to not be strictly necessary
Jeppesen 2005), would foster remixing.
to remixing within these communities. The necessity
of prior consumption, which restricts user innovation 5.2. Limitations and Future Research Directions
in traditional diffusion environments, does not con- Based on our goal to understand the determinants
strain remixing within online innovation communities. of an innovation being remixed by an online inno-
Remixing without prior consumption is commonplace vation community, we have investigated constructs
within online innovation communities. That said, the constituent to innovation diffusion theory alongside
consumption of an innovation does prompt remix- two forms of contemporary communication that char-
ing, as members of an innovation community conceive acterize these online communities. We do not claim
of improvements and personalizations based on their this to be an exhaustive set of the determinants of
consumption experiences with the source innovation. remixing; other meaningful predictors of remixing
fsQCA results provide additional configural nuance likely exist. Variables congruent to learning may be a
in our understanding of the determinants of remix- good starting point. Some categories of online com-
ing that could have been only very partially intu- munity interaction (see Kudaravalli and Faraj 2008,
ited based on the regression analyses. For innovations von Krogh et al. 2003) may prove more impactful
that are characterized by relative advantage, com- on remixing; for instance, it is conceivable that the
patibility, and trialability as well as high degrees of posting of particular types of questions (e.g., requests
online community interaction and consumption, those for technical resources) may signal the community’s
innovations that are nonobservable and noncomplex further learning interest that may manifest through
are consistently remixed, as are innovations that are remixing. Commentary containing feedback on the
observable and complex. Within these configurations, innovation may prompt remixes due to heightened
proper alignment between observability and complex- motivation to improve the existing innovation. It is not
ity is impactful on the propensity to remix. Noncom- entirely clear whether positive or negative comments
plex innovations with lower levels of observability regarding the innovation have a greater effect on com-
will be accessible to innovation community members munity remixing.
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 15

It will likely prove worthwhile to apply user-centric well as to Johan Broer and Eric Mortensen (MakerBot) for
(rather than product-centric) theoretical lenses to pre- participating in an interview with the author. Amir Lowery,
dict remixing. Importantly, this study did not attempt Krishnan Narayan, and Will Bernholz provided excellent
to investigate differences between users. Given that research assistance. The author also thanks NC State Uni-
users have different interests as well as levels of exper- versity’s Hunt Library Makerspace. Any remaining errors
tise and influence, these factors may be important in are the responsibility of the author.
determining remixing in other contexts (e.g., Susarla
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

et al. 2012). It is also possible that in UGC contexts with Appendix A. Constructs and Measurement
conspicuous user-level differences, these differences (C = Construct, D = Definition, M = Measurement)
may mediate or moderate the innovation level effects
C Relative advantage
found here. On the other hand, the identity ambiguity
D The degree to which an innovation is perceived as better
inherent to online communities (Faraj et al. 2011) may than the idea it supersedes
have interesting creativity effects and social implica- M Number of likes from members of the innovation community
tions. Investigation of possible nuances in determin- divided by number of page views
ing remixing within a team setting is also warranted. C Compatibility
While we have drawn implications for innovating D Whether an innovation is congruent with existing
firms, further investigation into firm-sponsored activ- understandings of members of a social system
ity is needed (West and O’Mahony 2008). M Indicator variable—Whether the innovation is remixed from
a previous product (1) or not (0)
As Rogers (2003, p. 225) acknowledges, the diffusion
factors will necessarily have slightly different mean- C Trialability
ings across innovation contexts. For instance, there D The degree to which an innovation may be experimented
with
may be other valuable perspectives on complexity; the M The number of allowances made by the license applied to the
innovation’s interdependence (i.e., standalone versus innovation
part of a larger system) may impact its propensity to be
C Observability
remixed. While this study’s measure of compatibility D The degree to which an innovation is available and visible
focuses on compatibility with previously introduced M The number of pictures posted by the creator within the
ideas, other dimensions (such as compatibility with innovation’s Thingiverse page
values and beliefs) may prove impactful on commu- C Complexity
nity remixing. The measure of trialability used here D The degree to which an innovation is perceived as relatively
focuses on nontechnical aspects of experimentation; difficult to understand and to use (by members of a social
other measurement perspectives on trialability are system)
M Number of files required to produce the innovation
warranted. Finally, the measures of consumption and
remixing employed here rely on members of the inno- C Front page presence
D Promotion of the innovation on the online community’s
vation community to collectively report these activities
home page
when prompted. Replication of this study in a UGC M Indicator variable—Whether the innovation was selected to
context in which consumption and remixing are moni- be “featured” on www.thingiverse.com (1) or not (0)
tored by investigators, such as controlled experiments, C Online community interaction
is worthwhile. D The degree to which an innovation is extensively discussed
By identifying factors influential in determining the online by members of the innovation community
innovations that are likely to be extensively remixed M Number of comments posted pertaining to the innovation by
by online innovation communities, this research rep- members of the online community
resents a step forward in understanding this type of C Consumption
knowledge collaboration. There is ample room for fur- D The innovation community’s use of an existing innovation in
ther research in this area, and we hope that researchers its current form
M Cumulative number of times the innovation is reported
will build on these findings in the future. to be “made” (i.e., 3D printed) by members of the online
community
Acknowledgments
C Remixing
The author thanks Rosanna Garcia, Gaia Rubera, Stefanie
D The innovation community’s use of an existing innovation as
Robinson, Claude Rubinson, Jeffrey Schmidt, and Pankaj source material or inspiration to aid in the development of
Setia for thoughtful comments on previous drafts of this further innovations
paper. The efforts of the special issue editors as well M Cumulative number of times the innovation is reported to be
as the entire review team are greatly appreciated. Input remixed (or cited as a starting point) for other innovations
from seminar participants at NC State University and
the Product Development and Management Association Appendix B. fsQCA Calibration, Results, and
Research Forum was also valuable. The author thanks Robustness Tests
Jarkko Moilanen (Statistical Studies of Peer Production) for As is good practice in fsQCA, the thresholds used to cal-
making available a portion of the data used for this study as ibrate the data are reported in Table B.1. Note that only
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
16 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

Table B.1 fsQCA Calibration Thresholds

Threshold for Crossover Threshold for

Condition full membership point nonmembership

Relative advantage 1 0.01 likes from the online community 00005 0

per page view
Relative advantage 2 0.009 000045 0
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Relative advantage 3 0.011 000055 0

Compatibility 1 Is remixed from a previous innovation Is not
Trialability 1 2 license accommodations 1 0
Observability 1 10 images of innovation posted 5 1
Observability 2 9 405 1
Observability 3 11 505 1
Complexity 1 3 distinct files are necessary to 2 1
make the innovation
Online community interaction 1 6 posted comments from members of 3 0
the online community
Online community interaction 2 5 205 0
Online community interaction 3 7 305 0
Consumption 1 2 makes by members of the online 1 0
innovation community
Remixing 1 2 remixes by members of the online 1 0
innovation community

Note. In the first column, “1” indicates original threshold, “2” indicates lower alternative threshold, and “3” indicates higher
alternative threshold.

the thresholds marked “1” correspond to the analysis in the (see Table B.1) to detect whether the configurations reported
body of the article. Calibration anchors should be meaning- are robust to small changes in the calibration of these sets.
ful and interpretable—they should not be set through the Next, 26 replication analyses are manually conducted.
use of sample statistics. Here, we consistently select easily Systematic combinations of the original calibration thresh-
understandable anchor points (i.e., round numbers when olds and the alternatives are specified to observe the impact
possible) consistent with the approach used by Ragin (2007). of changes to the calibration thresholds. As with the origi-
Ten images posted serves as the threshold for full inclu- nal analysis, all configurations in which n ≤ 1 are removed,
sion in the set of observable innovations, and 0.01 likes per and a consistency cutoff of 0.75 is used. Following Skaaning
page view is the threshold for full inclusion in the set of (2011), additional configurations are disregarded. Table B.3
relatively advantageous innovations. For both these sets, the presents replication findings in comparison to the original
crossover point is set at half the threshold for full member- configurations (i.e., Table B.2). Identical indicates that the par-
ship. The crossover point for online community interaction ticular configuration is replicated exactly. Superset indicates
is set at three, since to our judgment this is the point of max- an immediate superset is found, in which one of the condi-
imum ambiguity regarding whether it constitutes enough tions found in the original configuration is relaxed (but not
comments to be characterized as a conversation. Double this reversed).
amount (six) is the threshold for full inclusion. The nature The remix outcome’s first configuration replicates identi-
of the other measures used for this analysis largely dictate cally in 21 of the 26 fsQCAs (81%), with an immediate super-
other calibration decisions due to limited range and/or a set found in the remaining 5 cases. Since these immediate
high number of zero values (see Table B.1). Full results of
the fsQCA are reported in Table B.2. Table B.2 fsQCA: Remixing Outcome
While the methodological triangulation used in this study
represents one form of robustness check, it is also possi- 1 2
ble to specifically test the calibration thresholds used for Relative advantage • •
fsQCA. We follow the approach of Skaaning (2011) to exam- Compatibility • •
ine calibration thresholds for robustness. Alternative thresh- Trialability • •
olds are set close to the original thresholds, making “only Observability ž •
minor changes 0 0 0 to ensure that similar theoretical justifica- Complexity ž •
tions could apply to the original as well as the new anchors” Online community interaction • •
Consumption • •
(Skaaning 2011, p. 395). It is not practical to set alterna-
tive thresholds for compatibility (an indicator), trialability, Consistency 00788 00780
Raw coverage 00132 00126
complexity, consumption, or remixing (these are all integers
Unique coverage 000518 000459
lacking the range to meaningfully lower the thresholds, with
Solution coverage: 0.178
a high number of zero values for consumption and remix-
Solution consistency: 0.732
ing). Thus, we establish alternative thresholds for relative
advantage, observability, and online community interaction Note. • = Causal condition present; ž = causal condition absent.
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 17

Identical
Identical
Identical
supersets relax only one condition of the original configura-

Config2
tion (without reversing any conditions), these would lead to

OCI3
very similar conclusions. The remix outcome’s second con-
figuration replicates identically in 22 of the 26 fsQCAs (85%),

Identical
Identical
Identical
Config1
with an immediate superset found in the remaining 4 cases.
Again, these immediate supersets would lead to very similar
conclusions. This is very strong evidence of the robustness of

Identical
Identical
Identical
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Config2
both explanatory configurations found to consistently lead
Observability 3

OCI2 to remixing.

Superset
Identical
Identical
Config1

References
Agarwal R, Prasad J (1998) The antecedents and consequents of
Identical
Identical
Identical
Config2

user perceptions in information technology adoption. Decision

Support Systems 22(1):15–29.
OCI1

Amabile TM (1996) Creativity and Innovation in Organizations (Har-

Identical
Identical
Identical

vard Business School, Boston).

Config1

Amenta E, Caren N, Olasky SJ (2005) Age for leisure? Political

mediation and the impact of the pension movement on U.S.
old-age policy. Amer. Sociol. Rev. 70(3):516–538.
Superset
Identical

Identical

Arts JW, Frambach RT, Bijmolt TH (2011) Generalizations on con-

Config2

sumer innovation adoption: A meta-analysis on drivers of

intention and behavior. Internat. J. Res. Marketing 28(2):134–144.
OCI3

Barrett M, Heracleous L, Walsham G (2013) A rhetorical approach

Superset
Identical

Identical
Config1

to IT diffusion: Reconceptualizing the ideology-framing rela-

tionship. MIS Quart. 37(1):201–220.
Baum CF, Schaffer ME, Stillman S (2003) Instrumental variables
and GMM: Estimation and testing. Stata J. 3(1):1–31.
Superset
Superset
Identical
Config2

Belk RW, Llamas R (2013) The Routledge Companion to Digital Con-

Observability 2

sumption (Routledge, London).

OCI2

Bonaccorsi A, Rossi C (2003) Why open source software can suc-

ceed. Res. Policy 32(7):1243–1258.
Superset
Superset
Identical
Config1

Bozarth CC, Warsing DP, Flynn BB, Flynn EJ (2009) The impact of
supply chain complexity on manufacturing plant performance.
J. Oper. Management 27(1):78–93.
Brancheau JC, Wetherbe JC (1990) The adoption of spreadsheet
Superset
Identical

Identical
Config2

software: Testing innovation diffusion theory in the context of

end-user computing. Inform. Systems Res. 1(2):115–143.
OCI1

Buijs J (2003) Modelling product innovation processes, from lin-

ear logic to circular chaos. Creativity Innovation Management
Superset
Identical

Identical
Config1

12(2):76–93.
Cox A (1998) Cathedrals, bazaars and the town council. Slashdot
(October 13), http://slashdot.org/story/98/10/13/1423253/
Identical
Identical
Identical

featurecathedrals-bazaars-and-the-town-council.
Config2

Crowston K, Howison J (2006) Hierarchy and centralization in free

Note. RA, Relative advantage; OCI, online community interaction.

and open source software team communications. Knowledge,

OCI3

Tech. Policy 18(4):65–85.

Identical
Identical
Identical
Config1

Dabbagh N, Kitsantas A (2012) Personal learning environments,

social media, and self-regulated learning: A natural formula
fsQCA Calibration Robustness Analysis

for connecting formal and informal learning. Internet Higher

Ed. 15(1):3–8.
Identical
Identical
Identical
Config2

Dahlander L, Frederiksen L (2012) The core and cosmopolitans: A

Observability 1

relational view of innovation in user communities. Organ. Sci.

OCI2

23(4):988–1007.
Davis FD (1989) Perceived usefulness, perceived ease of use,
Identical
Identical
Identical
Config1

and user acceptance of information technology. MIS Quart.

13(3):319–340.
Desmarais BA, Harden JJ (2013) Testing for zero inflation in count
Identical
Identical

models: Bias correction for the Vuong test. Stata J. 13(4):

Config2

810–835.
Dewett T (2006) Exploring the role of risk in employee creativity.
OCI1

J. Creative Behav. 40(1):27–45.

Identical
Identical
Config1

Dockser Marcus A (2014) Hackathons target health-care ills. Wall

Street Journal (April 5), A6.
Table B.3

El Sawy OA, Malhotra A, Park Y, Pavlou PA (2010) Research

commentary–Seeking the configurations of digital ecodynam-
RA1
RA2
RA3

ics: It takes three to tango. Inform. Systems Res. 21(4):835–848.

 Stanko: Toward a Theory of Remixing in Online Innovation Communities
18 Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS

Faraj S, Jarvenpaa SL, Majchrzak A (2011) Knowledge collaboration Kozinets RV, De Valck K, Wojnicki AC, Wilner SJS (2010) Net-
in online communities. Organ. Sci. 22(5):1224–1239. worked narratives: Understanding word-of-mouth marketing
Fiss PC (2011) Building better causal theories: A fuzzy set approach in online communities. J. Marketing 74(2):71–89.
to typologies in organization research. Acad. Management J. Kudaravalli S, Faraj S (2008) The structure of collaboration in
54(2):393–420. electronic networks. J. Assoc. Inform. Systems 9(10/11):706–726.
Fleming L, Waguespack DM (2007) Brokerage, boundary spanning, Labay DG, Kinnear TC (1981) Exploring the consumer decision
and leadership in open innovation communities. Organ. Sci. process in the adoption of solar energy systems. J. Consumer
18(2):165–180. Res. 8(3):271–278.
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Lakhani KR, Wolf RG (2005) Why hackers do what they do: Under-
Franke N, Hader C (2014) Mass or only “Niche Customization”?
standing motivation and effort in free/open source software
Why we should interpret configuration toolkits as learning
projects. Feller J, Fitzgerald B, Hissam S, Lakhani KR, eds.
instruments. J. Product Innovation Management 31(6):1214–1234.
Perspectives on Free and Open Source Software (MIT Press, Cam-
Füller J, Matzler K, Hoppe M (2008) Brand community members bridge, MA), 3-22.
as a source of innovation. J. Product Innovation Management Lenhart A, Purcell K, Smith A, Zickuhr K (2010) Social media
25(6):608–619. and mobile Internet use among teens and young adults. Pew
Gerlitz C, Helmond A (2013) The like economy: Social buttons and Research Center, Washington, DC.
the data-intensive Web. New Media Soc. 15(8):1348–1365. Lerner J, Tirole J (2002) Some simple economics of open source.
Goh JM, Gao GG, Agarwal R (2016) The creation of social value: J. Indust. Econom. 50(2):197–234.
Can an online health community reduce rural-urban health Lessig L (2008) Remix: Making Art and Commerce Thrive in the Hybrid
disparities? MIS Quart. 40(1):247–263. Economy (Penguin, New York).
Gourville JT (2006) Eager sellers and stony buyers. Harvard Bus. Levitt T (1980) Marketing success through differentiation—Of any-
Rev. 84(6):98–106. thing. Harvard Bus. Rev. 58(1):83–91.
Greenhalgh T, Robert G, Macfarlane F, Bate P, Kyriakidou O (2004) Liu-Thompkins Y, Rogerson M (2012) Rising to stardom: An empir-
Diffusion of innovations in service organizations: Systematic ical investigation of the diffusion of user-generated content.
review and recommendations. Milbank Quart. 82(4):581–629. J. Interactive Marketing 26(2):71–82.
Mahajan V, Muller E, Bass FM (1995) Diffusion of new products:
Haefliger S, Von Krogh G, Spaeth S (2008) Code reuse in open
Empirical generalizations and managerial uses. Marketing Sci.
source software. Management Sci. 54(1):180–193.
14(3):G79–G88.
Han K, Mithas S (2013) Information technology outsourcing and Majchrzak A, Wagner C, Yates D (2013) The impact of shaping on
non-IT operating costs: An empirical investigation. MIS Quart. knowledge reuse for organizational improvement with wikis.
37(1):315–331. MIS Quart. 37(2):455–470.
Hargadon AB, Douglas Y (2001) When innovations meet institu- Makinen SJ, Kanniainen J, Peltola I (2014) Investigating adoption of
tions: Edison and the design of the electric light. Admin. Sci. free beta applications in a platform based business ecosystem.
Quart. 46(3):476–501. J. Product Innovation Management 31(3):451–465.
Hertel G, Niedner S, Herrmann S (2003) Motivation of software Mangold WG, Faulds DJ (2009) Social media: The new hybrid
developers in open source projects: An Internet-based survey element of the promotion mix. Bus. Horizons 52(4):357–365.
of contributors to the Linux kernel. Res. Policy 32(7):1159–1177. Maurseth PB, Verspagen B (2002) Knowledge spillovers in Europe:
Hilbe JM (2011) Negative Binomial Regression (Cambridge University A patent citations analysis. Scandinavian J. Econom. 104(4):
Press, Cambridge, UK). 531–545.
Hill BM, Monroy-Hernandez A (2012) The remixing dilemma: The McCarthy R (2012) Our lawyer explains the Thingiverse terms of
trade-off between generativity and originality. Amer. Behav. Sci. service. MakerBot Newsletter (September 26), www.makerbot
57(5):643–663. .com/blog/2012/09/26/our-lawyer-explains-the-thingiverse
Hsu CL, Lu HP, Hsu HH (2007) Adoption of the mobile Internet: -terms-of-service/.
An empirical study of multimedia message service (MMS). Moreau CP, Lehmann DR, Markman AB (2001) Entrenched knowl-
edge structures and consumer response to new products.
Omega 35(6):715–726.
J. Marketing Res. 38(1):14–29.
Jacobs D, Paxton PM, Jackson AL, Malone CA (2013) Murder,
Mortensen E (2015) Telephone interview regarding the Thingi-
political resources, and women’s political success. Social Sci. verse innovation community with Eric Mortensen (Thingiverse
Res. 42(2):513–526. community and audience development manager), June 26,
James KH, Humphrey GK, Vilis T, Corrie B, Baddour R, conducted by Michael Stanko.
Goodale MA (2002) “Active” and “passive” learning of three- Nambisan S, Agarwal R, Tanniru M (1999) Organizational mecha-
dimensional object structure within an immersive virtual real- nisms for enhancing user innovation in information technol-
ity environment. Behav. Res. Methods, Instruments, Comput. ogy. MIS Quart. 23(3):365–395.
34(3):383–390. Navas E (2012) Remix Theory (Springer, Vienna).
Jeppesen LB (2005) User toolkits for innovation: Consumers sup- Neter J, Kutner MH, Nachtsheim CJ, Wasserman W (1996) Applied
port each other. J. Product Innovation Management 22(4):347–362. Linear Regression Models, 3rd ed. (Irwin, Chicago).
Jeppesen LB, Frederiksen L (2006) Why do users contribute to firm- Oborn E, Barrett M, Davidson E (2011) Unity in diversity: Electronic
hosted user communities? The case of computer-controlled patient record use in multidisciplinary practice. Inform. Systems
music instruments. Organ. Sci. 17(1):45–63. Res. 22(3):547–564.
Jeyaraj A, Rottman JW, Lacity MC (2006) A review of the predictors, Ordanini A, Parasuraman A, Rubera G (2014) When the recipe
linkages, and biases in IT innovation adoption research. J. is more important than the ingredients: A qualitative com-
Inform. Tech. 21(1):1–23. parative analysis (QCA) of service innovation configurations.
J. Service Res. 17(2):134–149.
Keen A (2008) The Cult of the Amateur: How Blogs, MySpace, YouTube,
Orlikowski WJ (2007) Sociomaterial practices: Exploring technology
and the Rest of Today’s User-Generated Media Are Destroying Our
at work. Organ. Stud. 28(9):1435–1448.
Economy, Culture and Values (Random House, New York). Pavlou PA, Fygenson M (2006) Understanding and predicting elec-
Kent R (2008) Using fsQCA: A brief guide and workshop for tronic commerce adoption: An extension of the theory of
fuzzy-set qualitative comparative analysis. Workshop note, planned behavior. MIS Quart. 30(1):115–143.
University of Stirling, Stirling, UK. Poetz MK, Schreier M (2012) The value of crowdsourcing: Can
Klein J (2013) Reputation Economics (Palgrave Macmillan, New users really compete with professionals in generating new
York). product ideas? J. Product Innovation Management 29(2):245–256.
Kolb DA (1984) Experiential Learning: Experience as the Source of Ragin CC (2007) Fuzzy sets: Calibration versus measurement. Col-
Learning and Development (Prentice-Hall, Englewood Cliffs, NJ). lier D, Brady H, Box-Steffensmeier J, eds. Methodology Volume
 Stanko: Toward a Theory of Remixing in Online Innovation Communities
Information Systems Research, Articles in Advance, pp. 1–19, © 2016 INFORMS 19

of Oxford Handbooks of Political Science (Oxford University Press, Straub ET (2009) Understanding technology adoption: Theory and
New York), 174–198. future directions for informal learning. Rev. Educational Res.
Ragin CC (2008a) Redesigning Social Inquiry (University of Chicago 79(2):625–649.
Press, Chicago). Susarla A, Oh JH, Tan Y (2012) Social networks and the diffusion
Ragin CC (2008b) User’s Guide to Fuzzy-Set/Qualitative Comparative of user-generated content: Evidence from YouTube. Inform.
Analysis (University of Arizona, Tucson, AZ). Systems Res. 23(1):23–41.
Ragin CC, Rubinson C (2011) Comparative methods. Badie B, Sutanto J, Palme E, Tan CH, Phang CW (2013) Addressing the
Berg-Schlosser D, Morlino L, eds. International Encyclopedia of personalization-privacy paradox: A field experiment on smart-
Downloaded from informs.org by [128.197.26.12] on 15 September 2016, at 01:48 . For personal use only, all rights reserved.

Political Science (Sage, London), 331–341. phone users. MIS Quart. 37(4):1141–1164.
Rifkin J (2014) The Zero Marginal Cost Society: The Internet of Things, Tornatzky LG, Klein KJ (1982) Innovation characteristics and inno-
the Collaborative Commons, and the Eclipse of Capitalism (Macmil- vation adoption-implementation: A meta-analysis of findings.
lan, New York). IEEE Trans. Engrg. Management 29(1):28–45.
Roberts JA, Hann IH, Slaughter SA (2006) Understanding the moti- Venkatesh V (2000) Determinants of perceived ease of use: Integrat-
vations, participation, and performance of open source soft- ing control, intrinsic motivation, and emotion into the TAM.
ware developers: A longitudinal study of the Apache projects. Inform. Systems Res. 11(4):342–365.
Management Sci. 52(7):984–999. Venkatesh V, Morris MG, Gordon BD, Davis FD (2003) User accep-
Rogers EM (2003) Diffusion of Innovations, 5th ed. (Free Press, New
tance of information technology: Toward a unified view. MIS
York).
Quart. 27(3):425–478.
Rullani F, Haefliger S (2013) The periphery on stage. Properties,
Vis B (2012) The comparative advantages of fsQCA and regression
functions, and dynamics of the periphery in the free/open
analysis for moderately large-N analyses. Sociol. Methods Res.
source software community. Res. Policy 42(4):941–953.
Setia P, Rajagopalan B, Sambamurthy V, Calantone R (2012) 41(1):168–198.
How peripheral developers contribute to open-source software von Hippel E (1988) The Sources of Innovation (Oxford University
development. Inform. Systems Res. 23(1):144–163. Press, New York).
Shah SK (2006) Motivation, governance, and the viability of hybrid von Krogh G, Spaeth S, Lakhani KR (2003) Community, joining,
forms in open source software development. Management Sci. and specialization in open source software innovation: A case
52(7):1000–1014. study. Res. Policy 32(7):1217–1241.
Skaaning S-E (2011) Assessing the robustness of crisp-set and von Krogh G, Haefliger S, Spaeth S, Wallin MW (2012) Carrots
fuzzy-set QCA results. Sociol. Methods Res. 40(2):391–408. and rainbows: Motivation and social practice in open source
Snodgrass JG, Vanderwart M (1980) A standardized set of 260 pic- software development. MIS Quart. 36(2):649–676.
tures: Norms for name agreement, image agreement, familiar- West J, O’Mahony S (2008) The role of participation architecture
ity, and visual complexity. J. Experiment. Psych.: Human Learn. in growing sponsored open source communities. Indust. Inno-
Memory 6(2):174–215. vation 15(2):145–168.
Ståhlbröst A, Bergvall-Kåreborn B (2011) Exploring users motiva- Yoo Y, Henfridsson O, Lyytinen K (2010) Research commentary–
tion in innovation communities. Internat. J. Entrepreneurship The new organizing logic of digital innovation: An agenda for
Innovation Management 14(4):298–314. IS research. Inform. Systems Res. 21(4):724–735.
Stanko MA, Olleros X (2013) Industry growth and the knowledge Zhang X, Bartol KM (2010) Linking empowering leadership and
spillover regime: Does outsourcing harm innovativeness but employee creativity: The influence of psychological empower-
help profit? J. Bus. Res. 66(10):2007–2016. ment. Acad. Management J. 53(1):107–128.
 Journal of Manufacturing Technology Management
Value-driven clustering of industrial additive manufacturing applications
Filippo Fontana, Christoph Klahn, Mirko Meboldt,
Article information:
To cite this document:
Filippo Fontana, Christoph Klahn, Mirko Meboldt, (2018) "Value-driven clustering of industrial
additive manufacturing applications", Journal of Manufacturing Technology Management, https://
doi.org/10.1108/JMTM-06-2018-0167
Permanent link to this document:
https://doi.org/10.1108/JMTM-06-2018-0167
Downloaded on: 11 December 2018, At: 00:59 (PT)
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

References: this document contains references to 71 other documents.

To copy this document: [email protected]
The fulltext of this document has been downloaded 7 times since 2018*
Access to this document was granted through an Emerald subscription provided by emerald-
srm:178665 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.

*Related content and download information correct at time of download.

 The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1741-038X.htm

Industrial AM
Value-driven clustering applications
of industrial additive
manufacturing applications
Filippo Fontana
D-MAVT, Eidgenossische Technische Hochschule Zurich, Zurich, Switzerland
Received 5 June 2018
Christoph Klahn Revised 14 September 2018
27 October 2018
IPDZ, Inspire AG, St Gallen, Switzerland, and Accepted 12 November 2018
Mirko Meboldt
D-MAVT, Eidgenossische Technische Hochschule Zurich, Zurich, Switzerland

Abstract
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

Purpose – A prerequisite for the successful adoption of additive manufacturing (AM) technologies in industry
is the identification of areas, where such technologies could offer a clear competitive advantage. The purpose of
this paper is to investigate the unique value-adding characteristics of AM, define areas of viable application in a
firm value chain and discuss common implications of AM adoption for companies and their processes.
Design/methodology/approach – The research leverages a multi-case-study approach and considers
interviews with AM adopting companies from the Swiss and central European region in the medical and
industrial manufacturing industries. The authors rely on a value chain model comprising a new product
development process and an order fulfillment process (OFP) to analyze the benefits of AM technologies.
Findings – The research identifies and defines seven clusters within a firm value chain, where the
application of AM could create benefits for the adopting company and its customers. The authors suggest
that understanding the AM process chain and the design experience are key to explaining the heterogeneous
industrial maturity of the presented clusters. The authors further examine the suitability of AM technologies
with agile development techniques to pursue incremental product launches in hardware. It is clearly a field
requiring the attention of scholars.
Originality/value – This paper presents a value-driven approach for use-case identification and reveals
implications of the industrial implementation of AM technologies. The resultant clustering model provides
guidance to new AM adopters.
Keywords Value, 3D printing, Technology implementation, Additive manufacturing, Case studies,
Process innovation
Paper type Research paper

1. Introduction
New technologies find their way into real-world value chains, where they offer a leap
forward in terms of value creation. Production technologies are no different than others and
follow the same adoption patterns. In fact, companies adopt new manufacturing
technologies only if their implementation generates a substantial increase in productivity or
if it provides an unprecedented competitive advantage (Reichstein and Salter, 2006;
Schrettle, 2013; Conner et al., 2014). It has been previously demonstrated that innovation in
process and equipment technology can sometimes provide improved quality, shorter
delivery cycles, lower inventory levels and shorter product development cycles. It can
enable unprecedented products, reduce economies of scale, enrich the offered product
portfolio and allow more customer specials (Skinner, 1984).
Evidence supporting these statements is particularly common for additive
manufacturing (AM) technologies. In their first adoption, AM technologies found vast
application in prototyping (Campbell et al., 2012). AM made it possible to physically realize Journal of Manufacturing
Technology Management
and visualize product concepts beyond virtual representations (e.g. CAD) in a simple and © Emerald Publishing Limited
1741-038X
economic manner. The implementation of AM for prototyping purposes allowed the DOI 10.1108/JMTM-06-2018-0167
 JMTM adopting companies to validate products early and avoid costly mistakes. Furthermore,
adopters of AM prototyping technologies could compare the performance and feasibility of
different design alternatives and ease the selection of the most suitable variants for further
development (Lopez and Wright, 2002).
In recent decades, industry has manifested a growing interest in AM technologies and
their broader adoption, mostly in prototyping, and has attracted new investments, therefore
increasing the innovation pace of the AM ecosystem (Wohlers Associates, 2016).
A subsequent surge in the number of available AM technology improvements (i.e. process
reliability and improved availability of production-grade materials) has pushed them ever
more toward industrial production (Bak, 2003).
In this context, this paper investigates the unique value-adding characteristics of AM,
justifying its broader adoption in industry. Through this contribution, the authors elucidate
the underlying motives driving the industrialization of AM technologies and their
increasing adoption for serial manufacturing purposes. Manufacturing companies
considering the introduction of AM face two fundamental questions. The first is how to
determine the potential applications within their company to generate value. The second is
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

how to understand the implications of AM adoption on established processes. Companies

addressing these questions undergo long-learning cycles, and require time, effort and
experimentation to obtain results.
If a firm lacks a structured approach in the implementation of AM by failing to correctly
scope suitable opportunities in its value chain, a successful and timely implementation
might be compromised. In this work, the authors suggest that a value-driven approach to
the identification of AM application domains and its introduction will increase the likelihood
of success, as it proved beneficial in the circumstances of other managerial challenges
(Slywotzky, 1996). In this context, the paper investigates the following research questions:
RQ1. What are the elemental domains of a focal firm value chain where AM can be
applied to generate unprecedented value?
RQ2. What are the implications of AM adoption in the identified elemental domains?
To address these questions, Section 2 reviews the relevant literature in the domain of AM
management. In Section 3, the authors introduce a value-driven conceptual framework
leveraged for the analysis of the case interviews, describe the selected case-study approach and
introduce the methods employed for data collection. Section 4 reports the results of the study in
the form of seven recurring clusters of value creation, presenting the identified value generated
in implementation as well as for a few implications. Section 5 highlights the contribution of the
present work to the existing literature, and Section 6 provides final considerations.

2. Literature review
Relevant studies have been identified in four distinct research streams of the AM
management literature: AM value, barriers to AM implementation, impact of AM
implementation and identification of AM suitable parts and applications.

2.1 AM value
Previous studies showed that AM technologies offered the largest advantages in terms of
enabling unprecedented freedoms of design, and provided greater manufacturing flexibility.
Two fundamental properties were recurrent in the literature, providing the source of many
advantages achieved by the implementation of AM. Such fundamental properties are the
manufacturing complexity advantage and the small lot-size advantage (Holmström et al.,
2010; Petrovic et al., 2011; Berman, 2012; Mellor et al, 2014). In literature, manufacturing
complexity advantage is often referred to as “complexity for free” (Gibson et al, 2010;
 Eisenhut and Langefeld, 2013; Weller et al., 2015). However, scholars had already Industrial AM
demonstrated the existence of a relationship between increasing design complexity and applications
increasing manufacturing costs (Pradel et al., 2018).
Regarding manufacturing complexity advantage, the layer-by-layer characteristics of
AM processes enable the achievement of highly complex geometries with few limitations
imposed by technology-specific design rules (Baumers et al., 2011). Engineers can leverage
the manufacturing complexity advantage to provide customers with more efficient, more
lightweight, more functional and more customized products (Ahuja et al., 2015). When
compared to other manufacturing technologies, AM shows a lower increase in total cost
with increasing design complexity (Hague et al., 2004). Nevertheless, post-processing
activities can be a major driver for increasing overall manufacturing costs for AM,
especially at high degrees of complexity (Baumers et al., 2016).
Regarding small lot-size advantages, the absence of tooling, equipment changeover and
product-specific set-up procedures makes AM favorable over other manufacturing processes
for low-volume series, up to lot-size 1 (Tuck et al., 2008). Pioneers in the AM domain can achieve
cost reductions in the manufacture of high-variety component families (Fontana et al., 2018).
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

The implementation of AM technologies in a company directly influences its value-adding

processes. Two processes are often directly influenced by such an implementation: the new
product development (NPD) process and the order fulfillment process (OFP). Several scholars
have identified and described how the fundamental properties of AM can enable
unprecedented value creation in NPD and OFP. Such value enablers are summarized
in Table I and are further described below.
First, in the domain of NPD, the manufacturing complexity advantage enables engineers
to apply function-driven product design (Gibson et al., 2010; Klahn et al., 2015). Through
function-driven design, engineers have the opportunity to focus on the function of the
designed components, rather than on their manufacturability. Thus, engineers can integrate
design elements and features that are not manufacturable by means of other manufacturing
technologies (Yadroitsev et al., 2007; Vayre et al., 2012).
Second, the small lot-size advantage in NPD enables test-driven, iterative development.
Via iterative development, engineering teams can economically manufacture several test
versions of a product and drastically increase the communication and understanding of the
designed system (Lopez and Wright, 2002). The approach enables new development
paradigms based on rapid design iterations and is driven by physical testing instead of by
meticulous planning.
Third, in OFP, the manufacturing complexity advantage enables value creation through
the application of integral designs. Integral designs allow manufacturers to achieve their
desired functionality with a reduced number of parts. With the application of integral
design, complex monolithic designs enable the substitution of assemblies (Türk et al., 2016).
This approach reduces the need for assembly, simplifies manufacturing processes and
potentially streamlines related supply networks.

New product development process Order fulfillment process

Complexity Function-driven product design (Gibson et al., Integral product design (Türk et al., 2016)
advantage 2010; Klahn et al., 2015; Yadroitsev et al., 2007;
Vayre et al., 2012)
Small lot-size Facilitate test-driven, iterative development Reduce barriers to product variety and just- Table I.
advantage (Lopez and Wright, 2002) in-time (Deradjat and Minshall, 2017; AM value
Ballardini et al., 2018; Ghadge et al., 2018; enablers in a generic
Fontana et al., 2018) firm value chain
 JMTM Finally, the small lot-size advantage in OFP reduces the barriers responsible for surges of
manufacturing costs in the case of high-variety and just-in-time manufacturing (Fontana et
al., 2018). With smaller lot sizes, companies can achieve more responsiveness in production,
reduce delivery lead times, lower inventories and establish operational prerequisites to
address individual customer needs (Deradjat and Minshall, 2017; Ballardini et al., 2018;
Ghadge et al., 2018).

2.2 Strategic challenges and barriers to AM adoption

Previous publications in the domain of AM innovation identified the elements of
incremental, radical and disruptive AM technologies (Steenhuis and Pretorius, 2017).
To capitalize on the innovation potential of AM, companies must overcome several adoption
challenges. Hence, for successful adoption, companies should learn how to reconfigure and
adapt to AM (Khorram Niaki and Nonino, 2017). Several papers tackle these issues
and propose studies on the challenges and the barriers toward the implementation of AM
and propose ways for companies to overcome these limitations.
Some scholars have described the challenges of implementing metal AM, arising in the
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

specific case of mass customization (Deradjat and Minshall, 2017). Their study leveraged a
case-study approach with six companies from the dental industry and drew attention to the
challenges in the domains of corporate strategy, technology, operations, organizations and
external factors faced by adopters. The identified challenges included a limited degree of
automation offered by software, high costs of post-processing and high costs for equipment
maintenance and set-up. The authors concluded that the implementation challenges varied
from case to case according to the implementation phase, the degree of technological
maturity and company size.
Other researchers applied multi-case studies to deepen the understanding of business- and
patent-related factors influencing the adoption of AM for spare-parts manufacturing purposes
(Ballardini et al., 2018). Their studies confirmed the applicability of AM for spare-parts
manufacturing, but further highlighted restrictions posed by technological limitations and
introduced structural industry limitations and law uncertainties as challenges to
implementation. They further highlighted the risk for firms inadvertently approaching
suboptimal AM applications, then engaging in the manufacturing of parts with inferior
quality, being more expensive than other methods. Thus, this paper highlights the relevance
of studies deepening the methods and practices for scoping AM applications.
A multi-case-study approach, using collected data from 21 interviews from
17 companies, has been leveraged to understand and describe AM adoption challenges in
the specific context of small to medium-sized enterprises (Martinsuo and Luomaranta, 2018).
Their study confirmed previously identified challenges, such as a lack of knowledge of AM
design and skepticism of material availability, quality and durability of AM parts. It further
highlighted the dependency of the challenges upon the position of the firm in the supply
chain. A notable challenge emerging from their study, reported by original equipment
manufacturers (OEMs), is the difficulty in the identification of suitable applications for AM.

2.3 AM adoption and impact of implementation

Another relevant research stream in the AM management literature describes the AM adoption
process and addresses the impact of implementation. In this domain, scholars have
investigated the post-implementation effect to the organization, proven processes, supply
chains and operations. Ghadge et al. (2018) undertook a system-modeling approach to simulate
and explore the impact of AM deployment to aircraft spare-parts inventory management.
The study revealed that the responsiveness introduced by AM had beneficial effects on the
mitigation of high-inventory risks, on the increase of service levels and on the reduction of
downtime costs. Rylands et al. (2016) studied the introduction process of AM within companies’
 operations and described the related impact on business. Their research applied a case-study Industrial AM
approach, analyzing the cases of two AM adopters. Their contribution highlighted the effect of applications
shifting value proposition and creating new value streams for the adopting company.
The impact of AM in mechanical engineering and medical industries was explored with
an online survey approach (Muir and Haddud, 2018). There, the authors investigated
the potential impact of AM on inventory performance and customer satisfaction in the
spare-parts supply chain. In this setting, AM technology was found to positively influence
inventory performance, mitigate supply risks, reduce the impact of sudden surges in
demand and reduce stock obsolescence.
Another comprehensive study on supply management processes and components
systematically investigated the impact of AM technology usage in customized-parts
production (Oettmeier and Hofmann, 2016). They performed an empirical analysis via a
multi-case-study approach of two companies in the hearing systems industry. Their paper
identified important implications in the domains of flow management, supplier relationship
management, order fulfillment, customer relationships and returns management. Whereas
other contributions in this stream focused strongly on describing implications in SCM and
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

operations, this publication was the first to include important observations related to
product development. Interviewees reported the influence of AM in an earlier integration of
customer feedback in product development and the impact on internal communications.
This consideration highlighted a current gap in the literature and stressed the need for
further research considering not only the environment of supply chain and operations, but
also the perspective of AM adoption on the NPD process.

2.4 Identification of parts and applications for AM

Several industries have experienced major barriers to AM implementation via the difficulty
to identify promising AM applications (Martinsuo and Luomaranta, 2018). Scholars have
addressed this issue by investigating the common industry sectors of AM adoption and by
proposing decision-support tools, frameworks and methods to ease the scoping of parts and
applications suited for AM.
In the domain of component identification, a top-down approach to systematically identify
AM-feasible spare-parts was proposed in the context of the aviation industry (Knofius et al.,
2016). The component search strategy relies on an analytic hierarchical process, and the
proposed method was validated with a single case study. The method relied on data available
in common company databases, such as: logistic key performance indicators, supplier
information, inventory figures and parts geometries to rank candidates based on their expected
feasibility for AM processes. The method was suitable to assess a large spare-part portfolio and
was used to identify about 1,000 economically and technologically feasible parts for AM
manufacturing. Other scholars suggested a bottom-up search strategy based on a combination
of workshop concepts and a scoring method using multi-criteria analysis (Lindemann et al.,
2015). The study proposed a three-stage selection process based on the assessment of
economical, functional, geometrical, operational and manufacturing process-related
characteristics of the parts. However, search strategies based on the screening and
assessment of existing components could limit the scope to mere manufacturing technology
substitution or the re-engineering of already existing components and assemblies. Hence, it
might prevent the full exploitation of AM potential on a larger scale.
Conner et al. (2014) leveraged a key-attribute analysis framework to provide a detailed
map of possible application domains for AM technologies. The framework relied on three
product dimensions: geometrical complexity, manufacturing volume and degree of
customization. It provided a reference system to categorize potential AM applications in
eight regions. Other scholars have investigated promising industry sectors for AM
 JMTM application (Petrovic et al., 2011; Guo and Leu, 2013). They reported aerospace, automotive,
biomedical and energy sectors as the common adopters of AM technologies.
The analysis of prior publications in this subsection shows that extant literature
considered the search of possible AM applications at the component level or described AM
adoption at an industry-wide level. This highlights the need for further research at a firm level.
Furthermore, the review of literature highlights a lack of value-driven approaches. Therefore,
researchers should investigate the domains of a company value chain, where AM could be
leveraged to enhance value creation. The results of the entire literature review performed by
the authors reveal a literature gap best summarized by the following three statements:
(1) The identification of the right AM applications represents a challenge especially for
OEMs, and the issue should be further investigated.
(2) Despite NPD and OFP being the two processes where AM shows the greatest
potential for value creation, there is a lack of studies jointly considering AM impact
on both NPD and OFP.
(3) There is a need for literature undertaking a value-driven approach to the
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

identification of application domains of AM at the firm level.

3. Methods
3.1 Value-driven conceptual framework
This research analyzes value creation through AM technologies in the context of a generic
focal firm value chain. The authors use a holistic approach to analyze the benefits
originating from AM implementation in both NPD process and OFP. The model considers a
focal firm with an internal research and development unit and managed operations. The
firm is assumed to source raw materials and semi-finished products from a network of
suppliers and to supply its customers (not necessarily the final users) as shown in Figure 1.
Information
Materials
Focal Firm

Design Space Exploration

Concept Design

NPD
Detail Design

Validation

R&D Recurring
Operations Engineering
Customers
Suppliers

Purchasing Manufacturing Sales and

Distribution
Figure 1.
The proposed model
of a generic focal firm
OFP Aftersales and Service
value chain
 This presented value chain framework comprising NPD and OFP describes a relevant Industrial AM
portion of the lifecycle of a product within and beyond the boundaries of the focal firm, and applications
is thus key for deepening the understanding of the implications of AM.
On the upper part of Figure 1, the NPD process ensures that new ideas and business
opportunities are converted into marketable products and services. This process is executed
one time for the introduction of a new product or product family and is often a rather fuzzy
process (Heck et al., 2016). The development of new products entails a certain degree of
uncertainty. Therefore, the process includes several design iterations, feedback loops and
validation steps that converge toward meeting user requirements. Requirements set by users
include haptics, functionality, durability and other criteria, and often evolve to adaptive
requirements. The speed of production and number of new products that can be brought to
the market, given a certain amount of resources or capital, is a direct driver of total NPD
efficiency (Swink et al., 2006). Shareholders often manifest their interest in the reduction of the
development time or in the increase of NPD throughput to increase profitability. When new
products are ready for the market, the development team hand over the output of the
development project to manufacturing. During this phase (i.e. product launch), operations
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

develop and establish standards, procedures, quality control routines and continuous
improvement techniques. The long-term recurring OFP begins, and products are distributed
to the final users. Depending on the product and its characteristics, after-sales and reverse
logistics processes are sometimes established to guarantee customer support.
Regarding value, the scholars of several research streams have provided
several definitions (Srivastava et al., 1999; Van der Haar et al., 2001; Khalifa, 2004;
Sanchez-Fernandez and Iniesta-Bonillo, 2007). For this research, the authors restrict the
definition of value to include the actors of the above introduced value chain framework.
AM implementation can thus influence value creation and create benefits for both the focal
firm and its customers. Therefore, we provide definitions for value for the focal firm and
value for the customer. Value for the focal firm includes improvements that increase the
profitability of the company. Thus, it either reduces costs by making the processes more
efficient or reduces the materials cost because of reduced consumption, or it increases
revenue because of increased sales price or quantity. Value for the customer is anything that
increases customer willingness to pay for a product. Here, the authors define it as the
increased efficiency of the product in use, increased product quality and the achievement of
better service (e.g. delivery lead times) as perceived by the customer. Efficiency of the
product in use can be described as the reduction of total costs of ownership for the customer
(Klahn and Fontana, 2017).

3.2 Research design and selection of respondents

To address the research gaps listed in Section 2, the authors leverage a case-study approach.
Owing to the explorative nature of the research and the interests of the authors in
contributing novel theories, a case-study approach has been selected (Eisenhardt, 1989; Yin,
2013). Both the current lack of literature and the need for observing the issue in common
circumstances suggest the selection of a qualitative case-study method (Eisenhardt, 1989;
Meredith, 1998). A multiple-case-study approach is thus applied to increase external
validity, owing to the availability of multiple cases (Stake, 1995; Voss et al., 2002).
As the basis for data collection, the authors leveraged a network provided by the “Swiss
AM Expo 2016,” the largest Swiss professional fair for AM with about 70 exhibitors
showcasing applications representing the entire central European market. To address the
gap, the authors focused on firms recurrently applying AM technologies in the context of
series production. To ensure industrial relevance, companies engaging in art, research and
education, architecture, construction, jewelry and software development were dropped from
the initial sample. Furthermore, companies directly engaging in the development of AM
 JMTM products with their own managed operations were preferred to fully grasp the influence of
AM on both NPD and OFP. Institutions with whom the authors have already engaged in
professional relationships, or where the authors previously performed research activities
were excluded from the list of candidates for the sake of avoiding confirmation bias and
conflicts of interest. Finally, the focus was set on selecting cases representative for a wide
spectrum of industrial sectors, from companies of varying size to specifically targeting
better-established industry applications. It was assumed that the choice of domains where
AM was already established contributed to the completeness of the analysis. Cases
regarding one-time experiments, preliminary technology developments and early stage
adoption of AM were deliberately excluded from the sample. The final set includes seven
case studies, summarized in Table II. An in-depth description of the selected cases and of the
companies is available in the literature (Meboldt and Fontana, 2016).

3.3 Data collection and analysis

This research employs semi-structured interviews with managers, engineers, suppliers and
practitioners closely related to seven additively manufactured products. The authors
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

conducted interviews with 17 people from several functions related to the same case, thus
enabling triangulation and the collection of more perspectives and increasing the overall
reliability of the study (Patton, 1987). All case interviews were conducted on-site by the same
researchers with expert representatives from the focal firm and their closest AM suppliers.
Interviewing sessions were further combined with site visits and product showcases.
Owing to the heterogeneous background of the interviewees and the diverse industrial
sectors represented in the sample, the authors leveraged a generic and open-ended interview
outline. Thus, the outline was employed as an opener for further discussion, where the
interviewees were given the chance to provide further insights, or endorse statements from
other sources. All interviews covered the following central aspects: an introduction of the
respondents and their business relationship, a description of the implemented AM case, the
circumstances under which the project was initiated, the chances of creating unprecedented
value in the focal firm and for its customers observed in both NPD and OFP throughout the
project, and implications and effects on the focal firm activities after AM implementation.
All interviews were recorded and processed with the help of software specialized in the
analysis and coding of recordings for the purpose of qualitative research (Wittenburg et al.,
2006; Nijmegen: Max Planck Institute for Psycholinguistics, 2018). The relevant parts of the
interviews were transcribed, allowing for an in-depth analysis of the data. Secondary data
were collected from publicly available documents about the companies, site visits, notes
from the authors and product showcases.
The collected data were analyzed in three distinct phases. In the first phase, the authors
focused on each interview recording to isolate discussions and statements related to the
value added by AM. Transcriptions of all relevant portions of the interviews
were performed. In the second phase, the authors performed a cross-case analysis to
identify recurring patterns of value creation, to define a coding strategy and to compile the
value-adding clusters with single statements. In the third phase, the recordings were
analyzed to isolate and transcribe all discussions about the implications observed for each
identified cluster. The authors conducted most of the interviews in German. Results were
translated to English for the sake of documentation. During the translation process,
the authors transcribed as accurately as possible the original German statements. Therefore,
the form of the translated statements might not fully respect English grammar.
The analysis identified 51 statements about unprecedented value obtained from the
adoption of AM and the implications in the observed firms’ value chains. To enable more
in-depth discussions of the investigated research questions, the results of the study were
reported in the form of a cross-case analysis (Yin, 2013).
 Downloaded by INSEAD At 00:59 11 December 2018 (PT)

Units Interview
Company Sourcing produced duration
Case Product size Industry Interviewees Respondents positions strategy AM technology per year (min) Customers

Case A Vibratory bowl Medium Automation, 3 CEO Buy Selective laser 1–10 54 B2B
feeders industrials AM services supplier: sintering
head of R&D
AM services supplier:
executive VP
Case B Dental aligners Medium Dental, 3 Executive VP Make Stereolithography, W1,000 76 B2B
medical Medical treatment thermoforming
specialist
AM system supplier:
manager
Case C Surgical cutting Large Medical 3 Department head Buy Selective laser Undisclosed 45 B2C
guides Surgeon sintering, fused
AM services supplier: deposition
engineer modeling
Case D Tramcar lightbox Large Transportation 2 Engineer Buy Selective laser 100–1,000 33 B2B
indicator gears AM supplier: head of sintering
production
Case E Injection molding Medium Electronics 2 Engineer Buy Selective laser 1–10 49 B2B
feeding device manufacturing AM supplier: head of sintering
production
Case F Structural inserts Large Geospatial, 2 Manager Buy Selective laser 100–1,000 44 B2B
for UAV frame aerospace AM supplier: engineer sintering frames
Case G Structural Small Orthopedics, 2 CEO Buy Selective laser 10–100 68 B2B
components for ergonomics AM services supplier: melting
ergonomic business development
exoskeleton
applications
Industrial AM

information on the
Table II.

selected cases
Background
 JMTM 4. Results
Research identified seven elemental domains within a firm value chain, where the
application of AM could create benefits for the adopting company or for its customers.
These are listed with major findings in Table III. In each application cluster, the authors

Value-adding cluster Direct AM effect Implications on focal firm

Prototyping Quick iterations over different designs

Faster time to market
and concepts Ensuring that design will perform correctly
Trial versions with other non-series-like
in production conditions
materials to test product features New patterns of collaboration and team
Visualize important product dynamics in the product development team
characteristics not captured in CAD (e.g. agile, parallelization)
Enhanced designs Design follows the function, reduced Design impacts the performance of the
focus on the manufacturing constraintsproduct once in use (energy consumption,
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

Design practices aiming at weight throughput, ergonomic, usability, etc.)

reduction Design impacts the performance of the
Design practices aiming at integrationmanufacturing process and improve
of functions production of the product (variation,
assembly time, quality, etc.)
Incremental product Flexibility to update the design, even Learning from customer feedback reduces
launch after selling of first unit uncertainty for entry in unexplored markets
Avoidance of lock-in in tools and molds Smooth transition between product
generations
Requirement of new business models for
selling the product
Custom products Increasing convenience and simplicity Drastic increase of productivity in the
of manufacturing manufacturing of individual products
Reduced transaction costs in digital Often require the combination with a fully
domain (data can be stored, digital process chain
transmitted, accessed, modified at any- Increasing challenges in traceability
time and at very reduced costs) Late physicalization of products
Single point of storage of information
including meta-data
Hard-coded error detection and quality
assurance gates
Improved delivery Cost efficiency for small lot sizes, even Reduction in delivery lead times and in costs
in high variety Operation with less stock due to enhanced
Saves waiting caused by the absence of responsiveness
specific equipment changeover
Suitability for spare parts with reverse
engineering
Production tools Economic production of molds or tools Reduction of error rate, especially in
for indirect processing combination with manual operations.
Fast accessibility, reduced downtime Improvement of process stability,
ergonomics, etc.
Increasing overall throughput and cycle time
Process concentration Reduction of total process steps and Better scalability due to, one-time design
squeezed value creation overhead
Table III. Elimination of handover and of waiting Substitution of manual work as a major
Summary of identified time between separated process driver of profitability
value-adding clusters Transition from several fragmented Possibility of extending the portion of in-house
based on manual manufacturing steps to one value creation
interviews with integral manufacturing process De-skilling/re-skilling, reduction of skill
technology adopters barriers for workers
 identified direct effects caused by the application of AM and consequences of Industrial AM
implementation on other surrounding activities. applications
4.1 Prototyping
Prototyping simulates physical artifacts intended to reproduce a portion or the entirety of
characteristics of a final product to test and validate during development. Despite three
decades pioneering AM, interviewees reported prototyping as a relevant value-adding
application domain. According to interviewees, the direct advantages of adopting AM in
prototyping lay in the ability to perform quick design iterations and to, therefore, compare
the feasibility of different design approaches. An interviewee, responsible for the
development of case A, reported:
Adaptations can be achieved very quickly […] to manufacture such precise sliders, we
should perform a test, so that we have an idea of how the process performs, and then maybe
adapt some measures […] Here, one could manufacture more versions of the part, […] and then
test them all.
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

Via prototyping, adopters can reproduce important characteristics of the final product and,
therefore, gather new perspectives and insights on their design or scoped systems.
These are not offered by means of other techniques (e.g. CAD), even without applying the
final production-like materials. The interviewed engineer from Case E further stated:
[…] and this is for the purpose of validating automation. In order to run the first trials, we made a
plastic version.
Furthermore, from the medical industry, a surgeon in charge for the design of Case C stated:
[AM] allows us a complete new perspective […], that we cannot obtain with conventional 2D
imaging technologies.
According to interviewees, the adoption of AM in NPD in the field of prototyping has
three major implications. First, they report a change in the collaboration and team
dynamics in NPD from a plan-driven set-up to a more agile-driven approach. An engineer
from Case E stated:
We recently observed that, [during the development of a conventional tool manufactured by means
of CNC, we had to […] sit together to ensure everything works. It’s not possible to test a lot as we do
[here with AM]. There, [in the conventional case] we have to really discuss together in team and rely
on our expertise. With AM we do not need a lot of help from the collaborators. It works best to test!
[…] In teams, we have to sit together and discuss less.
The teams define interfaces, parallelize development tasks, rely on testing and therefore
achieve more efficiency with NPD. Compared to the development of conventionally
manufactured tools, where the team must extensively rely on the expertise of all members
and discuss each feature of the mold to avoid costly errors. In the case of AM, engineers
can work in parallel, and they have shorter sessions together for interface definition and
integration of independently developed elements. Second, physically testing the
performance of the designed system allows for the reduction of uncertainty, flaws and
errors in both production and the final system. A surgeon from Case C reported:
A big part of increasing importance is this 3D planning, from this we derive all such […] models.
Here it is possible to create for example prototypes […] and actually test how these perform.
Are these applicable? Do they work in the available entryway? These are all domains where this
technology plays an important role!
Therefore, highlighting the feasibility of prototyping technologies for the validation of the
designed system helps to verify its usability by simulating real production conditions.
 JMTM Third, interviewees reported the feeling of an increase in the pace and speed of
NPD, thanks to the other findings mentioned above. The CEO of the company from
Case G reported:
I mean, 3D printing is quite flexible and it helps obviously, you don’t want to spend time trying to
sort of make something. If you lose the time window you are done. […] First you ask people [to
manufacture a test system, and they reply:] uhm […] four weeks, five weeks using CNC. And then
we start to see, 3D printing takes about a week or even less.
This can relate directly into faster time to market for the developed products.

4.2 Enhanced designs

AM’s capability of generating complex and unprecedented structures is well understood and
documented in literature (Bletzinger and Ramm, 2001; Hague et al., 2004; Horn and Harrysson,
2012; Huang et al., 2016; Glasschroeder et al, 2015). In the application cluster of enhanced
designs, companies can leverage this AM process capability with the creativity of engineers to
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

enhance the functionality of the product. By leveraging design for AM practices, engineers
can follow an approach where the function of the designed system directly defines its shape,
and the conceptualization process is freed up from many manufacturing constraints.
Engineers can leverage such properties to achieve substantial mass reductions, especially in
the aerospace sector (Emmelmann et al., 2011). Scholars have identified three features allowing
the reduction of the weight of parts: bionic design, lattice structures and thin-walled structures
(Orme et al., 2017). According to interviewees, engineers can also pursue functional
integration. The head of research and development from Case A stated:
We try to integrate as many functions as possible in the bowl. This means, air ducts, vacuum ducts
to hold the part steady, or to blow them away and such things; also transition elements. […] If we
leverage this art of functional integration, then the process is faster.
Functional integration can achieve the incorporation of mechanical functions, fluid and
thermodynamic functions or electrical functions (Glasschroeder et al., 2015). Leveraging weight
reduction and functional integration in the design and engineering of components has important
downstream implications in the focal firm value chain. First, interviewees reported that such
design measures could impact product performance once in use. Here, in the scope of Case A,
functional integration reduced changeover time. The CEO of the Case A focal firm reported:
Changeover time is of course much smaller. Two screws, and the bowl is changed. Usually
[speaking of conventional feeders] no blue-collar workers in the line can change and calibrate the
bowl feeders. When we apply several different formats in a line, let’s say two or three, then these
have to be changed and calibrated by specialized workforce. When these are set-up with just two
screws [as the case of AM feeders] and two pins, then anyone can make the changeover. And this is
of increased interest, and also speed is increased.
However, this is not the only case where improvements of the product in use could be
achieved. Scholars have reported improvements in terms of efficiency, usability and mass
(Gibson et al., 2010; Huang et al., 2016; Klahn et al., 2014; Oettmeier and Hofmann, 2016).
Second, interviewees reported the influence of design for AM on the manufacturing process
of the product. An engineer in the domain of Case F reported:
So, what is nice there, is that the connection pieces they have an internal structure that, […] makes
sure we can glue the carbon rods easily together. […] we came up with a solution with a sort of glue
channels, where we inject glue in one hole, and it’s a sort of trap. […] Furthermore, we know that we
have a specific amount of glue added, and will always be the same on every connection point.
And it’s also a kind of quality check. We know that if we inject the glue, and it comes out on the
other end of the threads, we know that it’s secure and it’s holding together.
 Here, the design complexity can be leveraged to ensure quality in complex assembly Industrial AM
operations, improve cycle time or simplify labor-intensive assembly steps. Therefore, it can applications
positively influence the production process and the product.

4.3 Incremental product launch

With incremental product launches, a company leverages AM to reduce the uncertainty and
increase the speed of a new product launch. Incremental product launch finds an application
at the interface between NPD and OFP at the market-launch step. To pursue incremental
product launch, a company relies on a test-driven, iterative product development approach,
where product increments are frequently released and validated by lead customers on the
market. The feedback from early adopters can be used to fine-tune the product until a stable
design is reached. Through this approach, the uncertainty of NPD can be reduced, time to
market can be shortened and product-to-user fit can be optimized. The interviewees saw the
biggest advantage of AM in this domain in the possibility of updating the design, even after
commercialization without incurring major cost penalties, therefore, in the absence of
tooling lock-in. The team of engineers supporting Case F focal firm reported:
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

The biggest advantage that they have by using AM, is that they can make several iterations of an
existing product […] and in that way, they could launch the product, but they can still change the
design ongoing. They can make changes because they didn’t have any mold, or tools that were
specifically made for producing the parts.
Through this approach, the central implication on NPD is its ability to integrate customer
feedback earlier in the loop:
They sell a drone to their customer and in that way, the customer gives feedback. Yeah, I would like to
add something to the GPS tracker, something extra to the model, and they make a new iteration […].
Therefore, companies can continuously adapt the product to the evolving needs of the
market or can learn more quickly when addressing a new market niche. Interviewees
reported the lack of distinct product generations in the case of incremental product
launches. Instead, there was a smooth transition or evolution of the product:
So that was the biggest advantage. They don’t take a lot of stock of these drones, so they can
continuously improve their designs and immediately once the design has been improved, they can
get it into the manufacturing process. The purpose for them is to keep continuing product
development to smoothly go from version A to version B.
The novel approach of iteratively releasing and updating products has commonalities with
patterns known in the field of software. However, in the case of hardware, such an approach
poses major challenges on the pricing model, the supply chain and the communications with
the customers. In fact, customers are usually reluctant to buy products that are going to be
updated soon. To successfully apply such an approach, companies may need to profoundly
alter their business models and move toward a more service-centric or pay-per-use model.

4.4 Custom products

Customization is defined as the alteration of product characteristics, the modification of a
product model or even the one-time design of a product from scratch to specifically meet the
individual needs of a customer. In the field of customization, the major advantage of AM can
be found in the tool-less batch fabrication of lot-size 1 components, where, because of AM
implementation, companies experience an increasing convenience and simplicity in
manufacturing. Practically, if products share a common material, the build envelope of a
machine can be filled with different three-dimensional models with no additional efforts in
set-up, and the batch manufacturing process can be started. Thus, if different product
 JMTM variants fit into a machine, these can be manufactured in the same job. In the domain of
customization, AM is the direct effect of extending the digital process chain to the point of
manufacturing. Digital information can be stored at one single point of access and shared by
all actors in the value chain. The executive vice president of the company producing the
Case B product stated:
The dentist can load the digital model in his practice account, […] and we collect the data from there.
We get a notification from the system once he uploaded his data. We download the dataset with
pictures, scanned 3D models, x-ray and a so-called treatment plan. Meaning, the planner thinks about
how he wants to handle the case, which steps in what order, what should be the end result so that the
teeth are correctly positioned […] Such information is sent to us together with the 3D scan.
Digital models can be transferred, modified and re-worked with much lower transactional
costs and increased speeds in the digital domain compared to their analog counterparts.
Digital work flows applied to the generation of specific customer variants can be standardized
and programmatically coded. Therefore, the underlying process can be executed with built-in
gates for quality assurance. From Case C:
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

[…] information is standardized, so that nothing can be forgotten, and so that the process is
executed in the right order. This is also as a source of security for the patient, to ensure that the
final result is reproducible from patient to patient.
Furthermore, from Case B:
The software provides us the possibility to move the teeth only within a certain feasible range in
each step. In this way, the force applied to the tooth is not too high. Too large forces can lead to the
loss of the tooth.
Owing to the digital process chain, the adopters of AM experience an increase in productivity
in the field of customization. The medical treatment specialist from Case B reported:
[…] the more I invest on this digital process chain, […] the faster it gets. And in the same time the
doctor was able to address 10 cases, he is now able to deal with 20.
It is also clear from the interviews that AM adoption in this domain is often the consequence
of the pre-existence or the compatibility of fully digitalized process chains:
One result is the usage of such surgical cutting guides, which also are born from the trend, that
today surgeons rely more and more on 3D planning. Also, in other cases […] not only in our clinic.
It’s a routine to rely on 3D planning, and a logical consequence that such surgical guides are
manufactured [using AM].
Such statements highlight the fact that AM technologies, for customization, must be
integrated in already digitalized chains. Otherwise, multiple switching between the digital
and the analog domains can impede the full exploitation of benefits. Most productivity
increases are therefore achieved when products are manufactured at the latest point
possible in the process, thus applying most changes and necessary adaptations digitally.

4.5 Improved delivery

AM can be applied in operations to improve and streamline the process of delivering goods
to customers. When applied within the OFP, AM can impact the planning of manufacturing
operations and the execution of the delivery orders. The process of delivering physical
goods to a customer usually involves the managing of information flow, material handling,
production, packaging, inventory and transportation. Applications of AM in this domain
can influence the efficiency and speed of the OFP. Fontana et al. (2018) demonstrated that
the application of AM in the case of small lot sizes and high-variety manufacturing could
substantially reduce the per-item cost and improve delivery lead times. At an operational
 level in manufacturing, improvements offered by AM are facilitated by the batch Industrial AM
manufacturing of lot-size 1 items combined with the absence of equipment changeover. applications
This unique feature can drastically reduce the impact of fixed manufacturing costs on the
total per-item manufacturing costs. Interviewees report a particular suitability of AM
processes for the manufacture of spare-part components, especially with reverse
engineering. The AM supplier and manufacturer of Case D states:
The business unit maintenance and repair contacted us in relation to non-availability of a spare
part. They provided us with an original part, […] but no drawings and no CAD. The problem was
that the supplier went out of business, as typical. The supplier was located in Eastern-Europe. […]
The first iteration was delivered in 5 days. The verification required then 3–4 weeks, because the
part was tested on the real vehicle.
Regarding the impact on established processes, interviewees reported an increase in the
speed of sourcing and a reduction of costs for low-volume series compared to machining.
Here, the engineering team from Case E reported:
It is impossible with conventional manufacturing to achieve the same lead times as here. I would
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

say as a rough estimate it takes three times as much conventionally. […] Costs are for sure lower
than a factor eight to ten.
Interviewees also mentioned the potential of AM implementation in the reduction of stocks
and works in process because to the increased responsiveness of the process. The service
provider and manufacturer of unmanned aerial vehicle frame inserts disclosed:
I think stock-wise there is a big advantage. They don’t have a stock, they have a few frames, and
they know when there are several ones running in production.
This suggests a more pull-oriented set-up in production.

4.6 Production tools

Production tools include all the shape-giving devices that can be applied during the
manufacturing of a product (e.g. molds). As an example, from Case B, a vacuum forming
process is performed with the help of AM produced molds:
The 3D printed model is then used in a vacuum forming device, where sheets with different
thickness are formed.
Second, production tools include non-shape-giving devices, such as jigs and fixtures used to
ease manual operations during production, assembly and quality control. From Case F:
You can use the technology to build a quality fixture for measuring the frames, so when we
assemble the frames […], we assemble them on 3D printed, or partially 3D printed fixtures, in order
to keep track of the tolerances and the exact dimensions of the frame.
Jigs and fixtures are specially designed so that large numbers of components can be
machined, assembled, verified, transported and positioned identically. Via the use of such
devices, better ergonomics, quality and repeatability is achieved during assembly.
Production tools conventionally require complex and time-consuming steps to be produced
and are often manufactured in lot-size 1. AM offers several possibilities to speed up sourcing
and lower the costs of such devices. In Case E, an interviewee reported:
We manufactured this soldering jig quickly. That’s always very practical, because if I try to
machine these geometries, the part ends up very complex and expensive. [With AM] in one or two
days, the parts are delivered.
AM can be applied to the manufacturing of production tools for the processing of
components made of metal (Bassoli et al., 2007), polymers (Levy et al., 2003; Rännar et al.,
 JMTM 2007; Ahn, 2011) and composites, such as carbon fiber reinforced polymers (Türk, Triebe
and Meboldt, 2016).

4.7 Process concentration

In the domain of process concentration, AM technologies can be applied to reduce the total
number of steps needed to manufacture a product, assembly or component. Thanks to the
manufacturing complexity advantage, multiple complex features can be achieved in one or a
few manufacturing steps. The head of research and development of Case A supplier disclosed:
Milled bowls had to be purchased, at least the raw work-piece and then further processed in house.
[…] Now we can integrate all of this in the first SLS manufacturing step. That is an advantage.
Interviewees reported that the substitution of manual-labor steps could be particularly
beneficial for profitability. From Case A:
The more manual work is needed to manufacture a conventional bowl feeder, and the more we are
able to substitute this manual work with an AM process, the more it gets profitable. Here we
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

usually talk about weeks of manual work in a conventional bowl feeder.

The reduction of process steps has a twofold effect on the total cycle time. The processing
time of the excluded operations is subtracted, and the waiting time between the excluded
steps is gained. Therefore, process concentration can achieve major reductions of total cycle
time. This shows the potential to profoundly alter the supply networks of products. Shorter
process chains are less complex, and thus easier to operate. Therefore, there is demand for
integration in a single company. As mentioned in the Case A interview:
Very important, the value chain grew. Before, they used to outsource many milled parts, and now
they are able to perform much more in house. The value-adding in house increased extremely.
When deploying AM for process concentration, design represents a significant part of value
creation. However, once a design is available, it can be digitally reproduced at much reduced cost:
Today much more happens in CAD, and manufacturing a second bowl was a challenge. […] When
one need to manufacture a second bowl, in the conventional case, even if the same person is
manufacturing it, then he needs the same time as the first one.
In simpler terms, more automated processes have implications for employees. The substitution
of process steps with AM not only requires substantial reskilling of the workforce toward
machine operations, but it also allows for de-skilling in the case of simpler and automated
process chains. An example was given in the case of dental aligners:
The staff is trained on the digital method and most of them are able to take scans. There is the new
possibility of delegating within the practice to more unskilled labor such as assistants, and the
check-up visits are also much faster.

5. Discussion
With the first research question the authors addressed the question:
RQ1. What are the elemental domains of a focal firm value chain where AM can be
applied to generate unprecedented value?
Compared to other studies addressing the identification of feasible applications for AM
(Guo and Leu, 2013; Conner et al., 2014; Lindemann et al., 2015; Knofius et al., 2016), and
focusing on component identification driven by product characteristics, or on describing the
industry-wide sectors of application, the present publication delivers an analysis at the firm
level. The authors contributed the proposed clustering model by extending the propositions
suggested by previous research. The development of AM implementation strategies by
 identifying the benefits of AM, with the creation of new assessment criteria for AM Industrial AM
manufacturing, was previously proposed as a key issue to ease the implementation of applications
AM (Martinsuo and Luomaranta, 2018). The current formulation of the clustering model
contributes to this domain and provides guidance to new adopters of AM technologies in
two distinct ways. First, it enables a systematic approach to the evaluation of potential
applications. By leveraging the clusters and their definition, companies can analyze and
evaluate step by step which application domain shows the greatest potential, given the
company’s offered products and established processes. Chances are that the screening can
highlight one or more application clusters that are particularly attractive for the company to
start with. Second, once such clusters are identified, adopters can focus their resources to
accomplish value creation and can therefore select existing parts or engage in the
development of new designs for AM. Regarding the targeted cluster, it can therefore achieve
more effectiveness in the implementation project. This second aspect contributes in the
mitigation of the risk of inadvertently engaging in the manufacturing of components
resulting more expensive or inferior quality goods, as previously highlighted by scholars
(Ballardini et al., 2018). This adds a new layer in the assessment process.
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

The results further highlight that clusters are not mutually exclusive to particular
products. As shown in Table IV, from a single interview focusing on a specific product,
value-adding statements belonging to different value-adding clusters were identified.
This suggests that a single product can be optimized for and therefore leverage
value-adding elements from different domains. However, existing component selection
strategies undertake approaches focusing on the identification of parts residing in a narrow
subset of the whole potential application spectrum identified by this study. A clear example
is the fact that both the top-down (Knofius et al., 2016) and bottom-up approaches
(Lindemann et al., 2015) target the identification of components residing in the scope of the
improved delivery and enhanced designs clusters. Thus, this finding suggests that a certain
relationship might exist between an application cluster and a suitable method for
component selection. To be more precise, each cluster or group might have a certain
component screening procedure, resulting in more effective scoping. Therefore, the
introduction of a new layer given by the clustering approach at the firm level contributes to
AM component identification streams by highlighting the need for the further development
of component search strategies tailored to the other AM application domains.
The second research question asked:
RQ2. What are the implications of AM adoption in the identified elemental domains?
Here, the results confirm some of the impacts previously identified by other publications.
This paper extends previously identified implications of AM adoption in NPD (Oettmeier
and Hofmann, 2016), especially in the application domains of prototyping and incremental
product launches. In these settings, the study confirms both changes of communication in
engineering teams and the earlier occurrence of testing and validation during the
development process. Our study further adds insights at the customer relationship level,
where interviewees confirmed a growing importance of integrating customer feedback early
during the product launch to reduce uncertainty. It identifies a rather smooth transition
between product generations for AM adopters. These two aspects suggest that, to some
extent, agile development techniques for hardware might be best suitable in combination
with AM technologies at this stage of NPD, specifically for incremental product launch.
The study reveals that this interplay requires further investigation. In the domain of custom
products, our study highlights the importance of pre-established digital process chains to
automate the customization process prior to manufacturing. This result agrees with
previous findings, where the degree of automation in software was identified as an
important challenge of AM implementation (Deradjat and Minshall, 2017).
 Downloaded by INSEAD At 00:59 11 December 2018 (PT)

JMTM

Table IV.

case studies
among clusters and
Identified statements
and their distribution
Enhanced Incremental Custom Improved Production Process
Prototyping designs launch products delivery tools concentration

Case A Vibratory bowl feeders Automation, 2 6

industrials
Case B Dental aligners Dental, medical 9 1 1
Case C Surgical cutting guides Medical 2 3 4 1
Case D Tramcar lightbox indicator gears Transportation 4
Case E Injection molding feeding device Electronics 3 1 3 1 1
manufacturing
Case F Structural inserts for UAV frame Geospatial, 1 1 3 1 1
aerospace
Case G Structural components for Orthopedics, 2
ergonomic exoskeleton ergonomics
Total 8 3 4 12 8 7 9
 Other studies listed knowledge-related barriers to the implementation of AM (Mellor et al., Industrial AM
2014; Martinsuo and Luomaranta, 2018), suggesting AM-specific training or applications
experimentation to internalize such knowledge. However, once the technology was
established in production, our study identified an impact on de-skilling or reskilling of
employees in the domain of process concentration. Hence, our work highlights that some
tasks leveraging highly automated AM enabled process chains can be executed by
operators with lower degrees of specialization. Such evidence confirms patterns already
seen in cases of computer integrated manufacturing (Agnew et al., 1997), highlighting their
relevance for the domain of AM.
Shifts in value creation identified in a previous study (Rylands et al., 2016) were
confirmed in the domain of process concentration, where interviewees reported a substantial
growth of in house value creation, after the introduction of self-operated AM machines.
Furthermore, previous studies investigated the effect of technological maturity on the
implementation aspects of AM (Deradjat and Minshall, 2017). The results of our study
provide evidence of different degrees of application maturity in the various identified
AM application clusters. Some clusters, such as prototyping and production tools, are
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

widely applied in industry (Wohlers Associates, 2016), and their implications are well
understood. Clusters (e.g. enhanced designs and improved delivery) have been
demonstrated in pioneering applications and are currently growing in adoption at larger
scales (Holmström et al., 2010; Emmelmann et al., 2011; Sirichakwal and Conner, 2016;
Türk et al., 2016; Türk, Triebe and Meboldt, 2016; Fontana et al., 2018). However, clusters
such as process concentrations, custom products and incremental product launches are
described in the literature, but need to be further understood to disclose their full potential at
an industrial level. Whereas further research and ad hoc studies are required, as a first
attempt to explain such heterogeneous degrees of application maturity among the identified
clusters, the authors suggest two dimensions: the degree of understanding of design for AM
and the degree of expertise in the AM process chain.
Figure 2 shows the prerequisites for applying AM in a specific value-adding cluster
according to the above introduced aspects. As the figure shows, applications for
prototyping and for production tools do not require a profound understanding of the entire
AM process chain or an advanced know-how in AM design. They rather replicate
conventional designs and leverage existing supply chain models. Hence, these clusters
Required Understanding of Design For AM
Deep

Custom Products
Enhanced Designs Incremental Product Launch
Process Concentration
Narrow

Prototyping Improved Delivery

Production Tools
Figure 2.
Two-dimensional
portfolio showing
prerequisites for the
Narrow Deep adoption of AM in a
specified cluster
Required Understanding of AM Process Chain
 JMTM present minor barriers in the implementation and could spread in industry earlier than
other, more demanding clusters. On the contrary, value-adding clusters, including custom
products, incremental product launch and process concentration, are positioned in the upper
end of the portfolio. These application domains are the ones demanding the most advanced
knowledge in AM, mastering of the process chain and, in some cases, even requiring new
business models to be successfully implemented. Therefore, such applications lag in
industrial adoption and present the largest implementation challenges. Value-adding
clusters, such as enhanced designs and improved delivery, require only a major focus of
either process chain or design for AM. It can therefore be positioned in-between.
The study further provides some insights about the priorities of different industry
sectors regarding the adoption of AM. In fact, the cases presented in this study can be split
into two overarching categories: industrial manufacturing and medical. Cases A, D, E and F
are part of industrial manufacturing, whereas Cases B, C and G are part of the medical
category. By observing the distribution of statements found in the interviews reported in
Table IV, further insights can be understood. First, all the statements related to the cluster,
“custom products,” are mentioned by companies in the medical category only. Therefore,
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

industry-wide efforts toward achieving specific solutions are facilitated by establishing fully
digitalized process chains (Oettmeier and Hofmann, 2016; Deradjat and Minshall, 2017).
Second, statements related to the clusters of enhanced designs, incremental launch and
improved delivery are mentioned only by companies residing in the industrial
manufacturing category, and they appear to be a priority. Finally, application clusters of
prototyping, production tools and process concentration show that statements mentioned
by companies reside in both categories.

6. Conclusion
Regarding the industrialization of AM technologies, newly adopting companies demand
tools and methods to structure the technology adoption process. The identification of
high-potential fields of application and the awareness of the implications of AM adoption on
the focal firm level are of central interest to scholars eager to facilitate the spread of AM
technologies in industry. This publication leveraged a multi-case-study approach to identify,
define and describe seven value-adding clusters, where AM technologies could be applied
by a focal firm to create unprecedented value. The study further draws attention to several
implications of AM technology adoption in both NPD and OFP.
The present work contributes to the body of knowledge of AM management in the
specific domains of application identification, AM adoption and impact of implementation.
Contrary to other seminal works in the field (Conner et al., 2014; Achillas et al., 2015;
Lindemann et al., 2015; Muir and Haddud, 2018), the present publication undertook a value-
driven approach, and jointly considers both the operations and the product development
functions of a focal firm. The publication further postulates the suitability of AM
technologies in combination with agile development techniques to achieve incremental
product launches of hardware. The authors highlighted the need for further research
addressing the dynamics of incremental product launch.
As a pioneering work in the domain, this study has some clear limitations. These include
a geographically local sample of observed case studies. Despite Switzerland being
oft-reported among the most competitive and innovative countries in the world (Schwab
et al., 2017) and is therefore a representative country for research in innovation matters.
The observed case studies were selected from a restricted geographical area, and may,
therefore, not be fully representative of the global ecosystem. Second, a larger sample of case
studies might be necessary to perform a solid validation of the presented clustering model to
ensure its completeness and resilience. A further limitation is the strong single technology
focus. Despite SLS being one of the most mature technologies for industrial applications,
 and therefore widely employed, it is necessary to perform further investigations to ensure Industrial AM
the validity and completeness of the results of this study for other AM technologies. applications
Further work is not only limited to more case studies, but can also conceptualize specific
methods and processes leveraging the clustering model to provide procedures for structuring
the process of introducing AM. In this direction, the interplay between value-adding clusters
and cluster-specific methods for business-case considerations can be developed. Furthermore,
component search strategies dependent on the specific definition of clusters can be considered.
The authors identified a different degree of application maturity and suggested two
dimensions to explain the differences of adoption rate. This interplay could also be further
investigated and verified with ad hoc studies.

References
Achillas, C., Aidonis, D., Iakovou, E., Thymianidis, M. and Tzetzis, D. (2015), “A methodological framework
for the inclusion of modern additive manufacturing into the production portfolio of a focused factory”,
Journal of Manufacturing Systems, Vol. 37, Part 1, pp. 328-339, doi: 10.1016/j.jmsy.2014.07.014.
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

Agnew, A., Forrester, P., Hassard, J. and Procter, S. (1997), “Deskilling and reskilling within the labour
process: the case of computer integrated manufacturing”, International Journal of Production
Economics, Vol. 52 No. 3, pp. 317-324, available at: https://doi.org/10.1016/S0925-5273(97)00092-3
Ahn, D.-G. (2011), “Applications of laser assisted metal rapid tooling process to manufacture of
molding & forming tools – state of the art”, International Journal of Precision Engineering and
Manufacturing, Vol. 12 No. 5, pp. 925-938, doi: 10.1007/s12541-011-0125-5.
Ahuja, B., Karg, M. and Schmidt, M. (2015), “Additive manufacturing in production: challenges
and opportunities”, Proceedings Volume 9353, Laser 3D Manufacturing II, SPIE LASE, 2015,
San Francisco, CA, March 16, available at: https://doi.org/10.1117/12.2082521
Bak, D. (2003), “Rapid prototyping or rapid production? 3D printing processes move industry towards
the latter”, Assembly Automation, Vol. 23 No. 4, pp. 340-345, doi: 10.1108/01445150310501190.
Ballardini, R.M., Flores Ituarte, I. and Pei, E. (2018), “Printing spare parts through additive
manufacturing: legal and digital business challenges”, Journal of Manufacturing Technology
Management, Vol. 29 No. 6, pp. 958-982, doi: 10.1108/JMTM-12-2017-0270.
Bassoli, E., Gatto, A., Iuliano, L. and Violante, M.G. (2007), “3D printing technique applied to rapid
casting”, Rapid Prototyping Journal, Vol. 13 No. 3, pp. 148-155, available at: https://doi.org/10.110
8/13552540710750898
Baumers, M., Dickens, P., Tuck, C. and Hague, R. (2016), “The cost of additive manufacturing: machine
productivity, economies of scale and technology-push”, Technological Forecasting and Social
Change, Vol. 102, pp. 193-201, available at: https://doi.org/10.1016/j.techfore.2015.02.015; www.
sciencedirect.com/science/article/pii/S0040162515000530
Baumers, M., Tuck, C., Bourell, D.L., Sreenivasan, R. and Hague, R. (2011), “Sustainability of additive
manufacturing: measuring the energy consumption of the laser sintering process”, Proceedings
of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, Vol. 225
No. 12, pp. 2228-2239, available at: https://doi.org/10.1177/0954405411406044
Berman, B. (2012), “3-D printing: the new industrial revolution”, Business Horizons, Vol. 55 No. 2,
pp. 155-162, doi: 10.1016/j.bushor.2011.11.003.
Bletzinger, K.-U. and Ramm, E. (2001), “Structural optimization and form finding of light weight structures”,
Computers & Structures, Vol. 79 Nos 22-25, pp. 2053-2062, doi: 10.1016/S0045-7949(01)00052-9.
Campbell, I., Bourell, D. and Gibson, I. (2012), “Additive manufacturing: rapid prototyping comes of
age”, Rapid Prototyping Journal, Vol. 18 No. 4, pp. 255-258, doi: 10.1108/13552541211231563.
Conner, B.P., Manogharan, G.P., Martof, A.N., Rodomsky, L.M., Rodomsky, C.M., Jordan, D.C. and
Limperos, J.W. (2014), “Making sense of 3-D printing: creating a map of additive manufacturing
products and services”, Additive Manufacturing, Vols 1-4, pp. 64-76, available at: https://doi.org/
10.1016/j.addma.2014.08.005
 JMTM Deradjat, D. and Minshall, T. (2017), “Implementation of rapid manufacturing for mass
customisation”, Journal of Manufacturing Technology Management, Vol. 28 No. 1, pp. 95-121,
doi: 10.1108/JMTM-01-2016-0007.
Eisenhardt, K.M. (1989), “Building theories from case study research”, Academy of Management
Review, Vol. 14 No. 4, pp. 532-550.
Eisenhut, M. and Langefeld, B. (2013), “Additive manufacturing”, A Game Changer for the
Manufacturing Industry, available at: www.rolandberger.com/en/Publications/Additive-
manufacturing-2013.html
Emmelmann, C., Petersen, M., Kranz, J. and Wycisk, E. (2011), “Bionic lightweight design by laser
additive manufacturing (LAM) for aircraft industry”, Proceedings of the SPIE 8065, SPIE
Eco-Photonics 2011: Sustainable Design, Manufacturing, and Engineering Workforce Education
for a Green Future, Strasbourg, April 29, p. 80650L.
Fontana, F., Marinelli, E. and Meboldt, M. (2018), “Selection of high-variety components for selective
laser sintering: an industrial case study”, Industrializing Additive Manufacturing – Proceedings
of Additive Manufacturing in Products and Applications, Springer International Publishing,
Cham, pp. 238-251, doi: 10.1007/978-3-319-66866-6_23.
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

Ghadge, A. et al. (2018), “Impact of additive manufacturing on aircraft supply chain performance”,
Journal of Manufacturing Technology Management, Vol. 29 No. 5, pp. 846-865, doi: 10.1108/
JMTM-07-2017-0143.
Gibson, I., Rosen, D.W. and Stucker, B. (2010), Additive Manufacturing Technologies: Rapid Prototyping
to Direct Digital Manufacturing, Additive Manufacturing Technologies: Rapid Prototyping to
Direct Digital Manufacturing, Springer, Boston, MA, doi: 10.1007/978-1-4419-1120-9.
Glasschroeder, J., Prager, E. and Zaeh, M.F. (2015), “Powder-bed-based 3D-printing of function integrated
parts”, Rapid Prototyping Journal, Vol. 21 No. 2, pp. 207-215, doi: 10.1108/RPJ-12-2014-0172.
Guo, N. and Leu, M.C. (2013), “Additive manufacturing: technology, applications and research needs”,
Frontiers of Mechanical Engineering, Vol. 8 No. 3, pp. 215-243, doi: 10.1007/s11465-013-0248-8.
Hague, R., Mansour, S. and Saleh, N. (2004), “Material and design considerations for rapid
manufacturing”, International Journal of Production Research, Vol. 42 No. 22, pp. 4691-4708,
doi: 10.1080/00207840410001733940.
Heck, J., Rittiner, F., Steinert, M. and Meboldt, M. (2016), “Iteration-based performance measurement in
the fuzzy front end of PDPs”, Procedia CIRP, Vol. 50, pp. 14-19, available at: https://doi.org/10.10
16/j.procir.2016.04.183
Holmström, J. et al. (2010), “Rapid manufacturing in the spare parts supply chain”, Journal of Manufacturing
Technology Management, Vol. 21 No. 6, pp. 687-697, doi: 10.1108/17410381011063996.
Horn, T.J. and Harrysson, O.L.A. (2012), “Overview of current additive manufacturing technologies and
selected applications”, Science Progress, Vol. 95 No. 3, pp. 255-282, doi: 10.3184/
003685012X13420984463047.
Huang, R., Riddle, M., Graziano, D., Warren, J., Das, S., Nimbalkar, S., Cresko, J. and Masanet, E.R.
(2016), “Energy and emissions saving potential of additive manufacturing: the case of
lightweight aircraft components”, Journal of Cleaner Production, Vol. 135, pp. 1559-1570,
available at: https://doi.org/10.1016/j.jclepro.2015.04.109
Khalifa, A.S. (2004), “Customer value: a review of recent literature and an integrative configuration”,
Management Decision, Vol. 42 No. 5, pp. 645-666, doi: 10.1108/00251740410538497.
Khorram Niaki, M. and Nonino, F. (2017), “Additive manufacturing management: a review and future
research agenda”, International Journal of Production Research, Vol. 55 No. 5, pp. 1419-1439,
doi: 10.1080/00207543.2016.1229064.
Klahn, C. and Fontana, F. (2017), “Impact and assessment of design on higher order benefits”, in
da Silva, F.M., Bártolo, H.M., Bártolo, P., Almendra, R., Roseta, F., Almeida, H.A. and Lemos, A.C.
(Eds), Challenges for Technology Innovation-An Agenda for the Future, CRC Press-Taylor &
Francis Group, Lisbon, pp. 237-242.
 Klahn, C., Leutenecker, B. and Meboldt, M. (2014), “Design for additive manufacturing – supporting the Industrial AM
substitution of components in series products”, Procedia CIRP, Vol. 21, pp. 138-143, available at: applications
https://doi.org/10.1016/j.procir.2014.03.145
Klahn, C., Leutenecker, B. and Meboldt, M. (2015), “Design strategies for the process of additive
manufacturing”, Procedia CIRP, Vol. 36, pp. 230-235, available at: https://doi.org/10.1016/j.
procir.2015.01.082
Knofius, N., van der Heijden, M.C. and Zijm, W.H.M. (2016), “Selecting parts for additive manufacturing in
service logistics”, Journal of Manufacturing Technology Management, Vol. 27 No. 7, pp. 915-931,
doi: 10.1108/JMTM-02-2016-0025.
Levy, G.N., Schindel, R. and Kruth, J.P. (2003), “Rapid manufacturing and rapid tooling with layer
manufacturing (LM) technologies, state of the art and future perspectives”, CIRP Annals-
Manufacturing Technology, Vol. 52 No. 2, pp. 589-609, doi: 10.1016/S0007-8506(07)60206-6.
Lindemann, C., Reiher, T., Jahnke, U. and Koch, R. (2015), “Towards a sustainable and economic
selection of part candidates for additive manufacturing”, Rapid Prototyping Journal, Vol. 21
No. 2, pp. 216-227, available at: https://doi.org/10.1108/RPJ-12-2014-0179
Lopez, S.M. and Wright, P.K. (2002), “The role of rapid prototyping in the product development
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

process: a case study on the ergonomic factors of handheld video games”, Rapid Prototyping
Journal, Vol. 8 No. 2, pp. 116-125, doi: 10.1108/13552540210420989.
Martinsuo, M. and Luomaranta, T. (2018), “Adopting additive manufacturing in SMEs: exploring the
challenges and solutions”, Journal of Manufacturing Technology Management, Vol. 29 No. 6,
pp. 937-957, doi: 10.1108/JMTM-02-2018-0030.
Meboldt, M. and Fontana, F. (2016), “Additive Fertigung in der industriellen Serienproduktion:
Ein Statusreport”, AM Network, available at: www.research-collection.ethz.ch/handle/
20.500.11850/122247?locale-attribute=en; www.amnetwork.ch/img/download/pdf/AM_
Report_16.pdf
Mellor, S., Hao, L. and Zhang, D. (2014), “Additive manufacturing: a framework for implementation”,
International Journal of Production Economics, Vol. 149, pp. 194-201, available at: https://doi.org/
10.1016/j.ijpe.2013.07.008
Meredith, J. (1998), “Building operations management theory through case and field research”, Journal
of Operations Management, Vol. 16 No. 4, pp. 441-454, doi: 10.1016/S0272-6963(98)00023-0.
Muir, M. and Haddud, A. (2018), “Additive manufacturing in the mechanical engineering and medical
industries spare parts supply chain”, Journal of Manufacturing Technology Management, Vol. 29
No. 2, pp. 372-397, available at: https://doi.org/10.1108/JMTM-01-2017-0004
Nijmegen: Max Planck Institute for Psycholinguistics (2018), “ELAN (Version 5.2) [Computer
software]”, available at: https://tla.mpi.nl/tools/tla-tools/elan/ (accessed September 11, 2018).
Oettmeier, K. and Hofmann, E. (2016), “Impact of additive manufacturing technology adoption on
supply chain management processes and components”, Journal of Manufacturing Technology
Management, Vol. 27 No. 7, pp. 944-968, doi: 10.1108/JMTM-12-2015-0113.
Orme, M.E. et al. (2017), “Additive manufacturing of lightweight, optimized, metallic components
suitable for space flight”, Journal of Spacecraft and Rockets, Vol. 54 No. 5, pp. 1050-1059,
doi: 10.2514/1.A33749.
Patton, M.Q. (1987), How to Use Qualitative Methods in Evaluation, Sage Publications, Thousand
Oaks, CA.
Petrovic, V. et al. (2011), “Additive layered manufacturing: sectors of industrial application shown
through case studies”, International Journal of Production Research, Vol. 49 No. 4, pp. 1061-1079,
doi: 10.1080/00207540903479786.
Pradel, P., Bibb, R., Zhu, Z. and Moultrie, J. (2018), “Exploring the impact of shape complexity on
build time for material extrusion and material jetting”, in Meboldt, M. and Klahn, C. (Eds),
Industrializing Additive Manufacturing – Proceedings of Additive Manufacturing in Products
and Applications – AMPA2017, AMPA 2017, Springer, Cham and Zürich.
 JMTM Rännar, L., Glad, A. and Gustafson, C. (2007), “Efficient cooling with tool inserts manufactured by
electron beam melting”, Rapid Prototyping Journal, Vol. 13 No. 3, pp. 128-135, doi: 10.1108/
13552540710750870.
Reichstein, T. and Salter, A. (2006), “Investigating the sources of process innovation among UK
manufacturing firms”, Industrial and Corporate Change, Vol. 15 No. 4, pp. 653-682,
doi: 10.1093/icc/dtl014.
Rylands, B., Böhme, T., Gorkin, R., Fan, J. and Birtchnell, T. (2016), “The adoption process and impact
of additive manufacturing on manufacturing systems”, Journal of Manufacturing Technology
Management, Vol. 27 No. 7, pp. 969-989, available at: https://doi.org/10.1108/JMTM-12-2015-0117
Sanchez-Fernandez, R. and Iniesta-Bonillo, M.A. (2007), “The concept of perceived value:
a systematic review of the research”, Marketing Theory, Vol. 7 No. 4, pp. 427-451, doi: 10.1177/
1470593107083165.
Schrettle, S. (2013), Managing Manufacturing Process Innovation? New Manufacturing Technology
Adoption as a Dynamic Capability, University of St. Gallen, St. Gallen.
Schwab, K. and Sala-i-Martín, X., World Economic Forum (2017), The Global Competitiveness Report
Downloaded by INSEAD At 00:59 11 December 2018 (PT)

2017-2018, World Economic Forum, Cologny.

Sirichakwal, I. and Conner, B. (2016), “Implications of additive manufacturing for spare parts inventory”,
3D Printing and Additive Manufacturing, Vol. 3 No. 1, pp. 56-63, doi: 10.1089/3dp.2015.0035.
Skinner, W. (1984), “Operations technology: blind spot in strategic management”, Wickham Skinner
Interfaces, Vol. 14 No. 1, pp. 116-125.
Slywotzky, A.J. (1996), Value Migration, Harvard Business School Press, Boston, MA.
Srivastava, R.K., Shervani, T. and Fahey, L. (1999), “Marketing, business processes, and shareholder
value: an organizationally embedded view of marketing activities and the discipline of
marketing”, Journal of Marketing, Vol. 63 No. 1999, pp. 168-179, available at: www.jstor.org/
stable/1252110
Stake, R.E. (1995), The Art of Case Study Research, Sage, Thousand Oaks, CA.
Steenhuis, H.-J. and Pretorius, L. (2017), “The additive manufacturing innovation: a range of
implications”, Journal of Manufacturing Technology Management, Vol. 28 No. 1, pp. 122-143,
doi: 10.1108/JMTM-06-2016-0081.
Swink, M., Talluri, S. and Pandejpong, T. (2006), “Faster, better, cheaper: a study of NPD project efficiency
and performance tradeoffs”, Journal of Operations Management, Vol. 24 No. 5, pp. 542-562,
doi: 10.1016/j.jom.2005.09.004.
Tuck, C.J., Hague, R.J.M., Ruffo, M., Ransley, M. and Adams, P. (2008), “Rapid manufacturing facilitated
customization”, International Journal of Computer Integrated Manufacturing, Vol. 21 No. 3,
pp. 245-258, doi: 10.1080/09511920701216238.
Türk, D., Kussmaul, R., Zogg, M., Klahn, C., Spierings, A., Könen, H., Ermanni, P. and Meboldt, M.
(2016), “Additive manufacturing with composites for integrated aircraft structures”,
Society for the Advancement of Material and Process Engineering, Long Beach, CA, 15pp.,
doi: 10.3929/ethz-a-010691526.
Türk, D.A., Triebe, L. and Meboldt, M. (2016), “Combining additive manufacturing with advanced
composites for highly integrated robotic structures”, Procedia CIRP, Vol. 50, pp. 402-407,
available at: https://doi.org/10.1016/j.procir.2016.04.202
Van der Haar, J.W., Kemp, R.G.M. and Omta, O. (2001), “Creating value that cannot be copied”, Industrial
Marketing Management, Vol. 30 No. 8, pp. 627-636, doi: 10.1016/S0019-8501(99)00128-5.
Vayre, B., Vignat, F. and Villeneuve, F. (2012), “Designing for additive manufacturing”, Procedia CIRP,
Vol. 3, pp. 632-637, available at: https://doi.org/10.1016/j.procir.2012.07.108
Voss, C., Tsikriktsis, N. and Frohlich, M. (2002), “Case research in operations management”, International
Journal of Operations & Production Management, Vol. 22 No. 2, pp. 195-219, available at:
https://doi.org/10.1108/01443570210414329
 Weller, C., Kleer, R. and Piller, F.T. (2015), “Economic implications of 3D printing: market structure Industrial AM
models in light of additive manufacturing revisited”, International Journal of Production applications
Economics, Vol. 164, pp. 43-56, available at: https://doi.org/10.1016/j.ijpe.2015.02.020
Wittenburg, P., Brugman, H., Russel, A., Klassmann, A. and Sloetjes, H. (2006), “ELAN: a professional
framework for multimodality research”, in Calzolari, N., Choukri, K., Gangemi, A., Maegaard, B.,
Mariani, J., Odijk, J. and Tapias, D. (Eds), Proceedings of the 5th International Conference on
Language Resources and Evaluation (LREC’06), European Language Resources Association,
Genoa, May.
Wohlers Associates (2016), Wohlers Report 2016, Wohlers Associates, Fort Collins, CO.
Yadroitsev, I., Thivillon, L., Bertrand, Ph. and Smurov, I. (2007), “Strategy of manufacturing
components with designed internal structure by selective laser melting of metallic powder”,
Applied Surface Science, Vol. 254 No. 4, pp. 980-983, available at: https://doi.org/10.1016/j.
apsusc.2007.08.046
Yin, R.K. (2013), Case Study Research, 5th revise ed., Sage Publications, Thousand Oaks, CA.

About the authors

Downloaded by INSEAD At 00:59 11 December 2018 (PT)

Filippo Fontana is Research Associate and PhD Candidate at the Chair of Product Development and
Engineering Design, ETH Zürich since 2015. He received the BSc degree in Mechanical Engineering
and the MSc degree in Industrial Management. Through his research, Filippo Fontana supports
pioneering companies in the field of additive manufacturing to speed up the technology adoption
process. For this purpose, he executes pilot projects together with industrial partners and gathers data
to assess the industrial potential of additive manufacturing. Filippo Fontana is the corresponding
author and can be contacted at: [email protected]
Christoph Klahn is Head of the Design for New Technologies Group at Inspire AG since 2013.
The group is closely linked to the Product Development Group pd|z of ETH Zurich. Since 2008,
Christoph Klahn has been working in various functions on the potentials of additive manufacturing for
series part and the challenges for product development. The aim of his group is to transfer knowledge
about additive manufacturing from research to industry and, together with companies, to identify the
potential of new manufacturing processes in their products and to realize successful AM products.
Mirko Meboldt is Full Professor of Product Development and Engineering Design at ETH Zurich
since 2012. He received his MSc (2002) and his PhD (2008) degrees in Mechanical Engineering from the
University of Karlsruhe (KIT), Germany. In 2008, he started his industrial career at Hilti AG,
Liechtenstein, where he was responsible for global technology and product development processes.
He is interested in human-centered product development and regards the connection of research and
education as key to excellence in training. His main research focuses on the development of new
products in the field of mechanical engineering, biomedical applications and associated technologies.

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]
38 Int. J. Technology Management, Vol. 69, No. 1, 2015

Open innovation 2.0: is co-creation the ultimate

challenge?

Thierry Rayna*
Novancia Business School,
3 rue Armand Moisant, 75015 Paris, France
Email: [email protected]
*Corresponding author

Ludmila Striukova
University College London,
Gower Street, London WC1E 6BT, UK
Email: [email protected]

Abstract: Web 2.0 has considerably increased the possibilities of user

involvement in the production process and, thereby, has given rise to new
forms of co-creation (open innovation with customers). Because the roles of
consumers, now ‘prosumers’, have radically changed, specific challenges have
emerged. The aim of this article is to provide possible routes to address such
challenges. To do so, an input-output typology of co-creation is created. This
typology is then used to discuss the main challenges of co-creation: incentives,
risks and costs, IPRs.

Keywords: open innovation; co-creation; Web 2.0; crowdsourcing;

mass-customisation; 3D printing

Reference to this paper should be made as follows: Rayna, T. and Striukova, L.

(2015) ‘Open innovation 2.0: is co-creation the ultimate challenge?’, Int. J.
Technology Management, Vol. 69, No. 1, pp.38–53.

Biographical notes: Thierry Rayna is a Professor of Economics at Novancia

Business School in Paris. Previously, he spent ten years in the UK, where
he held positions at the Imperial College London, the London School of
Economics, UCL, and the University of Cambridge. His research investigates
the consequences of technological change and digitisation on IP strategies,
business models, and innovation ecosystems. He has served as an advisor for
national and international institutions, as well as for major companies in the
media, telecommunications and cultural industries. He also mentors start-ups.

Ludmila Striukova is a Senior Lecturer in the Department of Management

Science and Innovation at the University College London. Her previous
experience includes working as a market analyst for a statistical agency and
as a researcher at King’s College, University of London. She has published
extensively in the area of innovation and technology management and her
research work has been used in numerous EU and governmental reports. She
also has used her telecommunications engineering background to mentor and
advise technology-based start-ups and multinationals.

Copyright © 2015 Inderscience Enterprises Ltd.

 Open innovation 2.0 39

1 Introduction

Though the concept of open innovation has been increasingly popular, its implementation
can be often challenging (Enkel, 2010; Jaspers and den Ende, 2010). However, put in the
context of Web 2.0, these challenges have been taken to an entirely new level. Web 2.0
has become, quite rightly so, synonymous with user empowerment and user generated
content. Web 2.0 technologies have considerably lowered the barriers to user creation
and, in fact, many Web 2.0 services derive their value almost exclusively from the
creation and content provided by users.
Although the concept of open innovation spreads far further, user-based open
innovation is of significant importance and has been gaining momentum. Open source,
crowdsourcing and mass customisation, just to name a few, are now often part of
corporate strategies and, in some cases, are at the very heart of business models.
Hence both Web 2.0 and (user-based) open innovation correspond to the changing
role of users. From a purely passive role, they have become part of the creative and/or
productive process. The increasingly blurred line between the roles of users as consumer
and producer (and, in particular, the growing importance of this latter role) has changed
the view firms and academics have of users, who are now often branded as ‘prosumers’.
In this context, users are no longer recipients of a finished product, but are, instead,
integral part of a co-creation process.
A greater involvement of users in the production process poses, undoubtedly, new
challenges. Relationships between a firm and its customers are generally ruled by far
looser (if any) contractual obligations than partner firms or even workers inside the firm.
Consumers taking part in open innovation are seldom significantly rewarded for their
contribution. On the other hand, firms do not have the same power of monitoring and
enforcement they would have with industrial partners. Hence, logically, though
co-creation has a very large potential, it also leads to significant challenges in terms of
incentives, costs and risks, and intellectual property rights (IPRs).
Co-creation can take many forms. For instance, co-creation can be conducted with
one customer or a group of them. Customers might have similar roles as the employees of
the firms, or very different ones. Finally, the co-created product can be either custom
made for one individual or mass-produced for the whole market. In each case, the
challenges faced by the firms are likely to differ.
The aim of this article is to investigate the challenges posed by co-creation and
provide a framework to help overcoming these challenges. Since open innovation existed
long before the advent of Web 2.0, the first part of this article (Sections 1 and 2)
investigates the actual changes brought about by internet technologies. Next, a typology
of co-creation is constructed and the specific challenges posed by each type are analysed
with respect to motivations, costs and IPRs.

2 Open innovation and co-creation

Though open innovation is often associated with collaboration among firms,

collaboration with consumers is also an essential part of the concept popularised by
Chesbrough (2003).
40 T. Rayna and L. Striukova

Open innovation can typically take three forms (Gassmann and Enkel, 2004). The
first form, also known as inbound open innovation (Chesbrough et al., 2006), is
‘outside-in’ and consists in integrating suppliers and customers as external source of
knowledge. The second form, ‘inside-out’, or outbound open innovation (Chesbrough
et al., 2006) aims to increase profits by bringing ideas to the outside environment, so that
‘outsiders’ can develop ideas and technologies that are not used by the firm. The last
form is a combination of both forms and is referred to as ‘coupled’ open innovation.
Until recently, open innovation with customers was mainly ‘inside-out’. Companies
were using customers as a source of ideas for new products or improvements of existing
products (von Hippel, 1988; Christensen, 1997; Trott, 2003; von Hippel, 2005). Lately,
however, there has been a shift from simply exploiting customer knowledge to
co-creating knowledge with customers (Sawhney and Prandelli, 2000). Nowadays, open
innovation with customers has even taken the form of a coupled process focused on
co-creation. Customers are involved in generating ideas for new products, co-creating
products with firms, testing finished products and in providing end user product support
(Nambisan, 2002). Furthermore, consumers are no longer simply external sources of
ideas (outside-in), but can also become external paths to market (inside-out) (Baldwin
and von Hippel, 2006; Shah and Tripsas, 2007). Co-creation is the customer-related part
of open innovation.
Co-creation was originally defined by Kambil et al. (1996) as co-creation of value by
a firm’s customers. This definition was later on refined by Piller et al. (2010) who see
co-creation as an active, creative and social collaborative process between producers and
users aimed at creating value for customers.
Since co-creation implies an active role for both consumers and firms, Zwass (2010)
distinguishes between autonomous and sponsored co-creation. Autonomous co-creation
means that individuals or consumer communities produce value through voluntary
activities, which they conduct independently from established organisations (even though
they might be doing so using platforms provided by these organisations). In contrast,
sponsored co-creation relates to co-creation activities conducted by individuals or
consumer communities at the initiative of an organisation. Both autonomous and
sponsored co-creation can take place before and after the launch of the product, for
example, to develop a subsequent version of a product or when mass-customisation takes
place.
In general, co-creation with customers can take place at two different stages in the
production process:
1 Product design: consumers suggest a design, create their own design or
improve/amend an existing design.
2 Product manufacturing (and distribution): consumers manufacture products
themselves (for example, using a 3D printer) and might even distribute it.
The actual contribution of consumers depends on the type of co-creation practice setup
by the firms. Furthermore, even when both options are available, some consumers
participate only in one of the co-creation stages, whereas some both co-design and
co-manufacture.
One of the most famous examples of co-creation is crowdsourcing. Coined by Howe
(2006), crowdsourcing is defined as a new web-based business model that uses a
distributed network of individuals to find creative solutions for existing problems.
 Open innovation 2.0 41

Crowdsourcing practices may vary depending on processes of examining contributions

(Geiger et al., 2011) or marketing applications (Whitla, 2009). Crowdfunding has also
been, over the past years, an increasingly popular way to involve consumers in the
production process. This form of co-creation also relates to consumer involvement in the
production process (though the resource provided by consumers is cash, which is then
used to pay for manufacturing, instead of actual manufacturing).
Mass customisation, which involves the production of personalised or custom/tailored
goods or services, is also often associated with co-creation. Indeed, in many cases,
mass-customisation is a form of co-creation, enabling firms to learn from their customers
and to develop new ideas. However, as noted in Prahalad and Ramaswamy (2004), not all
forms of mass-customisation correspond to co-creation activities. For instance, when a
company designs a set of predetermined options and then manufactures these options on
demand, this is not co-creation, as customers do not have any input (besides indicating
their choice).
While co-creation can take place between companies and individual consumers,
co-creation communities are also formed between consumers. Such communities,
generally referred to as ‘communities of creation’ (Sawhney and Prandelli, 2000) or
‘communities of co-design’ (Piller et al., 2004), create common knowledge and value not
only for the members of the community, but also for ‘outsiders’. Co-creation
communities can be extremely valuable as they can provide a company with a range of
new ideas and help save on R&D, which is especially critical when the development of
products and technologies are beyond a company’s resources.
In the next section, we are going to examine how the introduction of Web 2.0 has
changed the nature of co-creation in particular and open innovation in general.

3 Open innovation and Web 2.0: Something old, something new

The idea of using customers to innovate is hardly a new one. Companies have always
been looking for new sources of innovation and one of such sources has always been
customers, whose ideas have been used to improve products and even to create new ones.
Though the term ‘open innovation’ is relatively a new one, it corresponds to practices
that (at least some) companies have been using for a long time. As companies started to
be more proactive in looking for new ideas, they also started to look for solutions that
would enable them to make this process more effective.
Logically, Web 2.0 has since its advent been considered as one of these solutions.
Though they are many different technologies that qualify as ‘Web 2.0’ technologies, they
all have in common that they enable users to share and publish content online in a very
simple manner. Whereas publishing a simple webpage with a few pictures beforehand
required having dedicated software (to design the page and upload it, or to reframe
pictures) and special skills and knowledge (e.g., HTML, JavaScript, FTP), Web 2.0
platforms, such as Facebook, Instagram, YouTube, enable users to do so without any
software (besides a web browser) or knowledge of web technologies.
Thus, Web 2.0 has considerably lowered the ‘entry cost’ for users to publish and
collaborate online and, thereby, it offers new technological possibilities that enable to
enhance the dialogue, on the one hand, between companies and consumers and, on the
other hand, between consumers themselves. Consequently, Web 2.0 has the potential to
42 T. Rayna and L. Striukova

make it much easier for customers to participate in the improvement of existing ideas and
the development of new ones. Furthermore, Web 2.0 also makes it easier for consumers
to form communities around co-creation, thereby providing economies of scale.
It was, thus, expected that Web 2.0 would not only make it easier for consumers to
have a dialogue with the companies, but also would enlarge the pool of prospective
contributors. Yet, as noted in Fournier and Lee (2009), many initiatives aimed at
integrating Web 2.0 in open innovation strategies have failed to become anything more
than focus groups created in the hope that consumers will bond.
Other successful forms of co-creation based on Web 2.0 have, however, emerged
recently. Interestingly, none of these forms are actually new. Mass-customisation,
crowdsourcing or crowdfunding (just to name a few) all existed before the advent of
Web 2.0. Nevertheless, Web 2.0 has significantly increased the scale of participation,
both on firms’ and consumers’ side.
Nonetheless, this increased participation in co-creation activities is not the only
change, as there are nowadays also more sources of co-creation. For example, in the past,
mass-customisation used to be mainly about asking consumers about their preferences
amongst a range of predetermined options. Today, because it is easier for consumers to
develop each other’s product, mass customisation can be done together with customers
and not simply for customers.
Another significant change in the co-creation process brought about by Web 2.0 is
that, in the past, mostly large companies were able to co-create with their customers,
whereas nowadays, even small firms are able to engage in co-creation. Furthermore,
co-creation with customers used to play a rather minor role in the innovation strategy.
Instead, nowadays, co-creation has increased in importance to such extent that some
companies base their business models entirely on co-creation activities, which would not
have been possible before the advent of Web 2.0 technologies. Beyond known
companies, such as YouTube, Facebook or Threadless, companies like Tattoodo (custom
tattoo designs) and Quirky (develops customer innovations chosen by other community
members) are just some examples of companies that have co-creation as the core of their
business model. In all these cases, companies rely heavily on Web 2.0 technologies and
would not have been able to develop such a business model before.
Yet, successfully implementing open innovation and, in particular, co-creation, is
generally not straightforward and requires addressing significant challenges. Thus, it is
important to examine how co-creation can be motivated and to consider the costs and
risks associated with implementing co-creation strategies. This is particularly the case
with Web 2.0, because the significantly more active part played by consumers has
radically changed their role.

4 Uncovering challenges ahead: a typology of co-creation

It has been acknowledged that co-creation represents a great opportunity for firms and
society alike, to the point that it is considered by some as the ‘next stage’ capitalism
(Ritzer and Jurgenson, 2010; Prahalad and Ramaswamy, 2004; Tapscott and Williams,
2006). However, just like any form of open innovation, co-creation comes with
associated challenges mainly in terms of motivation, costs and IPR issues. Yet, to this
respect, not all forms of co-creation are alike and it is sensible to expect major differences
 Open innovation 2.0 43

in regard to the type of challenges, for instance, mass-customisation and crowdsourcing

might face.
As mentioned in the previous section, a key issue of co-creation is whether the roles
and actions of consumers in the production process can be identified and measured. With
regard to the co-creative process, the roles of the firms and the consumers can either be
‘vertically integrated’, in which case firms and consumers have different roles1, or
‘horizontally integrated’, in which case firms and consumers are both engaged in
activities at the same level in the production process, generally in a collaborative manner.
Hence, from the firms’ perspective, it is first of all important to distinguish whether
co-creation corresponds to a clear and separate contribution of consumers or whether it is
a collaborative effort with consumers. A second critical aspect is whether the resulting
product is targeted at/has value for a single individual or is manufactured for the mass
market. Interestingly, while the former aspect relates to the ‘input’ side, or indeed to the
production process, the latter pertains to the ‘output’ and the market. Therefore, the
typology developed in this article covers both inside-out and outside-in aspects of open
innovation (Gassmann and Enkel, 2004).
Based on this, it is possible to build a typology of co-creation to identify challenges
common to the various existing (and future) forms of co-creation. In Figure 1, co-creation
forms are categorised according to the degree to which they imply a ‘differentiated’2 or
integrated3 consumer input (vertical axis) and according to the extent to which the output
corresponds to a tailored/customised product, valuable to one, or to a ‘mass’ product,
valuable to many (horizontal axis).

Figure 1 Typology of co-creation

44 T. Rayna and L. Striukova

The first quadrant (top/right) of Figure 1 corresponds to forms of co-creation where the
resulting product is mainly targeted at the mass market (or, at least, at a significant
segment of the market) and in which consumers and firms tend to assume different roles.
Some forms of ‘creative crowdsourcing’ fall within this quadrant. The company
threadless, for instance, ‘crowd-sources’ t-shirt designs, which are then mass-produced.4
In this case, consumers’ and firm’s activities are clearly differentiated, as consumers are
in charge of producing and selecting the designs while threadless takes care of the
manufacturing.5 Crowdfunding would also fall within this first category. Crowdfunding
implies different roles for consumers (funding) and firms (design, manufacturing, etc.)
and is targeted at a large number of consumers.
The second quadrant (down/right) corresponds to forms of co-creation for which the
output is valuable to many and which involve firms and consumers taking similar roles.
The most famous example of this form of co-creation is, without a doubt, open source.
Though not all the users of open source software contribute to its development, many do.
They do so in collaboration with the software developers of the firm that carries out the
open source project.
The third quadrant (down/left) relates to co-creation practices for which firms’ and
consumers’ roles in the co-creation process overlap (at least at some level) and which
relate to the production of a customised/bespoke product.
While this form of co-creation was seldom used during the early years of Web 2.0, it
is now on the rise. Indeed, 3D-printing technologies enable collaborative design and full
mass customisation to take place. Start-ups like digital forming and UCODO offer
co-design platforms that enable users to modify the shape, aspect and look of 3D models
of objects professionally designed. Users can then purchase the resulting object, which is
then ‘3D-printed’ and shipped. In such a case, though the manufacturing side of the
creation is taken care of by the firm, the designing process is conducted jointly by the
user and the firm. Nonetheless, considering the current rate of adoption of 3D printers6, it
is reasonable to expect that co-creation between companies and users will further extend
to include manufacturing activities (for instance, plastic parts could be manufactured at
home, while metal parts would be still produced by firms).
The last quadrant (up/left) is where co-creation activities that involve both
differentiated/segregated actions and customised/bespoke products are categorised.
Overall, this quadrant can be related to inside-out open innovation (Gassmann and Enkel,
2004), where external paths to market are sought. There are, so far, fewer examples of
such co-creation. User manufacturing, such as Ikea’s co-creation model, which includes
transportation and assembly (Kambil et al., 1996; Payne et al., 2009), fits this category.
Other examples may include ‘print-at-home’ tickets or stamps (because firms create and
consumers manufacture) and similar initiatives. Digital downloads can also belong to this
category (as users ‘manufacture’ a copy of the product on their hard drives).7 Users
printing at home spare/replacement parts using 3D printers also fit into this category. In
the future, the democratisation of 3D printing technologies is likely to enable more
advanced types of co-creation. In fact, this quadrant could be where two opposites –
crowdsourcing and mass-customisation – merge. Just as with crowdsourcing, customers
would be solely in charge of design and firms of manufacturing, just as with
mass-customisation, the resulting product would be meant for one customer only and not
for the mass market.
 Open innovation 2.0 45

The examples provided above illustrate the most extreme degrees in each category.
Yet, it is important to emphasise that this is not always the case with actual co-creation.
Even bespoke co-creation can be targeted at least at a few consumers and even ‘mass’
products are generally aimed at a particular market segment (not everyone likes to wear
the same t-shirt or jewellery). Likewise, even when activities of consumers and firms are
differentiated, they often overlap and it is exceedingly rare that firms and consumers have
purely integrated activities. Hence, a very large range of co-creation practices can be
analysed using this typology.
Just as any other form of open innovation, the actual benefits of co-creation for firms
strongly depend on the motivations of partners, on the costs associated with this activity
and on the intellectual property produced as an outcome of the collaboration. The
following sections will investigate how the challenges associated with each of these three
aspects differ in various types of co-creation.

5 Motivational challenges of co-creation

In order to make co-creation successful, it is important to understand what drives

customers’ desire to contribute. Web 2.0 made it easier to communicate with consumers
and also to enlarge the potential base of contributors, however, it also made motivating
customers more challenging. It is already difficult to make a ‘stranger’ contribute and
collaborate in an offline, face-to-face, environment but it becomes even more difficult to
do so within an online environment. Indeed, one of the critical issue in online and virtual
environment relate to the lack of social capital, in general, and trust, in particular
(Striukova and Rayna, 2008).
Some motivation factors are, in fact, rather general and can be used to encourage
co-creation in all four quadrants, both when the roles of consumers and firms in
co-creation can be easily distinguished and when they overlap and also both in cases
when co-creation results in producing individual products or targets mass market. These
motivation factors include enjoyment and fun (von Hippel and von Krogh, 2003) as well
as objectives and intellectual stimulations (Ridings and Gefen, 2004; Wasko and Faraj,
2000).
As mentioned above, trust is another critical factor. For instance, if customers do not
trust a particular brand, they are much less likely to ‘do anything to help it’ (Bughin et al.,
2008). Yet, while trust is an important motivational element in general (and, thus, relate
to all quadrants), it is particularly critical for some types of co-creation. For example,
with user manufacturing, customers need to trust that their e-tickets will be honoured
when they take the plane (Etgar, 2008). Similarly, customers engaging in integrated types
of co-creation need to be sure that the other party has the skills and resources required to
achieve the collaborative task at hand. In both cases, the issue is whether the other party
will ‘keep their side of the bargain’. Also, while in the case of ‘mass’ products trust
concerns mainly relate to whether one’s contribution will be acknowledged and
(possibly) rewarded, in the case of ‘custom’ products, trust issues are related to whether
the resulting ‘product’ will live up to expectations. In both cases, however, repeated
interactions are particularly instrumental in establishing trust (Striukova and Rayna,
2008).
46 T. Rayna and L. Striukova

Co-creation processes belonging to quadrants 1, 2, 3 may result in consumers

working together or amending products or designs created by other consumers.
Customers may decide to work together in order to have fun (Antikainen et al., 2010) or
because they are seeking friendships and want to ‘hang out together’ (Wasko and Faraj,
2000; Ridings and Gefen, 2004). It is important to remember that rewarding a group of
people working together is more challenging than rewarding individuals (Antikainen
et al., 2010) as the entertaining side of collective creativity does not fully fit with
monetary rewards. Monetary rewards would, therefore, increase participation, but not
collaboration. On the other hand, tools enabling to comment on others’ designs and to
make suggestions may be more efficient in terms of motivating collaboration.
While it is easy to understand why consumers might want to create products for their
own use (quadrants 3 and 4), not every consumer might be enthusiastic about co-creating
products for future mass production (quadrants 1 and 2). Those who do so are motivated
by the prospect of peer recognition (Lerner and Tirole, 2002), in a hope to enhance their
reputation and professional status (Lerner and Tirole, 2002; Wasko and Faraj, 2005), out
of altruism (Zeityln, 2003) and finally seek monetary rewards (Wasko and Faraj, 2000).
Motivation can be increased, when combining some of these factors together. For
example, a survey conducted by Antikainen et al. (2010) has shown that the combination
of monetary and non-monetary rewards (such as announcing the rewarded participants on
the website) works especially well.
Nevertheless, successful co-creation in quadrants 1 and 2 is not only about motivating
customers to co-create, but also identifying those consumers, whose needs will
correspond to those of the market, in other words, lead users (von Hippel, 1986). Li and
Bernoff (2011) show that Web 2.0 can be used to identify lead users. However, this can
be rather costly for a company, as it requires a team focusing on blogging, tweeting and
participating in forums.
Finally, motivation to work together with a company, instead of working (or creating)
an independent co-creation community should be considered. Currently, consumers are
rather motivated to become co-creators with companies. For example, 61% of US online
adults would consider helping companies design and build new products or improve
existing products, with more than half of them being interested in co-creation “regardless
of the product, service, or brand involved” (Williams et al., 2010). A major motivation
factor for them to participate in co-creation process is time involvement, which is a
scarce resource for consumers (Etgar, 2008). Also, 63% of consumers mentioned that
“participating is dependent upon receiving some form of compensation” (Williams et al.,
2010).
To sum up, a combination of monetary and non-monetary incentives is important for
successful co-creation. Finding the right mix of incentives is critical for firms as it is a
requirement of successful co-creation. Because motivations are different depending on
the type of co-creation, the incentive structure should be tailored to fit the particular
requirements of each type. It is also important to ensure that it is easy for consumers to
contribute, as lower costs of contributing will require weaker incentives.

6 Costs and risks of co-creation

Depending on the type of co-creation, the nature of the costs involved, as well as their
extent, vary greatly. Looking at Figure 1, it is clear, for instance, that the two quadrants
 Open innovation 2.0 47

on the left are more likely to be associated with higher manufacturing costs than the two
quadrants on the right, simply because custom/bespoke products do not allow for as much
economies of scale as goods which are mass produced. Of course, these greater costs can
be offset by the fact that consumers generally attribute a greater value to customised
products, as they normally fulfil their needs better. However, it is unlikely that consumers
will want all the products they consume to be custom made (there are also costs related to
customisation on the consumer side) and even when they choose customised products, it
is unclear whether the additional value they derive from customisation would lead to a
willingness to pay an amount high enough to cover the higher level of costs caused by
lower economies of scale.
In contrast, ‘mass’ products tend to carry a greater risk and uncertainty for firms than
custom ones. Indeed, it is far more difficult to predict the demand for a mass-market
product than for a customised one. Nonetheless, Web 2.0 enables to alleviate (at least
partially) this issue.
First of all, it enables to gather consumer feedback at an early stage. In the case of
crowdsourcing, the products that finally reach the mass-market are generally those
selected by Web 2.0 users. Though users taking part in ‘crowd-voting’ may not be
entirely representative of the whole target population of consumers, they are often ‘early
adopters’ or ‘visionaries’, in which case their choices are likely to impact those of the rest
of the consumer population (Rayna et al., 2009). Furthermore, crowdsourced products are
rarely targeted at the whole market but, instead, address market niches. Thus, co-created
mass products always embed some elements of customisation, which means less
uncertainty.
Crowdfunding, by nature, also alleviates part of the uncertainty, as the product is only
manufactured once a sufficient amount of funds has been put forward by the crowd.
Nevertheless, there are still risks involved because of the investment of resources at the
conception/design stage (most of which are sunk costs). Interestingly, crowdsourcing
(user-created design) and crowdfunding (user-funded manufacturing) could be run
together in order to minimise risk. In that case, however, there is a question of whether
users would engage in a co-creation process where they do, essentially, ‘everything’ and
of how much would a firm be able to charge for their really minimal input in the process.
Likewise, low profitability for a (possibly) minimal contribution is also often associated
with open source projects (Rayna and Striukova, 2010).
Aside from differences in cost due to scale, the various forms of co-creation differ
with respect to the transaction costs they involve. Transaction costs consist of search and
information costs (Coase, 1937; Williamson, 1981), bargaining and negotiation costs,
monitoring and enforcement costs (Tirole, 1988). According to Alston and Gillespie
(1989), research costs are likely to arise during the entire production process: in
pre-production (information constraint, asset specificity), production (agency and
coordination costs, contract enforcement) and post-production (measurement of output
and contract enforcement).
With regards to search and information costs of co-creation, they are likely to be
minimal. Indeed, in most cases, firms do not search for consumers to co-create with, but
are instead sought by consumers. Likewise, they do not normally need, except in
particular cases, to obtain information about consumers who engage in co-creation with
them. This is not surprising as search costs are, traditionally, borne by consumers rather
than firms.
48 T. Rayna and L. Striukova

Nonetheless, in some cases, firms are likely to face significant search and information
costs. This is generally the case when consumers compete to contribute to the co-creation
of one product (and thus, usually, relates to quadrants 1 and 2). For instance, threadless
does not sell all designs submitted by users, but only the ‘best’ ones. In fact, the vast
majority of crowdsourcing practices have the same shortcoming: firms need to spend a
large amount of resources ‘searching’ for the right designs/solutions among all those
submitted by the crowd. There might be, furthermore, duplication of efforts, as users
might submit the same or very similar contributions. In some cases, firms are able to
reduce these costs by ‘crowdsourcing’ the search (this is what threadless does by asking
users to vote for designs). However, this is only feasible when the ‘crowd’ and the firm
have the same (or close enough) opinion of what a satisfactory contribution is. A
difference of opinion between the firm and the contributors is also often mentioned as
one of the challenges of open source (Rayna and Striukova, 2010). It is important to note,
though, that not all forms of co-creations in quadrants 1 and 2 involve large search costs
(crowdfunding, for instance, does not).
Though co-creation activities in the two left quadrants are generally not subject
to search costs, they are, because of their ‘custom’ aspect, more subject to bargaining
and negotiation costs (which arise prior to the contract being signed): in the most
extreme version, one contract (possibly a different one) has to be signed with each
customer.
Integrated forms of co-creation (where firms and customers play similar roles) are,
because of their lack of explicit specialisation, prone to coordination costs. In such cases,
there is a risk of either duplication of efforts or gaps in the production process. This is
one of the risks that are often associated with open source projects (everyone in the
community focuses on the same parts of the project, while other parts are not covered),
but this can arise as well for integrated/custom co-creation forms. A solution to this issue
is modularity (Lerner and Tirole, 2002). In some cases, it is possible to divide the
co-creation process in well-defined tasks assigned to each participant. In the case of open
source projects, this means dividing a piece of software into smaller size modules (e.g.,
interface, file management) that can be assigned to a particular developer. In the case of
co-creation platforms, the firms can, for instance, decide to provide the original design
and not to contribute beyond that.
Enforcement costs, whether during production or post-production are likely to arise
for co-creation activities in all quadrants except quadrant 1. Though theoretically there
could be enforcement costs for this type of co-creation, it is usually not the case because
consumers supply a ‘finished product’ that is then selected or rejected by the company. In
contrast, all the other types of co-creation activities are potentially subject to monitoring
and enforcement costs during production and enforcement costs during post-production.
Information asymmetries, for instance, could be a source of such costs, as they always
create a gap between consumers and the firms (Franke and Piller, 2004). For instance,
one of the issues may be that both parties have a different view of what the outcome
should be, but are unable to communicate clearly about it. Another problem relates to the
expertise of consumers, as firms might over or underestimate consumers’ skills. As a
result, it may be the case that the co-created product is actually impossible to
manufacture (Magnusson et al., 2003). Finally, with user manufacturing, consumers may
not be able to manufacture the product (e.g., print a ticket) in a satisfactory manner,
which might lead to additional costs.
The costs discussed in this section are summarised in Figure 2.
 Open innovation 2.0 49

Figure 2 Typology of co-creation and associated costs

7 IPRs for co-creation

Managing IPRs is perhaps one of the major challenges for companies embracing open
innovation, in general, and co-creation in particular. According to West and Gallagher
(2006), it is rather challenging to motivate individuals to generate and contribute IP if
there are no financial returns. The main challenge, however, when it comes to
co-creation, is to determine who is the owner of the resulting IP and then to choose
between a weak and a strong IP enforcement strategy.
Assigning IPRs should be a rather straightforward task for co-creation in the
quadrants 1 and 4, since there is a clear separation between the roles of firms and
consumers. Assigning IPRs in quadrant 1 is also simplified by the fact that, in most
crowdsourcing projects, the company that initiates the project keeps all the IPRs. This
might change in the future, though. Some companies, for example, orange with its service
call and reward project, started to offer a joint IP ownership to the contributors. In
quadrant 4, IPRs become an issue when digital goods are involved and customers who
download these goods can potentially redistribute them without authorisation. Firms
attempted to use technical solutions, such as digital rights management systems, to
prevent redistribution, but such solutions have been, so far rather ineffective (Rayna and
Striukova, 2008), so companies have instead to change their business model, in order to
tackle this problem.
50 T. Rayna and L. Striukova

IPR management becomes even more problematic when the value of the respective
contributions is hard to identify and to measure (quadrants 2 and 3), as it is then even less
clear who should be the owner of the resulting IP. There are two possible solutions,
depending on whether the company is planning to choose a weak or a strong IPR
enforcement strategy. According to Farrell and Shapiro (2004), there are two opposing
IPR schools of thought: incentives and openness. Those who belong to the incentives
school of thought argue that creators are only likely to continue contributing in the future
if they can capture a sufficient value and, thus a strong IPR protection is necessary. When
such a strategy is chosen, the project can be divided into several parts to enable
contributors to participate in one portion of the project only (Bughin et al., 2008; Henkel
and Baldwin, 2010). In that case, the different parts of the project will be protected by
different IPRs. It is important to note, however, that such a modularity of IP is likely to
increase the cost of design, will add legal costs, and may imply a loss of performance.
In contrast, proponents of the openness school argue that weaker IPR regimes
stimulate creativity and innovation. In particular, artistic creators are intrinsically
motivated to create and continue producing even in the absence of strong IPR protection.
In that case, open source strategy or open design strategy can be chosen for co-creation
activities related to quadrants 3 and 4. Goals and philosophy of open design are similar to
open source. There is usually a centralised design repository (such as Thingiverse for 3D
printable designs) and participants use the designs of others and improve them or amend
to fit their needs.
A final challenge of co-creation and IP is that it may be difficult to maintain secrecy
when working with co-creators, which is often crucial during the prelaunch stage. This
risk is especially high for quadrants 1 and 2, since the output of co-creation is aimed at
mass market, and many are likely to be interested in information about new products
(unlike in the ‘custom’ case, where only one person is really interested).

8 Conclusions

Web 2.0 has created both challenges and opportunities for companies who implement
open innovation (co-create) with customers. In order to help see these challenges and
opportunities in a more structured way, this article has provided a two-dimensional
typology of co-creation, which classifies co-creation practices according to whether they
target individuals or mass-market (related to output) and whether the co-creation
activities of firms and consumers are differentiated or integrated (related to input). By
clearly identifying the particular type of a co-creation activity, it becomes possible to
devise strategies enabling to motivate contributors, minimise the costs and appropriate
value.
To this respect, this article has demonstrated that co-creation practices have a
different level of complexity, mainly related to how well the contribution of consumers
can be identified and measured. There are many simple co-creation processes, where the
roles are well defined, that companies can use. However, if firms opt for more advanced
co-creation structures, where the roles and contributions are blurrier, they should aim to
create long-term relationships with consumers. By creating innovative communities
around them, firms can act as a hub and devolve even more roles (e.g., examining ideas,
improving coordination and enforcement) to the community of consumers.
 Open innovation 2.0 51

In any case, technologies such as 3D printing will enable in the near future even more
co-creation opportunities. In this context, the future competitive advantage will probably
be less about ‘picking up good ideas here and there’, but instead having a stable
community of innovators around. For that to happen, costs and motivations should of
course, be thoroughly understood, but beyond that, the social relations and dynamics
within these communities need to be apprehended.
Besides the question of the role of social relations in co-creation, another avenue for
further research would be to empirically test the typology developed in this article.
While some types of co-creation are, at the moment, little developed, there is little
doubt that upcoming technologies, such as 3D printing, will enable more co-creation
(crowd-customisation, in particular) and, as such, will provide more opportunities to
validate empirically the co-creation typology presented in this article.

References
Alston, L. and Gillespie, W. (1989) ‘Resource coordination and transaction costs: a framework for
analyzing the firm/market boundary’, Journal of Economic Behavior & Organization, Vol. 11,
No. 2, pp.191–212.
Antikainen, M., Makipaa, M. and Ahonen, M. (2010) ‘Motivating and supporting collaboration in
open innovation’, European Journal of Innovation Management, Vol. 13, No. 1, pp.100–119.
Baldwin, C.H.C. and von Hippel, E. (2006) ‘How user innovations become commercial products:
A theoretical investigation and case study’, Research Policy, Vol. 35, No. 9, pp.1291–1313.
Bughin, J., Chui, M. and Johnson, B. (2008) ‘The next step in open innovation’, The McKinsey
Quarterly, No. 4, pp.112–122.
Chesbrough, H. (2003) Open Innovation: The New Imperative for Creating and Profiting from
Technology, Harvard Business School Press, Boston, MA.
Chesbrough, H., Vanhaverbeke, W. and West, J. (2006) Open Innovation: Researching a New
Paradigm, Oxford University Press, Oxford.
Christensen, C. (1997) The Innovators Dilemma: When New Technologies Cause Great Firms to
Fail, Harvard Business School Press, Boston, Massachusetts.
Coase, R.H. (1937) ‘The nature of the firm’, Economica, Vol. 4, No. 16, pp.386–405.
Enkel, E. (2010) ‘Attributes required for profiting from open innovation in networks’, International
Journal of Technology Management, Vol. 52, Nos. 3/4, pp.344–371.
Etgar, M. (2008) ‘A descriptive model of the consumer co-production process’, Journal of the
Academy of Marketing Science, Vol. 36, No. 1, pp.97–108.
Farrell, J. and Shapiro, C. (2004) ‘Intellectual property, competition, and information technology’,
in Varian, H., Farrell, J. and Shapiro, C. (Eds.): The Economics of Information Technology,
pp.49–87, Cambridge University Press.
Fournier, S. and Lee, L. (2009) ‘Getting brand communities right’, Harvard Business Review,
Vol. 87, No. 4, pp.105–111.
Franke, N. and Piller, F. (2004) ‘Toolkits for user innovation and design: an exploration of user
interaction and value creation’, Journal of Product Innovation Management, Vol. 21, No. 6,
pp.401–415.
Gassmann, O. and Enkel, E. (2004) ‘Towards a theory of open innovation: three core process
archetypes’, in Proceedings of the R&D Management Conference, pp.1–18, Lisbon, Portugal.
Geiger, D., Seedorf, S., Schulze, T., Nickerson, R. and Schader, M. (2011) ‘Managing the crowd:
Towards a taxonomy of crowdsourcing processes’, in Proceeding of Americas Conference on
Information Systems (AMCIS), pp.1–11, Detroit, Michigan.
52 T. Rayna and L. Striukova

Henkel, J. and Baldwin, C. (2010) Modularity for Value Appropriation: How to Draw the
Boundaries of Intellectual Property, Harvard Business School Finance Working Paper,
(11-054).
Howe, J. (2006) ‘The rise of crowdsourcing’, Wired Magazine, Vol. 14, No. 6, pp.1–4.
Jaspers, F. and den Ende, J.V. (2010) ‘Open innovation and systems integration: how and why
firms know more than they make’, International Journal of Technology Management, Vol. 52,
Nos. 3/4, pp.275–294.
Kambil, A., Ginsberg, A. and Bloch, M. (1996) Re-inventing Value Propositions, Stern Working
Paper IS-96-21, New York University.
Lerner, J. and Tirole, J. (2002) ‘Some simple economics of open source’, Journal of Industrial
Economics, Vol. 50, No. 2, pp.197–234.
Li, C. and Bernoff, J. (2011) Groundswell, Expanded and Revised Edition: Winning in a World
Transformed by Social Technologies, Harvard Business Review Press, Boston, MA.
Magnusson, P., Matthing, J. and Kristensson, P. (2003) ‘Managing user involvement in service
innovation experiments with innovating end users’, Journal of Service Research, Vol. 6,
No. 2, pp.111–124.
Nambisan, S. (2002) ‘Designing virtual customer environments for new product development:
toward a theory’, Academy of Management Review, Vol. 27, No. 3, pp.392–314.
Payne, A., Storbacka, K., Frow, P. and Knox, S. (2009) ‘Co-creating brands: diagnosing and
designing the relationship experience’, Journal of Business Research, Vol. 62, No. 3,
pp.379–389.
Piller, F., Ihl, C. and Vossen, A. (2010) ‘A typology of customer co-creation in the innovation
process’, available at SSRN 1732127.
Piller, F., Ihl, C., Fuller, J. and Stotko, C. (2004) ‘Toolkits for open innovation: the case of mobile
phone games’, in Proceedings of the 37th Annual Hawaii International Conference on System
Sciences, 2004, IEEE, pp.1–10.
Prahalad, C. and Ramaswamy, V. (2004) ‘Co-creation experiences: the next practice in value
creation’, Journal of Interactive Marketing, Vol. 18, No. 3, pp.5–14.
Rayna, T. and Striukova, L. (2008) ‘White knight or trojan horse? The consequences of
digital rights management for consumers, firms and society’, Communications & Strategies,
Quarter 1, No. 69, pp.109–125.
Rayna, T. and Striukova, L. (2010) ‘Large-scale open innovation: open source vs. patent pools’,
International Journal of Technology Management, Vol. 52, Nos. 3/4, pp.477–496.
Rayna, T., Striukova, L. and Landau, S. (2009) ‘Crossing the chasm or being crossed out: the case
of digital audio players’, International Journal of Actor-Network Theory and Technological
Innovation, Vol. 1, No. 3, pp.36–54.
Ridings, C.M. and Gefen, D. (2004) ‘Virtual community attraction: why people hang out online’,
Journal of Computer Mediated Communication, Vol. 10, No. 1, pp.0–1.
Ritzer, G. and Jurgenson, N. (2010) ‘Production, consumption, prosumption: the nature of
capitalism in the age of the digital ‘prosumer’’, Journal of Consumer Culture, Vol. 10, No. 1,
pp.13–36.
Sawhney, M. and Prandelli, E. (2000) ‘Communities of creation: managing distributed innovation
in turbulent markets’, California Management Review, Vol. 42, No. 4, pp.24–54.
Shah, S.K. and Tripsas, M. (2007) ‘The accidental entrepreneur: the emergent and collective
process of user entrepreneurship’, Strategic Entrepreneurship Journal, Vol. 1, No. 1,
pp.123–140.
Striukova, L. and Rayna, T. (2008) ‘The role of social capital in virtual teams and organisations:
corporate value creation’, International Journal of Networking and Virtual Organisations,
Vol. 5, No. 1, pp.103–119.
Tapscott, D. and Williams, A.D. (2006) Wikinomics: How Mass Collaboration Changes
Everything, Portfolio, New York.
 Open innovation 2.0 53

Tirole, J. (1988) The Theory of Industrial Organization, 14th ed., MIT Press, Cambridge, MA.
Trott, P. (2003) ‘Innovation and market research’, in Shavinina, L.V. (Ed.): International
Handbook on Innovation, pp.835–844, Pergamon/Elsevier, Oxford.
von Hippel, E. (1986) ‘Lead users: a source of novel product concepts’, Management Science,
Vol. 32, No. 7, pp.791–805.
von Hippel, E. (1988) The Source of Innovation, Oxford University Press, New York.
von Hippel, E. (2005) Democratizing Innovation, The MIT Press.
von Hippel, E. and von Krogh, G. (2003) ‘Open source software and the ‘private-collective’
innovation model: issues for organization science’, Organization Science, Vol. 14, No. 2,
pp.208–223.
Wasko, M. and Faraj, S. (2000) ‘It is what one does: why people participate and help others in
electronic communities of practice’, Journal of Strategic Information Systems, Vol. 9,
Nos. 2–3, pp.155–173.
Wasko, M. and Faraj, S. (2005) ‘Why should I share? Examining social capital and knowledge
contribution in electronic networks of practice’, MIS Quarterly, Vol. 29, No. 1, pp.35–57.
West, J. and Gallagher, S. (2006) ‘Challenges of open innovation: the paradox of firm investment
in open-source software’, R&D Management, Vol. 36, No. 3, pp.319–331.
Whitla, P. (2009) ‘Crowdsourcing and its application in marketing activities’, Contemporary
Management Research, Vol. 5, No. 1, pp.15–28.
Williams, D., Gownder, J.P., Wiramihardja, L. and Corbett, A.E. (2010) US Consumers Are
Willing Co-Creators – Activate Engaged Consumers with Social Technologies to Build Better
Products, Forrester Research, Cambridge, UK.
Williamson, O.E. (1981) ‘The economics of organization: the transaction cost approach’, The
American Journal of Sociology, Vol. 87, No. 3, pp.548–577.
Wohlers, T. (2012) Additive manufacturing and 3D printing: State of the industry, Wohlers Report
2012, Wohlers Associates.
Zeityln, D. (2003) ‘Gift economies in the development of open source software: anthropological
reflections’, Research Policy, Vol. 32, No. 7, pp.1287–1291.
Zwass, V. (2010) ‘Co-creation: toward a taxonomy and an integrated research perspective’,
International Journal of Electronic Commerce, Vol. 15, No. 1, pp.11–48.

Notes
1 E.g., in the extreme cases: consumers provide the design and firms manufacture and vice
versa.
2 In which case the activities of firms and consumers can be seen as vertically integrated.
3 Which refers to a collective work between firms and consumers. Firms and consumers’
activities are vertically integrated.
4 The t-shirt business model has been copied by dozens of other companies, e.g.,
design-by-humans. It has also been adapted to other industries, for instance to customised
jewellery (CrowdJewel, Mejuri).
5 And of course, distribution and sales. Promotion may be considered as a collaborative effort.
6 Wohlers (2012) reports a 289% unit growth in sales of 3D printers in 2011.
7 In that case the custom aspect relate to the ‘on demand’ manufacturing as well as the fact that
consumers choose a personalised bundle of digital goods.
 Journal of Manufacturing Technology Management
Additive manufacturing in the mechanical engineering and medical industries
spare parts supply chain
Melanie Muir, Abubaker Haddud,
Article information:
To cite this document:
Melanie Muir, Abubaker Haddud, (2017) "Additive manufacturing in the mechanical engineering and
medical industries spare parts supply chain", Journal of Manufacturing Technology Management,
https://doi.org/10.1108/JMTM-01-2017-0004
Permanent link to this document:
https://doi.org/10.1108/JMTM-01-2017-0004
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

Downloaded on: 18 January 2018, At: 19:20 (PT)

References: this document contains references to 94 other documents.
To copy this document: [email protected]
The fulltext of this document has been downloaded 75 times since 2017*
Users who downloaded this article also downloaded:
(2017),"The impact of additive manufacturing on supply chains", International Journal of
Physical Distribution &amp; Logistics Management, Vol. 47 Iss 10 pp. 954-971 <a href="https://
doi.org/10.1108/IJPDLM-11-2016-0332">https://doi.org/10.1108/IJPDLM-11-2016-0332</a>
,"Additive manufacturing in the wood-furniture sector: Sustainability of the technology, benefits and
limitations of adoption", Journal of Manufacturing Technology Management, Vol. 0 Iss 0 pp. - <a
href="https://doi.org/10.1108/JMTM-08-2017-0175">https://doi.org/10.1108/JMTM-08-2017-0175</a>

Access to this document was granted through an Emerald subscription provided by emerald-
srm:305060 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.

*Related content and download information correct at time of download.

 The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1741-038X.htm

Additive
Additive manufacturing in the manufacturing
mechanical engineering and
medical industries spare parts
supply chain
Melanie Muir and Abubaker Haddud Received 11 January 2017
Revised 15 June 2017
Management School, University of Liverpool, Liverpool, UK 11 November 2017
Accepted 28 November 2017

Abstract
Purpose – The purpose of this paper is to approximate the impact that additive manufacturing (AM) will
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

have on firm inventory performance (IP) and customer satisfaction (CS) when it is applied within the spare
parts (SP) supply chain of manufacturing organisations. This research also explores the influence of customer
sensitivity (CSy) to price and delivery lead time and supply risk (SR) within those approximations.
Design/methodology/approach – An online survey was used to collect the primary data for this research.
Data were collected from 69 respondents working for organisations in two industrial segments within the UK
manufacturing sector: “Industrial and Commercial Machinery and Computer Equipment” and “Measuring,
Analysing and Controlling Instruments, Photographic, Medical and Optical Instruments”. The respondents
worked for entities that were categorised in three groups: customers, suppliers, and entities that were both
customers and suppliers. The groups that were self-identified as “customers” or “suppliers” answered 20 survey
items each and the group that was identified as both “customers” and “suppliers” answered 40 survey items.
Findings – The results revealed that AM was considered a suitable vehicle for the fulfilment of SP demand.
However, AM appeared to make no material difference to CS; the scenario used improved delivery time of SP
but increased price. Also, AM was thought to improve IP through less reliance on buffer stock to manage
SR and spikes in demand and less carrying of SP at risk of obsolescence.
Research limitations/implications – The respondents worked for entities within two manufacturing
industry segments within the UK and the insights garnered may not be indicative of similar organisations
competing in other manufacturing industry segments within the UK or in other countries. In addition,
approximately 82 per cent of the surveyed respondents worked for small organisations with fewer than
100 employees and the results may differ for larger organisations. Further limitations were the relatively
small sample size and lack of open-ended questions used in the survey. Larger sample size and the usage of
open-ended survey questions may lead to more reliable and valuable responses and feedback.
Practical implications – The findings from this research are considered to be of interest to practitioners
contemplating adoption of AM and to developers of AM wishing to increase market share due to the positive
reaction of entities within the industrial and commercial machinery and computer equipment, and measuring,
analysing and controlling instrumentation industrial segments. This research raises awareness to the
possible risks and rewards – from a range of perspectives, of AM to practitioners considering its adoption in
the spare parts supply chain (SPSC).
Originality/value – The paper takes a novel perspective on AM in SPSCs by illuminating the supplier and
buyer perspective based on empirical data. This research provides new insights about the appreciation of the
use of AM in SPSCs of mostly small sized manufacturing companies located in the UK. This paper also gives
new insights about the willingness/conditions of manufacturing companies in the UK to adopt AM for the
provision of SP. The originality of this research is twofold: it broached the applicability of AM in the supply
chains of the two targeted industrial segments, and as far as the authors are aware, the influence of CSy
(e.g. to price or lead time) and SR on SPSC players’ attitude to AM is yet to be considered. Finally,
this research adopted a systems theory lens and considered system-wide impact of AM introduction.
Keywords Customer satisfaction, Inventory management, Additive manufacturing, Business performance,
Supply chains
Paper type Research paper

1. Introduction
1.1 Background Journal of Manufacturing
Technology Management
Satisfying the demand for spare parts (SP) is a challenge for both the supplier and the © Emerald Publishing Limited
1741-038X
customer. On the one hand, the supplier aims to reduce the requirement for holding of DOI 10.1108/JMTM-01-2017-0004
 JMTM slow-moving/excess inventory without causing adverse impact on customer satisfaction (CS).
On the other, the customer aims to maximise equipment availability, thus, requires access to
SP quickly without suffering the adverse impact of high procurement costs. Several studies
indicate that Additive Manufacturing (AM) technology usage may have an impact on different
actors in the supply chain such as: suppliers, manufacturing firms, and customers
(Oettmeier and Hofmann, 2016). For example, such impact may be narrower and shorter
supply chains (Holmström et al., 2010). AM technologies provide the opportunity to
integrate additional functionality into products and to optimise products for function
(Holmström et al., 2010; Glasschroeder et al., 2015). AM has the potential to further these
benefits due to AM’s ability to produce a wide variety of replacement and SP using a single
source of raw material (Knofius et al., 2016; Meisel et al., 2016). However, AM in the spare parts
supply chain (SPSC) and the collective opinions of supply chain entities regarding its use from
both an inventory management and CS perspective appear to be less researched and merits
investigation. The level in which AM will penetrate manufacturing industries and society
overall is not (yet) known and there are many research opportunities in this field (Kothman and
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

Faber, 2016; Meisel et al., 2016; Steenhuis and Pretorius, 2017).

Holmström et al. (2010), Khajavi et al. (2014), and Liu et al. (2014) examine the appropriate
position of AM in the SPSC. Holmström et al. (2010) approaches this conceptually, whereas
Khajavi et al. (2014) and Liu et al. (2014) utilise scenarios. Each of the three articles is set
within the aerospace industry, the respective authors utilising existing literature to conduct
their studies and form their conclusions. The benefits and shortcomings of AM in each
configuration are acknowledged but not challenged. In other words, would these benefits
and shortcomings be the same regardless of specific situation? It has been suggested,
for example, that current AM functionality for the production of SP would be beneficial in
remote locations where equipment outlives availability of parts or in situations where it
would be dangerous for parts to be delivered in the usual way (Peres and Noyes, 2006;
Tatham et al., 2015). To that end, perhaps the existence of supply risk (SR) increases the
attractiveness of AM. In addition, by concluding in favour of AM in a central position within
the supply chain, Holmström et al. (2010) and Khajavi et al. (2014) appear to consider the
short-term needs of the firm (e.g. return on investment) more important than utilising AM as
a dynamic capability to positively impact CS. However, according to Wits et al. (2016), some
end users may be willing to absorb the cost of AM themselves (possibly rendering the firm
superfluous) in order to minimise machine downtime as well as enjoy other benefits AM
offers. To that end, perhaps the existence of customer sensitivity (CSy) (e.g. to price or
delivery lead time) should be taken into consideration.
There are some deficiencies of existing studies within the aerospace SP environment,
which make application of the findings to other industries inappropriate. For example,
the Liu et al. (2014) study only focusses on total safety inventory in the aircraft SPSC.
Khajavi et al. (2014) recommend exploring the utilisation of AM technology in non-military
operations and industries with lower downtime costs such as in the auto industry.
In addition, Holmström et al. (2010) suggest further research be undertaken to conceptualise
the development of rapid manufacturing in the SPSC.
It is the authors’ contention that what is missing from the aforementioned body of work is a
comprehensive investigation of the potential impact of AM in the SPSC regardless of industry;
particularly the dyadic goals of firm inventory performance (IP) and CS. Moreover, as far as the
authors are aware, the influence of CSy and SR on SPSC players’ attitude to AM is yet to be
considered. The objective of the study is to explore the potential impact of AM in the SPSC,
expressly regarding firm IP and CS. Moreover, the influence of CSy and SR on SPSC players’
attitude to AM is also examined. The study will explore the following research questions:
RQ1. How would AM be perceived as a viable alternative to currently adopted methods to
fulfil SP demand and how does supplier risk influence a firm’s view on its viability?
 RQ2. What impact is expected as a result of AM adoption on CS and IP? Additive
RQ3. How would CSy to price, or delivery lead time, influence the impact of AM? manufacturing
The study solicits opinions of respondents engaging in a quantitative survey regarding
anticipated outcomes of AM were it to be adopted as an alternative means of SP creation in
supply chains of organisations within the manufacturing sector. In particular, entities
within the manufacturing sector sub-categories “Industrial and Commercial Machinery and
Computer E[quipment]” and “Measuring, Analysing and Controlling Instruments” are
targeted (Data HQ Limited, 2016b). Opinions sought are restricted to the outcomes of
relevance to the study including IP and CS, as well as, the influence of SR and CSy (to price
and lead time) on the formation of those opinions. Other outcomes, such as quality and
safety, are considered beyond the scope of the study. The authors opine a technical review
of AM performance would be required to provide these insights such as the research
undertaken by Zha and Anand (2015). The study focusses primarily on literature published
within the last five years because there is a dearth in material published pre-2012 containing
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

either “additive manufacturing” or “3D printing” coupled with “spare parts” search terms.
This corresponds with expansion of AM use from the production of prototypes to become a
vehicle for SP manufacture (Zistl, 2014).

1.2 Definition of constructs

The following definitions are provided in order to clarify the meaning of the four main
constructs explored in this research:
(1) CS
The Financial Times (2016a) citing Oliver (1997) defines CS “as a judgment
following a consumption experience it is the consumer’s judgment that a product
provided (or is providing) a pleasurable level of consumption-related fulfillment [sic]”.
In this study, “customers […] predicted product performance” pertains to delivery
(lead time) and cost of SP produced utilising AM, which when compared against both
actual performance and that experienced when traditional techniques were applied,
gives rise to a level of CS or dissatisfaction (The Financial Times, 2016a).
(2) CSy
According to The Financial Times (2016b) citing Longman Business English
Dictionary (n.d.), “Price Sensitivity” can be defined as “[t]he degree to which a change
in the price of something leads to a change in the amount sold or that could possibly
be sold”. However, this study expands upon this definition to include delivery
sensitivity and it relates both to CS instead of sales. Thus, adapting the above, the
definition of CSy becomes, “[t]he degree to which a change in the price [or lead time] of
something leads to a change in [level of CS]”.
(3) IP
As Eroglu and Hofer (2011, p. 359) hypothesised, this study also assumes
“[i]nventory leanness will have a positive effect on firm performance”. The quest
for lean inventory originated from the Toyota Production System and its
elimination of the seven wastes including the “[w]aste of stock on hand”
(Ohno, 1988, p. 20). That said, it is important for inventory (e.g. SP) to be available
in the right quantity when needed and this is the underpinning of “[j]ust-in-time”
philosophy (1988, p. 4). Hence, the definition of (ideal) IP according to this study is
the immediate availability of SP to satisfy specific demand, without reliance on
buffer inventory. Also, performance is measured via inventory costs, availability
and the fill rate (Hedenstierna and Disney, 2016).
 JMTM (4) SR
Zsidisin (2003, p. 222) defines SR as, “the probability of an incident associated
with inbound supply from individual supplier failures or the supply market
occurring, in which its outcomes result in the inability of the purchasing firm to meet
customer demand”. The definition goes on to include a further negative outcome
being, “threats to customer life and safety” (p. 222), however this falls beyond the
scope of the current study.
This paper consists of five main sections – introduction and definition of constructs, literature
review that includes: AM, AM in SPSC, and the research conceptual framework, research
methodology and methods, results and discussions, and conclusions and recommendations.

2. Literature review and conceptual framework

The literature review is divided into three parts. In Section 2.1, literature about AM
including definitions, overview and history, and key benefits it offers. In Section 2.2,
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

AM adoption within the spare part supply chain will be explored and key benefits and
challenges will be highlighted. In Section 2.3, the research conceptual framework will be
introduced and the four main research hypotheses will be presented. Finally, Section 2.4 will
summarise the existent research gap.

2.1 AM
AM is the umbrella term used to reflect a range of related technologies including
“stereolithography […] selective laser sintering […] fused deposition modeling [sic] […]
solid ground curing […] laminated object manufacturing [and] […] digital light processing”
(Ruan et al., 2014 cited in Mawale et al., 2016, p. 94). AM involves “fabricating an artefact […]
from a [three-dimensional computer aided design] model without the need for process
planning in advance of manufacturing” (Chen et al., 2015, p. 616). In simplistic terms, an item
is produced using powder-form raw materials (e.g. ground metals) and resin for binding in a
layer-by-layer repetitive process (unlike traditional manufacturing methods involving
numerous sequential steps), which in essence amount to deriving an item from a solid mass.
As reported by Kruth et al. (1998), AM began to take hold in industry within the 1980s. From
its introduction, AM was predominantly used to reduce the lead time required for the
creation of pre-production prototypes or “concept models” (Kruth et al., 1998, p. 527), thus
accelerating the design stage of bringing new products to the market. Whilst it is still
deployed in this way, AM also features in production processes of early adopting aerospace,
motor sports, dentistry and audiology industries (see Cooper et al., 2012; Liu et al., 2014;
Barone et al., 2016; Sandström, 2016).
AM is used in mass customisation, a case in point being the production of hearing aids.
Sandström (2016, p. 163) found the key reasons behind AM’s popularity in this sector
included the audiology industry’s desire to standardise the process of hearing aid
production – previously considered an “artisanship” enterprise. Moreover, AM afforded a
vehicle to reduce inherent manufacturing variability and, therefore, improve quality of
output. However, AM in a mass customisation setting is considered atypical – and when
applied, is not necessarily coveted by non-AM users competing in same market
(Sandström, 2016). Instead, the manufacturing method is apparently more suited to
production of single-item, small and medium sized batches albeit of customised artefacts
(Mellor et al., 2014). AM offers a number of benefits such as raw material waste reduction
(Campbell et al., 2011; cited in Thomas, 2016) – since the process makes use of all material
inputs, is executed right first time (depending on accuracy of drawings) and is additive
rather than deductive. Moreover, AM is a rapid-response technique hence a dynamic
capability (Kim and Park, 2013; cited in Thomas, 2016) to the extent that customers and
 suppliers can join forces to design and manufacture precisely what the customer requires Additive
with minimal delay. A further benefit of AM recognised in the aerospace industry is its manufacturing
potential to reduce weight of parts through creation of hollowed out versions without
impacting critical attributes such as strength and stability (Beyer, 2014). Unfortunately,
AM is not free of limitations. For example, initial investment is steep for industrial models.
However, Gardan (2016, p. 3121) suggests where AM is concerned, “low-cost machines
have a low performance”. Moreover, it is said that a reduced need for holding and
transporting inventory outweighs this shortcoming (Beyer, 2014).

2.2 AM in SPSC
Provision of SP is a fundamental element of after-sales service expected of manufacturing
entities in the age of “[s]ervitization” where the relationship with the customer no longer
ends with the sale of a product (Baines and Lightfoot, 2013; cited in Huxtable and
Schaefer, 2016, p. 49). Moreover, as Costantino et al. (2012, p. 451) observed, customers’
appetite for firm agility is ever more demanding requiring rapid response even to “demand[s]
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

for low-volume and high variety products”. That said, as well as serving the customer,
organisations are equally keen to minimise capital tied up in slow-moving inventory, such as
the aforementioned SP with intermittent demand. Thus, various methods are deployed
to balance capability for timely fulfilment of demand for SP whilst limiting incurrence of
inventory carrying costs involving intricate forecasting of future demand based on
intrinsically inadequate historical information. Essentially, accuracy of forecasts dictates
firms’ ability to satisfy their customers and manage working capital.
But achieving forecasting accuracy is the “Holy Grail” of inventory management. It is
especially difficult when demand quantity and frequency are unstable with patterns
including a “high fraction of zero-demand periods” (Hahn and Leucht, 2015, p. 543).
Unsurprisingly, success is often elusive; reliance on buffer inventory is therefore
commonplace and is even utilised albeit to a lesser extent when greater forecasting
accuracy is achieved (Barrow and Kourentzes, 2016). Avoidance of significant late
delivery penalties and loss of credibility are great motivators for maintenance of safety
stocks. Moreover, in some cases, buffer inventories are considered unavoidable when
supply side risks are present. For example, when SP is at the point of discontinuation but
the useful life of the product they serve is yet to expire. In these instances, panic bulk
purchases are likely – regardless of forecasted demand, in order to avoid reliance on the
unofficial “gray market [sic]” (Wolbert, 2014). However, it is possible that AM may provide
a reliable alternative.
AM is not without shortcomings. High initial investment and on-going cost of
“feedstock”, in addition to the need to retrain existing, or recruit new specialist labour
requires significant financial outlay (Royal Academy of Engineering, 2013, p. 17).
Furthermore, the finish achieved by AM is sometimes considered inadequate. Despite
these barriers, the functionality is gaining momentum in the aerospace industry supply
chain for the manufacture of SP. According to Holmström et al. (2010), Khajavi et al. (2014)
and Liu et al. (2014), in a centralised setting where significant costs can be spread over
greatest volume of units, the benefits of AM appear to outweigh the disadvantages.
However, what is not clear from these studies is the impact of AM on both firm IP and CS.
Moreover, whether AM is an attractive alternative to traditional methods applied in other
manufacturing settings.

2.3 Theoretical and conceptual framework

The purpose of this section is to introduce, with the aid of existing literature, the underlying
theories Dynamic Capabilities and Systems Theory, which have been selected to guide the
research. Combined, these theories form the “framework provid[ing] […] both structure and
 JMTM boundaries within which” the proceeding research will be undertaken (Ennis, 1999, p. 133).
In addition, the conceptual framework is both illustrated and discussed, and relationships
between key concepts explained.
2.3.1 Dynamic capabilities. Acquisition and continuous adaptation of companies’
capabilities is required to facilitate rapid response to changes in customer demands,
and is considered imperative in today’s highly competitive and unpredictable market.
Teece et al. (1997, p. 516) believe “dynamic capabilities” provide the answer and define them
as a “firm’s ability to integrate, build, and reconfigure internal and external competences to
address rapidly changing environments”. Dynamic capabilities offer an extension to the
largely static resource-based view (RBV ). Resources according to RBV being the current
“strength[s] or weakness[es] of a given firm” (Wernerfelt, 1984, p. 172). Instead, dynamic
capabilities involve the reorganising, adaptation and refining of existing resources to
counter new challenges as they emerge (Allred et al., 2011; cited in Blome et al., 2013).
The SPSC arguably presents such a challenge, its environment beset with unpredictable
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

lumpy demand, in terms of quantity, frequency, and type. Attempts have been made to
develop dynamic capabilities for this purpose, for example, through improved forecasting
(Saalmann et al., 2016). However, a firm’s ability to capitalise on dynamic capabilities is also
dependent on effective communication with supply chain partners and making sense of and
utilising pertinent insights gleaned (Saenz et al., 2014).
2.3.2 Systems theory. As Arnold and Wade (2015) discovered, many attempts have been
made to define systems theory – but alas a unilaterally agreed definition remains at large.
However, Meadows (2008, p. 11) defines systems as “an interconnected set of elements that
[are] coherently organised in a way that achieves something”. Therefore, it is perhaps
important when contemplating a change to consider its impact on the whole system rather
than basing the decision to go ahead on the success of a singular element. Moreover, it has
been suggested for this reflection to take place outside the system’s perimeter not in the
midst of it, for example “from the vantage point of the [overarching] metasystem”
(Van Gigch, 1979, p. 661) to ensure an unbiased opinion of the system’s holistic performance.
Failure to objectively consider the whole system could result in improvement in one aspect
being offset by deterioration elsewhere. Seemingly heeding this advice, Townsend and
Urbanic (2012) adopting a “teleological systems” theory lens, prove through use of a case
study that best outcomes can be achieved through the contemporaneous application of
“additive manufacturing” and “computer numerically controlled […] machining”, instead
of adopting one method but not the other (2012, pp. 324-326).
2.3.3 Conceptual framework. This research is based on the conceptual framework
illustrated in Figure 1. The framework considers the independent variable AM adoption as a
potential dynamic capability within SPSC; in other words, a tool to be flexed and moulded
allowing a firm to keep pace within an unpredictable business environment in the supply chain
arena. This is in line with the suggestion of Arafa and ElMaraghy (2011) that possession of
dynamic capabilities in the manufacturing industry is represented by a firm’s ability to wield

Customer Customer
H2
sensitivity (CSy) satisfaction (CS)

H1
Additive
Spare parts Improved spare parts
manufacturing H0
supply chains supply shain performance
adoption
Figure 1. H3
Conceptual framework Inventory
of the research Supply risk (SR) H4
performance (IP)
 flexibility at will. However, because AM adoption is considered a “process innovation”, care Additive
must be taken to assess the impact of its adoption on the entire supply chain performance manufacturing
(Caputo et al., 2016), e.g. in this research the dependent variables such as firms IP (internal) and
CS (external) are used. Moreover, consideration must also be given to the possible influence the
intermediate variables of CSy (to price and lead time) and SR may have on the tolerance of AM
by CS and IP dependent variables respectively. Thus, the null hypothesis for this research is:
Ho. AM adoption in SPSC does not influence CS and IP.
From the customer perspective, AM is expected to improve CS where CSy to delivery lead
time is prominent (i.e. short lead time for procurement of SP is desirable) and CSy to price is
negligible (Khajavi et al., 2014). That said CS is expected to deteriorate if CSy to price
is prominent and CSy to delivery lead time is negligible. Based on these expectations;
the following two hypotheses have been developed:
H1. AM adoption in SPSCs is expected to influence CSy to price and lead time.
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

H2. AM adoption in SPSCs will lead to improved CS.

From the supplier perspective, the authors predict a lack of SR will limit the attractiveness
of AM as a vehicle to improve IP. In such a scenario, AM is likely to be considered
an expensive solution for common after-sales parts despite its posited usefulness in
“just-in-time manufacturing” (Frazier, 2014, p. 1925). However, growth in attractiveness of
AM as a solution is anticipated where SR is present ( Janssen et al., 2014). Based on these
expectations, the following two hypotheses have been developed:
H3. AM adoption in SPSCs is expected to minimise SR.
H4. AM adoption in SPSCs will lead to improved IP.

2.4 Gaps in extant literature

It is the authors’ contention that what is missing from the aforementioned body of work is a
comprehensive investigation of the potential impact of AM in the SPSC particularly the dyadic
goals of firm IP and CS. Moreover, as far as the authors are aware, the influence of CSy (e.g. to
price or lead time) and SR on SPSC players’ attitude to AM is yet to be considered.

3. Research methodology
3.1 Research approach
This research utilised an online survey to solicit quantitative data. This methodology falls
outside the extremities of positivism since the reality is estimated through the capture of
collective opinions rather than through first-hand experiments (Easterby-Smith et al., 2012).
A 5-point Likert scale was adopted and the questionnaire distributed to a number of entities
included in lists maintained by Data HQ Limited (2016b) that were willing to be contacted
via e-mail. Since Likert scale had been utilised, the resultant data was considered ordinal
rather than continuous, which limited analytical approaches generally accepted by the
research fraternity.

3.2 Questionnaire design

The questionnaire comprised the following sections: introduction; a statement assuring
prospective survey participants of their anonymity; a statement explaining completion of
the survey implied issuance of informed consent; an approximation of duration; and
instructions on how to complete the survey. The survey participants were invited to select,
from a drop-down list, the industry segment most applicable to their activities – including
 JMTM the option “prefer not to say”. A multiple-choice question was then posed to establish
whether SP were procured for own use (i.e. respondent is a customer) or manufactured for
sale (i.e. respondent is a supplier) or both. “[S]kip-page logic” was then applied to direct
survey participants to a specific set of questions (e.g. customer, supplier, or both customer
and supplier) based on the answer to aforementioned question (Purch, 2016b). Each of the
sections customer or supplier contained four questions with five items each resulting in a
total of 20 examined items. Also, the customer and supplier section contained eight
questions with five items each resulting in a total of 40 examined items. The authors
recognised the possibility of respondents abandoning the survey due to the number of
questions posed (Chudoba, 2010), but opined fewer questions would not answer the research
objectives. On completion of the survey, all participants were invited to provide their
opinion regarding face validity of the survey. The number of questions for the entire survey
varied, being equivalent to the section completed plus initial two questions and final rating
of the survey. The questions adopted a Likert scale format, where numbers were attributed
to a brief description of opinion (Thwaites Bee and Murdoch-Eaton, 2016).
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

Collecting data from single respondents may raise the issue of common method bias.
To overcome this, in developing the survey, improving item scale as a recommended
procedural remedy was followed (Podsakoff et al., 2003). The reduction of method bias was
achieved through the careful development of the used items themselves. For example,
the developed items were kept: simple, specific, and concise. Also, use of ambiguous or
unfamiliar terms and vague concepts was avoided (Tourangeau et al., 2000). Furthermore,
it was clearly stated that any provided personal data would remain anonymous and this
should reduce respondents’ apprehension. In addition, to further test for common method
bias, eight items measuring variable unrelated to the research topic were included at the end
of the survey (Bagozzi, 2011; Lindell and Whitney, 2001). These items were about rating
each of the used items in the survey and their results were analysed separately from the rest
of examined items. In summary, all these used tests and measures provide evidence that the
collected data did not suffer from common method bias issues.

3.3 Research population, sample, and survey administration

The target research population were enterprises included in the “Industrial and Commercial
Machinery and Computer E[quipment]” and “Measuring, Analysing and Controlling
Instruments” mailing lists; the sub-groups being members of Data HQ Limited’s (2016b)
manufacturing category. A substantial number of entities in the lists were described as
small and medium size enterprises (SMEs), although some branches of larger firms were
also included (Data HQ Limited, 2016b). Access to funds can be a barrier for SMEs but
various financing options are said to be available and therefore AM may help SMEs achieve
their quest for production efficiencies (General Electric Company, 2011). Only enterprises
that had agreed to be contacted by e-mail were approached but the final sample was also
dependent on the quantity of survey responses received. Every effort was made to collect
data from a larger group from the two targeted industry segments. Due to a limited access to
more e-mail addresses and time constraints, a total number of 69 usable survey responses
were collected and used for this research. However, a total of 1,735 sent invitation emails
were opened and only 69 completed the survey.
Although one of the common drawbacks of using electronic surveys is a poor response
rates compared with other types of survey methodologies (Robson, 2011), we attempted to
figure out the reason(s) behind this low response rate. The first possible sample bias is
“response bias”. Response bias refers to the bias that results from problems in the
measurement process. As it is explained in Section 3.2 earlier, the collected data did not
suffer from common method bias issues as this was mitigated through: the careful
development of the used items; reducing respondents’ apprehension by indicating in the
 survey that any provided personal data would remain anonymous; and measuring variable Additive
unrelated to the research topic, included at the end of the survey (i.e. the rating questions). manufacturing
Thus, “response bias” did not occur. The second type of sample bias is the “non-response
bias”. Non-response bias is defined as “the likelihood that individuals who did not return a
completed questionnaire would have responded to the questionnaire items differently from
those who did respond” (Richardson-Harman et al., 1988, p. 184). Thus, the authors think
some individuals chosen for the sample were either unwilling or unable to participate in the
survey, due to either a limited knowledge about AM in general or a lack of knowledge about
the potential impacts its adoption may bring to SPSCs.
An e-mail was sent to the prospective participants containing a link to an online survey.
E-mail communication medium was selected due to the speed of distribution and the rapid
notification of failed delivery messages. Although this method, labelled “convenience
sampling”, is sometimes criticised (Easterby-Smith et al., 2012, p. 228), the authors opined
e-mail and the internet is a familiar combination offering greater convenience to survey
recipients than, say, a physical letter-based survey. Although it is recommended for a trial to
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

be undertaken to test “bounce rates” of e-mail distribution (Data HQ Limited, 2016a, p. 2),
this was not practicable due to time constraints. However, following initial distribution, in
an effort to increase the survey’s final response rate, a reminder e-mail was circulated.
The online survey tool selected was www.SurveyMonkey.com achieving second
place following an assessment against similar tools (Purch, 2016a). Of particular interest
was the “SurveyMonkey Gold […] skip-page logic”, which facilitated customisation of
questionnaire to different groups (Purch, 2016b).

3.4 Data analysis methods

The IBM Statistical Package for the Social Sciences version 24 was used to analyse the
collected data in this study. First, the reliability and validity of the data collection tool were
examined. In order to examine the consistency and reliability of the used items in the survey,
Cronbach α values and corrected item-total correlation coefficients were used. Also,
an exploratory factor analysis was used for all of the used items in the survey that utilised
the Likert five-point agreement scale to examine their construct validity. Second, descriptive
analyses were conducted to calculate frequency means for each item within the four
constructs that used the five-point Likert level of agreement scale. Furthermore, the χ2
residual values were calculated for each of the used items in the survey and illustrated in the
corresponding tables in Section 4. This test was used to examine the developed research
hypotheses in this study.

3.5 Research instrument validity and reliability testing

Three statistical tests have been used to examine and evaluate the used research
instrument’s validity and reliability. Face validity and an exploratory factor analysis were
used to validate the used items within each of the three used constructs. Also, Cronbach’s α
reliability coefficient was used to examine the internal consistency of the used items.
Testing the internal consistency was necessary to examine the ability of each used Likert
item to measure the underlying theme (Davenport Jr et al., 2015). Table I provides details

Number of Items factor Items Cronbach’s α if item Construct

Scale items loadings ranges deleted ranges Cronbach’s α value

Customers 20 0.704 to 0.979 0.616 0.732 0.695 Table I.

Suppliers 20 0.734 to 0.960 0.859 0.885 0.875 Validity and reliability
Customers and suppliers 40 0.757 to 0.785 0.859 0.885 0.770 tests results ranges
 JMTM about: name of used constructs, number of used items within each construct, factor loading
value ranges for the used items within each construct, Cronbach’s α values if items deleted,
and Cronbach’s α values for the entire three scales.
3.5.1 Instrument validity test ( factor analysis). Given the fact that this survey instrument
has been developed for this research and it has not been used before, an exploratory factor
analysis was used to validate the validity of the used scales. A rule-of-thumb is that factor
loadings greater than 0.30 are considered significant, loadings greater than 0.40 are considered
more important, and loadings that are 0.50 or greater are considered very significant
(Hair et al., 2005). As it can be seen in Table I, the factor loading values for the examined
20 items in the “customers” construct ranged from 0.704 to 0.979. Also, the factor loading
values for the examined 20 items in the “suppliers” construct ranged from 0.734 to 0.960.
Furthermore, the factor loading values for the examined 40 items in the “customers and
suppliers” construct ranged from 0.757 to 0.785. All the examined items within the three used
constructs were greater than 0.7; thus, these values can be considered very significant and that
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

the used constructs are valid in measuring the impact of AM on the SPSC.
3.5.2 Instrument reliability test (cronbach’s α). On the other hand, Cronbach’s α reliability
coefficient was used to examine the internal consistency of the used items. George and
Mallery (2003) provide the following role of thumb: (0.5 and below) unacceptable, (0.6 and
above) questionable, (0.7 and above) acceptable, (0.8 and above) good, (0.9 and above)
excellent. As shown in Table I, Cronbach’s α values for the “customers” scale was 0.695;
the “suppliers’ scale” was 0.875; and the “customers and suppliers” scale was 0.770. It is
suggested that Cronbach’s α coefficient values of 0.7 or more provide reliability evidence for
internal consistency of the measurement scales (Rivard and Huff, 1988). Other sources
suggest that minimum α of 0.600 typically considered acceptable (Manerikar and
Manerikar, 2015). Although the commonly acceptable reliable coefficient is normally 0.7 or
higher, lower thresholds are sometimes used in the literature (Santos, 1999). Thus,
the internal consistency of the three used measurement scales is reliable.
3.5.3 Items “face validity” evaluation. Face validity is defined as “a complex,
multidimensional construct useful for evaluating how test items appear to respondents and
others” (Cronbach, 1970, p. 182). At the end of the used survey, the respondents were asked to
rate a number of statements in an endeavour to evaluate the “face validity” of the used items.
As illustrated in Table II, the respondents who were self-identified as “customers” rated four
items and the mean values were between 3.53 and 3.87. In addition, the “suppliers”
respondents rated four different statements and the mean values ranged from 3.09 to 3.36.
Finally, the respondents who were self-identified as “customers and suppliers” rated eight
statements and the mean values ranged from 3.56 to 3.81. It can be noticed that all the ratings’
mean averages were within band 3 (Agree) of the used five levels of agreement. Thus,
the used test items appeared to the respondents as effective in measuring what they intended
to measure.

4. Analysis and results

As shown in Table III, data were collected from 69 respondents for this study. The respondents
worked for firms operating in two industrial segments; industrial and commercial machinery
and computer equipment (66.67 per cent); and measuring, analysing and controlling
instruments, photographic, medical and optical instruments (33.33 per cent). The majority of
respondents (82.61 per cent) worked for small organisations of 99 employees or below.
The respondents were self-identified as, and categorised into, three groups: customers
(21.74 per cent), suppliers (34.78 per cent), and customers and suppliers (43.48 per cent).
The following sub-sections provide results about the three used constructs. Mean values
for each item are presented in each table. Cronbach’s α (if item deleted) values are also
 Mean values
Additive
Customers and manufacturing
Survey question Customers Suppliers suppliers

Please rate the following statements (customers)

The survey effectively measured current level of customer
satisfaction 3.53 3.74
The survey effectively measured predicted level of customer
satisfaction on introduction of additive manufacturing 3.60 3.70
The survey effectively measured customer sensitivity to price 3.87 3.70
The survey effectively measured customer sensitivity to delivery
lead time 3.87 3.81
Please rate the following statements (suppliers)
The survey effectively measured current spare parts inventory
performance 3.36 3.70
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

The survey effectively measured predicted spare parts inventory Table II.
performance on introduction of additive manufacturing 3.23 3.67 Results of the used
The survey effectively measured additive manufacturing as spare survey rating
parts fulfilment method 3.09 3.56 questions for the
The survey effectively measured supply risk 3.14 3.70 “face validity”

Frequency
Survey question (n ¼ 69) Percentage

Industry segment
Industrial and commercial machinery and computer equipment 46 66.67
Measuring, analysing and controlling instruments, photographic, medical and
optical instruments 23 33.33
Organisation’s number of employees
Small (1-99 employees) 57 82.61
Medium (100-999 employees) 5 7.25
Large (1,000 and more employees) 7 10.14
Organisation’s most significant relationship to spare parts
Customer (i.e. spare parts are mainly procured for own use) 15 21.74
Supplier (i.e. spare parts are mainly manufactured or procured for sale) 24 34.78
Customer and Supplier (i.e. spare parts are both procured for own use, and Table III.
manufactured or procured for sale) 30 43.48 Profile of respondents

provided for each item. Furthermore, the factor analysis loading values for each item are
included in the tables. Further still, in order to test the main research hypotheses in this
study, a χ2 test is used. χ2 test is used to examine data for differences, associations,
and relationships to answer hypotheses (Waller, 2012). The χ2 test compares observed
frequencies (N) with the corresponding expected frequencies (N) under each of the used
five levels of agreement (Minaei-Bidgoli et al., 2004). The expected N can be determined by
dividing the number of total usable responses by the number of used Likert scale levels of
agreement. By dividing 15 by 5, the result is 3 and this is the expected N that is used to
calculate the residual values for the “customers” scale. For example, the first examined item
in Table IV has actual responses (observed N) as follows: Strongly Disagree (0); Disagree (1);
Neutral (2); Agree (9); and Strongly Agree (3). By subtracting the expected N values from the
observed N values, the results were residual values of −3, −2, −1, 6, and 0. It can be inferred
that the most significant category was the fourth one (Agree). The highest residual value
 Downloaded by University of Florida At 19:20 18 January 2018 (PT)

JMTM

Table IV.

“customers”)
(respondents as
and sensitivity to
price and lead time
Survey results about
customer satisfaction
χ2 test – ( frequencies
residual values)
Cronbach’s Factor Item
Survey question Mean α loading SD D N A SA supported

4. Based on your organisation’s experience of ordering spare parts, please rate the following statements
Spare parts are available when my organisation requires them 3.93 0.708 0.979 −3 −2 −1 6 0 Yes
Spare parts orders are delivered on time 3.93 0.701 0.876 −3 −2 −1 6 0 Yes
Spare parts are customised to my organisation’s needs 2.93 0.616 0.957 −1 0 1 −2 1 No
Spare parts are priced at a level my organisation is happy to pay 3.27 0.710 0.765 −3 1 1 3 −2 Yes
I would recommend my organisation’s spare parts supplier(s) to others 3.87 0.714 0.704 −3 −3 0 8 −2 Yes
5. Assuming that additive manufacturing, adopted by your supplier(s), both reduces lead time and increases price of spare parts by 10%
Spare parts are available when my organisation requires them 4.0 0.639 0.974 −2 −2 −3 5 2 Yes
Spare parts orders are delivered on time 4.07 0.658 0.950 −2 −2 −3 4 3 Yes
Spare parts are customised to my organisation’s needs 3.67 0.631 0.907 −1 −2 −1 2 2 Yes
Spare parts are priced at a level my organisation is happy to pay 2.87 0.633 0.872 −1 1 1 1 −2 No
I would recommend my organisation’s spare parts supplier(s) to others 3.47 0.633 0.947 −1 −2 −1 5 −1 Yes
6. Please rate the following statements concerning pricing of spare parts
My organisation prefers to wait longer and pay less for spare parts 2.73 0.672 0.934 −2 4 0 0 −2 No
My organisation prefers to buy spare parts in bulk to achieve lower price bracket 2.73 0.707 0.845 −1 3 −2 3 0 No
If price of spare parts increases, my organisation will look for alternative supplier(s) 4.00 0.695 0.795 −3 −2 −1 5 1 Yes
Lower price is more important than lower delivery lead time 2.47 0.693 0.872 −1 2 4 −2 0 No
My organisation prefers fixed pricing of spare parts 3.40 0.650 0.889 −2 −2 3 2 −1 No
7. Please rate the following statements concerning delivery lead time of spare parts
My organisation prefers to pay more to receive spare parts sooner 3.13 0.688 0.868 −1 0 2 2 −2 Yes
My organisation prefers to buy spare parts immediately when required to achieve lower
equipment downtime 4.33 0.702 0.857 −3 −2 −2 2 5 Yes
If delivery lead time of spare parts increases, my organisation will look for alternative supplier(s) 3.93 0.702 0.931 −3 −1 −2 5 1 Yes
Shorter lead time is more important than lower price 3.47 0.732 0.897 −3 −2 5 1 −1 No
My organisation prefers fixed delivery lead time of spare parts 3.60 0.714 0.835 −3 −1 1 4 −1 Yes
Notes: 1 ¼ Strongly disagree; 2 ¼ Disagree; 3 ¼ Neutral; 4 ¼ Agree; 5 ¼ Strongly agree; SD ¼ Strongly disagree; D ¼ Disagree; N ¼ Neutral; A ¼ Agree; SA ¼ Strongly agree
 will be the decisive factor when selecting under which of the five categories the majority of Additive
responses were. Any examined item in the scale that has the highest residual value under manufacturing
“Strongly Disagree”, “Disagree”, or “Neutral” will be considered as insignificant and the
examined item is considered as “Not Supported”.

4.1 Customers’ construct: impact of AM adoption on CS and sensitivity to price and lead time
Four main questions with five items each have been used to examine the potential impact
of AM adoption on CS and sensitivity to prices and lead times. Respondents who identified
themselves as “customers” have completed this part of the used survey and the results
were as presented in Table IV. Out of the examined 20 items, 11 had mean values between
3 and 4 and this falls within the “Agree” band on the used five-point Likert scale. Also,
five items had mean values between 2 and 3 and this represents “Disagree” and “Neutral”
on the used answers’ options. These include: “Spare parts are customised to my
organisation’s needs” (2.93); “Spare parts are priced at a level my organisation is happy to
pay” (2.87); “My organisation prefers to wait longer and pay less for spare parts” (2.73);
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

“My organisation prefers to buy spare parts in bulk to achieve lower price bracket” (2.72);
and “Lower price is more important than lower delivery lead time” (2.47). Furthermore,
4 items had mean values of 4.0 or higher and these match “Agree” and “Strongly agree” on
the used five-point Likert scale. These include: “Spare parts are available when my
organisation requires them” (4.00); “Spare parts orders are delivered on time” (4.07);
“If price of spare parts increases, my organisation will look for alternative supplier(s)”
(4.00); and “My organisation prefers to buy spare parts immediately when required to
achieve lower equipment downtime” (4.33).
The χ2 test results indicate that out of the examined 20 items that were answered by the
“Customers” respondents, seven items were not supported. The themes of these
unsupported items relate to seeking lower prices but maintaining fixed or shorter lead
times. This indicates a high level of sensitivity to higher prices and long lead times.
The results also indicate that the respondents felt the adoption of AM in the supply chain is
expected to lead to shorter lead times but may result in higher SP’ prices. Thus, the null
hypothesis of this research Ho is rejected and the first alternative hypothesis H1 accepted.
It can be clearly inferred that AM adoption would positively influence the lead times by
making them shorter and this result is in line with the findings of other researches (e.g. Pour
and Zanoni, 2017; Attaran, 2017). Although the “direct” price of the SP is expected to be
sometimes higher, organisations save in other indirect costs, e.g. inventory holding and
management costs (Tziantopoulos et al., 2016) and downtime costs (Eggenberger et al., 2017).
Overall, the results indicate that AM adoption would positively impact: the availability of SP,
customised SP, on-time delivery, shorter lead times, and less stock through the adoption of
zero-stock approaches and ordering parts when needed. Thus, customers’ satisfaction with
better lead time is likely to improve but it may not do so (at this stage of AM development)
when it comes to SP prices. Thus, it is fair to say that based on the highlighted benefits,
the second alternative hypothesis H2 is accepted but the hypothesis is not strongly supported
when the price factor is considered.

4.2 Suppliers’ Construct: impact of AM adoption on IP and SR

Four main questions with five items each have been used to examine the potential impact of
AM adoption on IP and SR. Respondents who identified themselves as “Suppliers” have
completed this part of the used survey and the results were as illustrated in Table V. Out of
the examined 20 items, the majority (17 items) had mean values between 3 and 4 and this
falls within the “Agree” band on the used five-point Likert scale. On the other hand,
two items had mean values within the “Disagree” band and these were: “Spare parts
procured for sale are likely not subject to supply risk” (2.83) and “Spare parts required by
 Downloaded by University of Florida At 19:20 18 January 2018 (PT)

JMTM

Table V.

“suppliers”)
(respondents as
and supply risk
Survey results about
inventory performance
χ2 test – ( frequencies
residual values)
Cronbach’s Factor Item
Survey question Mean α loading SD D N A SA supported

8. Based on your organisation’s current spare parts inventory performance, please rate the following statements
Spare parts buffer stock is maintained to offset possible supply risk 3.54 0.868 0.847 −4.8 −0.8 −0.8 5.2 0.2 Yes
Spare parts buffer stock is maintained to offset possible spikes in customer demand 3.75 0.867 0.772 −4.8 −2.8 −2.8 11.2 −0.8 Yes
Some spare parts stock held is at risk of obsolescence 3.33 0.862 0.851 −2.8 −1.8 −1.8 2.2 2.2 Yes
Spare parts demand is difficult to forecast with accuracy 4.08 0.865 0.752 −4.8 −4.8 −1.8 6.2 4.2 Yes
Spare parts with low demand have long lead times 3.29 0.867 0.840 −4.8 2.2 0.2 0.2 1.2 No
9. Assuming your organisation adopts additive manufacturing to manage spare parts inventory performance, please rate the following statements
Spare parts buffer stock is maintained to offset possible supply risk 3.21 0.871 0.824 −4.8 0.2 1.2 6.2 −3.8 Yes
Spare parts buffer stock is maintained to offset possible spikes in customer demand 3.54 0.865 0.603 −4.8 −1.8 −1.8 10.2 −2.8 Yes
Some spare parts stock held is at risk of obsolescence 3.08 0.866 0.752 −0.8 −0.8 −1.8 2.2 0.2 Yes
Spare parts demand is difficult to forecast with accuracy 3.92 0.866 0.778 −4.8 −2.8 −1.8 4.2 4.2 Yes
Spare parts with low demand have long lead times 3.08 0.866 0.774 −4.8 4.2 −0.8 1.2 −0.8 No
10. Assuming your organisation adopts additive manufacturing as its spare parts fulfilment method, please rate the following statements
Spare parts are likely available when my organisation’s customers require them 3.08 0.861 0.903 −3.8 −1.8 −1.8 7.2 −2.8 Yes
Customer orders for spare parts are likely delivered on time 3.08 0.862 0.960 −2.8 −3.8 −1.8 9.2 −3.8 Yes
Spare parts are likely customised to my organisation’s customers’ needs 3.08 0.859 0.811 −1.8 −2.8 −2.8 7.2 −1.8 Yes
Spare parts procured for sale are likely not subject to supply risk 2.83 0.864 0.928 −3.8 0.2 −0.8 5.2 −3.8 Yes
My organisation’s spare parts fulfilment method is likely a competitive strength 3.17 0.859 0.784 −3.8 −3.8 0.2 7.2 −2.8 Yes
11. Please rate the following statements concerning supply risk
My organisation has a reliable spare parts supply chain 3.88 0.884 0.850 −4.8 −3.8 −0.2 9.2 −1.8 Yes
My organisation adopts a multi-sourcing strategy for spare parts procurement 3.54 0.882 0.749 −3.8 −1.8 2.2 5.2 −0.8 Yes
Spare parts required by my organisation’s customers are readily available in the market 2.96 0.880 0.842 −3.8 4.2 −0.8 5.2 −4.8 Yes
My organisation has buying strength in the spare parts market 3.21 0.885 0.734 −3.8 2.2 1.2 1.2 −0.8 No
My organisation’s suppliers are subject to regular reliability review 3.25 0.884 0.769 −4.8 0.2 4.2 4.2 −3.8 Yes
Notes: 1 ¼ Strongly disagree; 2 ¼ Disagree; 3 ¼ Neutral; 4 ¼ Agree; 5 ¼ Strongly agree; SD ¼ Strongly disagree; D ¼ Disagree; N ¼ Neutral; A ¼ Agree; SA ¼ Strongly agree
 my organisation’s customers are readily available in the market” (2.96). In addition, only one Additive
item had a mean value greater than 4 and this falls within the “Strongly Agree” level of manufacturing
agreement and the item was: “Spare parts demand is difficult to forecast with accuracy” (4.08).
Using the χ2 test, the results indicate that out of the examined 20 items that the
“Suppliers” respondents answered, only three items were not supported. Results from
the first and fourth question indicate that organisations are very keen on ensuring the
availability and supply of the needed SP. The results also indicate that the SR levels are
high and it is difficult to forecast SP demand with accuracy and this leads to keeping of
more buffer stocks. However, the results from the second and third question indicate that
the adoption of AM is likely to influence the SPSC performance. Thus, the null hypothesis
of this research Ho is rejected. In addition, the respondents agreed with most of the
examined items related to SR and indicated that AM adoption is likely to minimise some of
these risks. Thus, H3 is accepted. Key benefits organisations gain from adoption of AM
included increased SP availability when needed and this is in line with other researchers’
findings (e.g. Holmström and Partanen, 2014; Pour and Zanoni, 2017). Delivering orders on
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

time was also seen as a potential benefit from adopting AM in the SPSC. This potential
benefit was also highlighted in other articles (e.g. Dwivedi et al., 2017; Sasson and
Johnson, 2016). Furthermore, the results from this study revealed that obtaining
customised SP to the needs of an organisation is a strong potential from adoption of AM.
Such finding is in line with results from others recent studies in the same field (e.g. Cohen
et al., 2014; Durach et al., 2017; Li et al., 2017; Oettmeier and Hofmann, 2016). This would
also lead to lower buffer stocks and therefore to improved IP. Thus, H4 is accepted.
However, it is important to mention that the respondents have cast some doubt about the
suggestion that all the SR would diminish because of AM adoption.

4.3 Customers’ and suppliers’ construct: impact of AM adoption on CS and sensitivity to

prices and lead times
Four main questions with five items each have been used to examine the potential impact of
AM adoption on CS and sensitivity to prices and lead times. Respondents who were
self-identified as “Customers and Suppliers” have completed this part of the used survey
and the results were as presented in Table VI. Out of the examined 20 items, 15 items had
mean values between 3 and 4 and this falls within the “Agree” band on the used five-point
Likert scale. On the other hand, two items had mean values within the “Disagree” band and
these were: “My organisation prefers to wait longer and pay less for spare parts” (2.39)
and “Lower price is more important than lower delivery lead time” (2.64). In addition, three
items had mean values greater than 4 and these values fall within the “Strongly agree” level of
agreement and these items were: “If price of spare parts increases, my organisation will look
for alternative supplier(s)” (4.04); “My organisation prefers to buy spare parts immediately
when required to achieve lower equipment downtime” (4.11); and “If delivery lead time of
spare parts increases, my organisation will look for alternative supplier(s)” (4.25).
The χ2 test results indicate that out of the examined 20 items that were answered by the
respondents who were self-identified as “Customers and Suppliers”, only four items were not
supported. The same observed results trends under the “Customers” group are also
observed here. Thus, the null hypothesis Ho is rejected and the first alternative hypothesis
H1 is accepted. Overall, the results indicate that AM adoption would positively impact:
the availability of SP, customised SP, on-time delivery, shorter lead times, and less stock
through the adoption of zero-stock approaches and ordering parts when needed. Thus,
customers’ satisfaction with better lead time is likely to improve but it may not do so (at this
stage of AM development) when it comes to SP prices. Thus, it is fair to say that based on
the highlighted benefits, the second alternative hypothesis H2 is accepted but the
hypothesis is not strongly supported when the price factor is considered.
 Downloaded by University of Florida At 19:20 18 January 2018 (PT)

JMTM

suppliers”)
Table VI.

“customers and
(respondents as
and sensitivity to
price and lead time
Survey results about
customer satisfaction
χ2 test – ( frequencies
residual values)
Cronbach’s Factor Item
Survey question Mean α loading SD D N A SA supported

4. Based on your organisation’s experience of ordering spare parts, please rate the following statements
Spare parts are available when my organisation requires them 3.75 0.773 0.920 −4.6 −2.6 −3.6 12.4 −1.6 Yes
Spare parts orders are delivered on time 3.54 0.772 0.871 −4.6 −1.6 0.4 7.4 −1.6 Yes
Spare parts are customised to my organisation’s needs 3.04 0.757 0.890 −4.6 2.4 3.4 3.4 −4.6 Yes
Spare parts are priced at a level my organisation is happy to pay 3.14 0.758 0.870 −3.6 −1.6 5.4 4.4 −4.6 No
I would recommend my organisation’s spare parts supplier(s) to others 3.68 0.759 0.917 −4.6 −4.6 1.4 10.4 −2.6 Yes
5. Assuming that additive manufacturing, adopted by your supplier(s), both reduces lead time and increases price of spare parts by 10%
Spare parts are available when my organisation requires them 3.96 0.756 0.838 −5.6 −4.6 −1.6 12.4 −0.6 Yes
Spare parts orders are delivered on time 3.82 0.750 0.894 −5.6 −2.6 −0.6 8.4 0.4 Yes
Spare parts are customised to my organisation’s needs 3.61 0.756 0.887 −5.6 −1.6 2.4 7.4 −1.6 Yes
Spare parts are priced at a level my organisation is happy to pay 3.11 0.768 0.841 −5.6 1.4 7.4 0.4 −3.6 No
I would recommend my organisation’s spare parts supplier(s) to others 3.61 0.763 0.905 −5.6 −4.6 3.4 7.4 −1.6 Yes
6. Please rate the following statements concerning pricing of spare parts
My organisation prefers to wait longer and pay less for spare parts 2.39 0.759 0.800 −1.6 9.4 −0.6 −3.6 −3.6 No
My organisation prefers to buy spare parts in bulk to achieve lower price bracket 3.25 0.753 0.869 −4.6 2.4 −0.6 5.4 −2.6 Yes
If price of spare parts increases, my organisation will look for alternative supplier(s) 4.04 0.764 0.935 −5.6 −5.6 0.4 9.4 1.4 Yes
Lower price is more important than lower delivery lead time 2.64 0.766 0.896 −3.6 5.4 5.4 −2.6 −4.6 No
My organisation prefers fixed pricing of spare parts 3.64 0.759 0.716 −5.6 −3.6 3.4 8.4 −2.6 Yes
7. Please rate the following statements concerning delivery lead time of spare parts
My organisation prefers to pay more to receive spare parts sooner 3.32 0.771 0.930 −5.6 1.4 1.4 6.4 −3.6 Yes
My organisation prefers to buy spare parts immediately when required to achieve lower
equipment downtime 4.11 0.773 0.888 −5.6 −4.6 −2.6 10.4 2.4 Yes
If delivery lead time of spare parts increases, my organisation will look for alternative
supplier(s) 4.25 0.775 0.888 −5.6 −5.6 −3.6 11.4 3.4 Yes
Shorter lead time is more important than lower price 3.36 0.777 0.895 −5.6 −1.6 3.4 5.4 −2.6 Yes
My organisation prefers fixed delivery lead time of spare parts 3.89 0.757 0.915 −5.6 −3.6 −0.6 9.4 0.4 Yes
Notes: 1 ¼ Strongly disagree; 2 ¼ Disagree; 3 ¼ Neutral; 4 ¼ Agree; 5 ¼ Strongly agree; SD ¼ Strongly disagree; D ¼ Disagree; N ¼ Neutral; A ¼ Agree; SA ¼ Strongly agree
 4.4 Customers’ and suppliers’ construct: impact of AM adoption on IP and SR Additive
Four main questions with five items each have been used to examine the potential impact of manufacturing
AM adoption on IP and SR. Respondents who identified themselves as “Customers and
Suppliers” have completed this part of the used survey and the results were as illustrated
in Table VII. Out of the examined 20 items, the majority (17 items) had mean values
between 3 and 4 and this falls within the “Agree” band on the used five-point Likert scale.
On the other hand, only one item had a mean value between 2 and 3 and this falls within the
“Neutral” level of agreement and this item was: “Some spare parts stock held is at risk of
obsolescence” (2.93). In addition, two items had mean values greater than 4 and these falls
within the “Strongly Agree” level of agreement and the items were: “Spare parts demand is
difficult to forecast with accuracy” (4.04) and “My organisation’s spare parts fulfilment
method is likely a competitive strength” (4.04).
The χ2 test results indicate that out of the examined 20 items that were answered by the
respondents who were self-identified as “Customers and Suppliers”, only one item was not
supported. This is a strong indication that the adoption of AM would influence the
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

performance of SPSC and, thus, the null hypothesis Ho is rejected. In addition,

the respondents’ agreement with the majority of examined items indicates that the adoption
of AM is perceived to positively impact supply and minimise certain SRs. Thus, the third
alternative hypothesis H3 is accepted. Furthermore, the respondents’ agreed that the
adoption of AM will lead to: SP availability when needed, orders delivered on time, and
customised SP and this will ultimately improve the overall IP. Thus, the fourth alternative
hypothesis H4 is accepted.

5. Conclusions and recommendations

This chapter provides insights about key research conclusions related to potential impact of
AM adoption on CSy to price and lead time and on overall CS. This part also provides views
about the potential impact of AM adoption on SR and on overall inventory management.
Furthermore, three main recommendations are developed based on the findings of this
research. This is followed by a summary of the research theoretical and practical
contribution. In addition, a number of research limitations and suggestions for future
research are presented. Finally, a “lessons learned” section is added that could simplify
future survey research in this field.

5.1 Conclusions
The study aimed to assess possible impact on firm IP and CS should AM replace current
methods to fulfil customer demand for SP. Moreover, particular focus was given to CSy to
price and lead time and SR in order to gauge their influence (if any) on attitudes towards AM
introduction. The opinions of suppliers and customers were sought in order to estimate
what impact the adoption of AM may have in the SPSC as a whole.
5.1.1 CSy to price and lead time. The used items in the surveys were designed to gauge
the respondents’ perception about pricing SP and order lead times. The results showed that
these two elements are crucial to the organisation and the perfect ideal option is to have the
required SP acquired on time and at the desired/acceptable prices. Assuming AM is
adopted, the respondents were asked to rate a number of items related to the potential
impacts of such adoption. The top perceived benefit from the adoption of AM was the
availability of SP when organisations require them. This indicates the concerns
organisations usually have regarding obtaining the needed SP on time and at the desired
prices. The second top perceived potential benefit organisations gain from adopting AM was
“on-time delivery”. The third most supported item was the fact that SP demand is difficult
to forecast accurately. Thus, the adoption of AM will lead to better availability of SP,
 Downloaded by University of Florida At 19:20 18 January 2018 (PT)

JMTM

suppliers”)
Table VII.

“customers and
(respondents as
and supply risk
Survey results about
inventory performance
χ2test – ( frequencies
residual values)
Cronbach’s Factor Item
Survey question Mean α loading SD D N A SA supported

8. Based on your organisation’s current spare parts inventory performance, please rate the following statements
Spare parts buffer stock is maintained to offset possible supply risk 3.86 0.765 0.924 −5.6 −3.6 −0.6 10.4 −0.6 Yes
Spare parts buffer stock is maintained to offset possible spikes in customer demand 3.61 0.762 0.773 −5.6 −0.6 1.4 4.4 −0.6 Yes
Some spare parts stock held is at risk of obsolescence 3.71 0.765 0.950 −4.6 −3.6 0.4 8.4 −0.4 Yes
Spare parts demand is difficult to forecast with accuracy 4.04 0.768 0.845 −5.6 −3.6 −3.6 11.4 1.4 Yes
Spare parts with low demand have long lead times 3.79 0.769 0.867 −5.6 −2.6 0.4 7.4 0.4 Yes
9. Assuming your organisation adopts additive manufacturing to manage spare parts inventory performance, please rate the following statements
Spare parts buffer stock is maintained to offset possible supply risk 3.07 0.758 0.902 −3.6 1.4 −3.6 10.4 −5.6 Yes
Spare parts buffer stock is maintained to offset possible spikes in customer demand 3.07 0.757 0.944 −3.6 0.4 −0.6 7.4 −4.6 Yes
Some spare parts stock held is at risk of obsolescence 2.93 0.757 0.889 −4.6 5.4 −3.6 4.4 −3.6 No
Spare parts demand is difficult to forecast with accuracy 3.79 0.769 0.790 −5.6 −2.6 −3.6 10.4 0.4 Yes
Spare parts with low demand have long lead times 3.14 0.764 0.939 −4.6 0.4 2.4 3.4 −2.6 Yes
10. Assuming your organisation adopts additive manufacturing as its spare parts fulfilment method, please rate the following statements
Spare parts are likely available when my organisation’s customers require them 3.89 0.776 0.929 −5.6 −3.6 −0.6 9.4 0.4 Yes
Customer orders for spare parts are likely delivered on time 3.96 0.778 0.870 −5.6 −3.6 −1.6 9.4 1.4 Yes
Spare parts are likely customised to my organisation’s customers’ needs 3.86 0.777 0.842 −5.6 −4.6 1.4 9.4 −0.6 Yes
Spare parts procured for sale are likely not subject to supply risk 3.50 0.773 0.870 −5.6 −1.6 3.4 6.4 −2.6 Yes
My organisation’s spare parts fulfilment method is likely a competitive strength 4.04 0.769 0.895 −5.6 −5.6 1.4 7.4 2.4 Yes
11. Please rate the following statements concerning supply risk
My organisation has a reliable spare parts supply chain 3.96 0.77 0.938 −5.6 −4.6 2.4 6.4 2.4 Yes
My organisation adopts a multi-sourcing strategy for spare parts procurement 3.89 0.763 0.929 −5.6 −3.6 −1.6 11.4 −0.6 Yes
Spare parts required by my organisation’s customers are readily available in the market 3.29 0.785 0.918 −3.6 0.4 1.4 2.4 −0.6 Yes
My organisation has buying strength in the spare parts market 3.43 0.765 0.888 −5.6 −0.6 0.4 6.4 −1.6 Yes
My organisation’s suppliers are subject to regular reliability review 3.68 0.762 0.873 −5.6 −1.6 −2.6 11.4 −3.6 Yes
Notes: 1 ¼ Strongly disagree; 2 ¼ Disagree; 3 ¼ Neutral; 4 ¼ Agree; 5 ¼ Strongly agree; SD ¼ Strongly disagree; D ¼ Disagree; N ¼ Neutral; A ¼ Agree; SA ¼ Strongly agree
 on-time order delivery, shorter lead times, less inventory holding and better inventory Additive
management and performance, and overall, less SR. Such potential benefits have also been manufacturing
highlighted by: Attaran (2017), Eggenberger et al. (2017), Pour and Zanoni (2017) and
Tziantopoulos et al. (2016).
In contrast, the respondents indicated that SP were not priced at a level that their
organisations are happy to pay and this was the least supported item. We think that, at this
early stage of AM adoption, SP’ direct prices maybe higher but organisations are likely to
save on other indirect costs e.g. inventory management and downtime associated cost.
Thus, organisations would ultimately save on the SP total acquisition costs. However,
the more AM advances the cheaper the SP prices become. Thus, CS with better lead time is
likely to improve but it may not do so (at this stage of AM development) when it is related to
SP prices.
5.1.2 SR. Further used items in the surveys were designed to gauge the respondents’
level of concern about SRs and IP. The respondents showed high level of agreement with the
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

negative impacts disruptions in supply may yield to the business. They also perceived the
adoption of AM within the SPSC as a good approach to minimise certain SRs and make
inventory management and performance more effective. Similar potential benefits have
also been highlighted by other researchers, e.g. Cohen et al. (2014), Durach et al. (2017),
Dwivedi et al. (2017), Holmström and Partanen (2014), Li et al. (2017), Oettmeier and
Hofmann (2016), Pour and Zanoni (2017) and Sasson and Johnson (2016). However, the
respondents of this study also believed that adopting AM would minimise certain SRs but
not eliminate all of them. Perhaps such scepticism was influenced by the fact that there are
other risks that may contribute to the supply disruption, and overcoming them cannot
be achieved by the adoption of AM. Overall, the study showed a strong agreement that the
adoption of AM would minimise SR and lead to better inventory management and
performance. Thus, the adoption of AM will change how SRs are managed and will change
the notion of supply chain risk management and resiliency forever (Caccamo, 2016).

5.2 Recommendations
As a result of the research findings, the authors propose the following recommendations: in
the light of the unanimous opinion held by the “Supplier” and “Customer & Supplier”
groups that AM would likely lead to improved IP, suppliers competing in the SPSC are
encouraged to consider AM adoption. Also, ensuing AM related costs be absorbed by
suppliers competing in the SPSC where possible, rather than recovery thereof being
attempted through increased sales prices. By doing so, the trade-off customers face between
high purchase price and short delivery lead time will be removed, allowing suppliers to
capitalise on increased CS and competitive advantage this dynamic capability affords. Or,
if AM related costs are recovered through increased sales prices, suppliers competing in the
SPSC monitor the impact to CS and in particular the rate of customer attrition this decision
has post AM introduction. This recommendation recognises the “Customer” and “Customer
& Supplier” groups’ increased dissatisfaction with price, and (in the case of the former
group) reduced willingness to recommend SP suppliers.

5.3 Theoretical contributions

The research takes a novel perspective on AM in the SPSCs by illuminating the supplier and
buyer perspective based on empirical data. This research provides new insights about the
appreciation of the use of AM in SPSCs of small and medium sized manufacturing
companies located in the UK. Through analysis of the disparate opinions of customers and
suppliers regarding impact of AM, the possible risks and rewards of the technology from a
range of perspectives have been gleaned. These perspectives are considered to be of interest
 JMTM to practitioners contemplating adoption of AM for fulfilment of SP demand. Moreover, the
study is considered of potential interest to researchers as it aims to expand extant literature
on AM in the context of SP fulfilment.

5.4 Practical contributions

This research is deemed to have contributed to extant literature in several ways. First,
whilst Holmström et al. (2010), Khajavi et al. (2014), and Liu et al. (2014) concentrated on AM
adoption in the aerospace industry, this research broaches its applicability within two other
industrial segments: “Industrial and Commercial Machinery and Computer Equipment” and
“Measuring, Analysing and Controlling Instruments, Photographic, Medical and Optical
Instruments”. Second, the study is considered of potential interest to practitioners in
multiple industries because it solicited opinions on actual or expected value of AM as a tool
to balance the contradictory goals of satisfying the customer whilst minimising firms’
investment in potentially slow-moving SP inventory. Finally, despite its age, evidence of
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

AM’s commercial viability is still in its infancy. Therefore, the study is also considered of
potential value to AM technology developers as it is likely to indicate demand for AM in
SPSC of unexploited industries. The research is also considered beneficial to developers of
AM wishing to increase market share due to positive reaction of entities within industrial
and commercial machinery and computer equipment, and measuring, analysing and
controlling instrumentation industrial segments.

5.5 Limitations and suggested future research

This research predominantly concentrated on the opinions of individuals working for
organisations in two industrial segments within the UK manufacturing sector. Future
research may explore the adoption of AM within other industrial segments in the UK and in
other industrialised counties. This will make the generalisability of the research findings
more applicable. A further possible limitation concerns size of organisations surveyed;
approximately 82 per cent of survey respondents were associated with organisations
consisting fewer than 100 employees. The results may differ for larger organisations and
further studies should target participants working for medium and large organisations.
Furthermore, every effort was made to collect data from a larger group but our study was
based on the opinions of 69 respondents. Thus, collecting data from a larger sample is
strongly recommended in future studies. Finally, the used survey in this research was
mainly developed using close-ended type questions. This limits the ability to collect more
personalised qualitative opinions about the tested items. Thus, it may be wise to add some
open-ended questions when necessary to collect more valuable feedback.

5.6 Lessons learned

The authors believe the following adaptations to survey administration and design may
lead to an increased response rate and improve future research in this field. First, more time
could be allotted to the collection of responses. Due to time constraints, the current survey
was open approximately ten days. This meant it was not possible to carry out a trial,
which – following adjustments – may have reduced the 221 emails out of 6,786-bounce rate.
Moreover, a longer window of opportunity to complete the survey would naturally facilitate
a greater yield. Second, the Likert items of questions measuring CSy to delivery lead time
should be written in such a way to encourage adequate internal consistency. The question
posed in the current study is found below.
Likert items measuring CSy to delivery lead time.
Please rate the following statement concerning delivery lead time of SP:
• My organisation prefers to pay more to receive SP sooner.
 • My organisation prefers to buy SP immediately when required to achieve lower Additive
equipment downtime. manufacturing
• If delivery lead time of SP increases, my organisation will look for alternative supplier(s).
• Shorter lead time is more important than lower price.
• My organisation prefers fixed delivery lead time of SP.

References
Allred, C.R., Fawcett, S.E., Wallin, C. and Magnan, G.M. (2011), “A dynamic collaboration capability as
a source of competitive advantage”, Decision Sciences, Vol. 42 No. 1, pp. 129-161.
Arafa, A. and ElMaraghy, W. (2011), “Manufacturing strategy and enterprise dynamic capability”,
CIRP Annals – Manufacturing Technology, Vol. 60, pp. 507-510.
Arnold, R.D. and Wade, J.P. (2015), “A definition of systems thinking: a systems approach”,
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

Procedia Computer Science, Vol. 44, pp. 669-678.

Attaran, M. (2017), “Additive manufacturing: the most promising technology to alter the supply chain
and logistics”, Journal of Service Science and Management, Vol. 10 No. 2017, pp. 189-205.
Bagozzi, R.P. (2011), “Measurement and meaning in information systems and organizational research:
methodological and philosophical foundations”, MIS Quarterly, Vol. 35 No. 2, pp. 261-292.
Baines, T. and Lightfoot, H. (2013), Made to Serve: How Manufacturers can Compete through
Servitization and Product Service Systems, John Wiley & Sons.
Barone, S., Casinelli, M., Frascaria, M., Paoli, A. and Razionale, A.V. (2016), “Interactive design of dental
implant placements through CAD-CAM technologies: from 3D imaging to additive
manufacturing”, International Journal on Interactive Design and Manufacturing, Vol. 10 No. 2,
pp. 105-117.
Barrow, D. and Kourentzes, N. (2016), “Distributions of forecasting errors of forecast combinations:
implications for inventory management”, International Journal of Production Economics,
Vol. 177, pp. 24-33.
Beyer, C. (2014), “Strategic implications of current trends in additive manufacturing”, Journal of
Manufacturing Science and Engineering, Vol. 136 No. 6, p. 64701.
Blome, C., Schoenherr, T. and Rexhausen, D. (2013), “Antecedents and enablers of supply chain agility
and its effect on performance: a dynamic capabilities perspective”, International Journal of
Production Research, Vol. 51 No. 4, pp. 1295-1318.
Caccamo, W. (2016), “3D printing blows up supply chain risk management [online]”, available at:
www.sdcexec.com/article/12241583/3d-printing-blows-up-supply-chain-risk-management
(accessed 24 December 2016).
Campbell, R.I., De Beer, D.J. and Pei, E. (2011), “Additive manufacturing in South Africa: building on the
foundations”, Rapid Prototyping Journal, Vol. 17 No. 2, pp. 156-162.
Caputo, A., Marzi, G. and Pellegrini, M. (2016), “The internet of things in manufacturing innovation
processes: development and application of a conceptual framework”, Business Process
Management Journal, Vol. 22 No. 2, pp. 383-402.
Chen, D., Heyer, S., Ibbotson, S., Salonitis, K., Steingrímsson, J.G. and Thiede, S. (2015), “Direct digital
manufacturing: definition, evolution, and sustainability implications”, Journal of Cleaner
Production, Vol. 107, pp. 615-625.
Chudoba, B. (2010), “Does adding one more question impact survey completion rate?”, SurveyMonkey®
Blog, 12 August, available at: www.surveymonkey.com/blog/210/12/08/survey_questions_and_
completion_rates/ (accessed 9 October 2016).
Cohen, D., Sargeant, M. and Somers, K. (2014), “3-D printing takes shape”, McKinsey Quarterly,
January, available at: www.mckinsey.com/business-functions/operations/our-insights/3-d-
printing-takes-shape (accessed 24 December 2016).
 JMTM Cooper, D. et al. (2012), “Additive manufacturing for product improvement at red bull technology”,
Materials and Design, Vol. 41, pp. 226-230.
Costantino, N., Dotoli, M., Falagario, M., Fanti, M.P. and Mangini, A.M. (2012), “A model for supply
management of agile manufacturing supply chains”, International Journal of Production
Economics, Vol. 135 No. 1, pp. 451-457.
Cronbach, L.J. (1970), Essentials of Psychological Testing, 3rd ed., Harper & Row, New York, NY.
Data HQ Limited (2016a), “Fact sheet: business mailing list [online]”, available at: https://datahq.co.uk/
uploads/documents/DATA-HQ-FACT-SHEET-BUSINESS-MAILING-LIST_LowRes.pdf (accessed
4 October 2016).
Data HQ Limited (2016b), “Manufacturing”, available at: https://datahq.co.uk/marketing-lists/business-
mailing-list/manufacturing (accessed 8 October 2016).
Davenport, E.C., Davison, M.L., Liou, P.Y. and Love, Q.U. (2015), “Reliability, dimensionality, and
internal consistency as defined by Cronbach: distinct Albeit related concepts”, Educational
Measurement: Issues & Practice, Vol. 34 No. 4, pp. 4-9.
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

Durach, C.F., Kurpjuweit, S. and Wagner, S.M. (2017), “The impact of additive manufacturing on
supply chains”, International Journal of Physical Distribution & Logistics Management, Vol. 47
No. 10, pp. 954-971.
Dwivedi, G., Srivastava, S.K. and Srivastava, R.K. (2017), “Analysis of barriers to implement additive
manufacturing technology in the Indian automotive sector”, International Journal of Physical
Distribution & Logistics Management, Vol. 47 No. 10, pp. 972-991.
Easterby-Smith, M., Thorpe, R. and Jackson, P. (2012), Management Research, 4th ed.,
Sage Publications, London.
Eggenberger, T., Oettmeier, K. and Hofmann, E. (2017), “Additive manufacturing in automotive spare
parts supply chains – a conceptual scenario analysis of possible effects”, International Conference
on Additive Manufacturing in Products and Applications, Springer, Cham, pp. 223-237.
Ennis, C.D. (1999), “A theoretical framework: the central piece of a research plan”, Journal of Teaching
in Physical Education, Vol. 18 No. 2, pp. 129-140.
Eroglu, C. and Hofer, C. (2011), “Lean, leaner, too lean? The inventory-performance link revisited”,
Journal of Operations Management, Vol. 29 No. 4, pp. 356-369.
Frazier, W. (2014), “Metal additive manufacturing: a review”, Journal of Materials Engineering &
Performance, Vol. 23 No. 6, pp. 1917-1928.
Gardan, J. (2016), “Additive manufacturing technologies: state of the art and trends”, International
Journal of Production Research, Vol. 54 No. 10, pp. 3118-3132.
General Electric Company (2011), “European SME capex barometer: SMEs’ business sentiment and
capital investment in Europe’s for biggest economies [online]”, available at: www.gecapital.com/
en/pdf/GE_Capital_European_SME_Capex.pdf (accessed 8 October 2016).
George, D. and Mallery, P. (2003), SPSS for Windows Step by Step: A Simple Guide and Reference,
4th ed., Allyn & Bacon, Boston, MA.
Glasschroeder, J., Prager, E. and Zaeh, M.F. (2015), “Powder-bed-based 3D-printing of function
integrated parts”, Rapid Prototyping Journal, Vol. 21 No. 2, pp. 207-215.
Hahn, G. and Leucht, A. (2015), “Managing inventory systems of slow-moving items”, International
Journal of Production Economics, Vol. 170 Part B, pp. 543-550.
Hair, J., Black, W., Babin, B., Anderson, R. and Tat ham, R. (2005), Multivariate Data Analysis, 6th ed.,
Prentice-Hall, Englewood Cliffs, NJ.
Hedenstierna, C.P.T. and Disney, S.M. (2016), “Inventory performance under staggered deliveries and
autocorrelated demand”, European Journal of Operational Research, Vol. 249 No. 3, pp. 1082-1091.
Holmström, J. and Partanen, J. (2014), “Digital manufacturing-driven transformations of service supply
chains for complex products”, Supply Chain Management: An International Journal, Vol. 19
No. 4, pp. 421-430.
 Holmström, J., Partanen, J., Tuomi, J. and Walter, M. (2010), “Rapid manufacturing in the spare parts Additive
supply chain: alternative approaches to capacity deployment”, Journal of Manufacturing manufacturing
Technology Management, Vol. 21 No. 6, pp. 687-697.
Huxtable, J. and Schaefer, D. (2016), “On servitization of the manufacturing industry in the UK”,
Procedia CIRP, 52, The Sixth International Conference on Changeable, Agile, Reconfigurable and
Virtual Production (CARV2016), pp. 46-51.
Janssen, G.R., Blankers, I.J., Moolenburgh, E.A. and Posthumus, A.L. (2014), TNO: The Impact of 3-D
Printing on Supply Chain Management, TNO, Hague.
Khajavi, S., Partanen, J. and Holmström, J. (2014), “Additive manufacturing in the spare parts supply
chain”, Computers in Industry, Vol. 65 No. 1, pp. 50-63.
Kim, B. and Park, C. (2013), “Firms’ integrating efforts to mitigate the tradeoff between controllability
and flexibility”, International Journal of Production Research, Vol. 51 No. 4, pp. 1258-1278.
Knofius, N., van der Heijden, M.C. and Zijm, W.H.M. (2016), “Selecting parts for additive manufacturing in
service logistics”, Journal of Manufacturing Technology Management, Vol. 27 No. 7, pp. 915-931.
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

Kothman, I. and Faber, N. (2016), “How 3D printing technology changes the rules of the game: insights
from the construction sector”, Journal of Manufacturing Technology Management, Vol. 27 No. 7,
pp. 932-943.
Kruth, J., Leu, M. and Nakagawa, T. (1998), “Keynote papers: progress in additive manufacturing and
rapid prototyping”, CIRP Annals – Manufacturing Technology, Vol. 47 No. 2, pp. 525-540.
Li, Y., Jia, G., Cheng, Y. and Hu, Y. (2017), “Additive manufacturing technology in spare parts supply
chain: a comparative study”, International Journal of Production Research, Vol. 55 No. 5,
pp. 1498-1515.
Lindell, M.K. and Whitney, D.J. (2001), “Accounting for common method variance in cross-sectional
designs”, Journal of Applied Psychology, Vol. 86 No. 1, pp. 114-121.
Liu, P., Huang, S.H., Mokasdar, A., Zhou, H. and Hou, L. (2014), “The impact of additive manufacturing
in the aircraft spare parts supply chain: supply chain operation reference (scor) model based
analysis”, Production Planning & Control, Vol. 25 Nos 13-14, pp. 1169-1181.
Manerikar, V. and Manerikar, S. (2015), “Cronbach’s alpha”, Aweshkar Research Journal, Vol. 19 No. 1,
pp. 117-119.
Mawale, M., Kuthe, A. and Dahake, S. (2016), “Additive layered manufacturing: State-of-the-art
applications in product innovation”, Concurrent Engineering: Research & Applications, Vol. 24
No. 1, pp. 94-102.
Meadows, D. (2008), Thinking in Systems: A Primer, Chelsea Green Publishing, Vermont.
Meisel, N.A., Williams, C.B., Ellis, K.P. and Taylor, D. (2016), “Decision support for additive
manufacturing deployment in remote or austere environments”, Journal of Manufacturing
Technology Management, Vol. 27 No. 7, pp. 898-914.
Mellor, S., Hao, L. and Zhang, D. (2014), “Additive manufacturing: a framework for implementation”,
International Journal of Production Economics, Vol. 149, pp. 194-201.
Minaei-Bidgoli, B., Tan, P.N. and Punch, W.F. (2004), “Mining interesting contrast rules for a web-based
educational system”, Machine Learning and Applications, 2004. Proceedings. 2004 International
Conference on IEEE, pp. 320-327.
Oettmeier, K. and Hofmann, E. (2016), “Impact of additive manufacturing technology adoption on
supply chain management processes and components”, Journal of Manufacturing Technology
Management, Vol. 27 No. 7, pp. 944-968.
Ohno, T. (1988), Toyota Production System: Beyond Large-Scale Production, Productivity Press,
Cambridge, MA.
Oliver, R.L. (1997), Satisfaction: A Behavioral Perspective on the Customer, New York, NY.
Peres, F. and Noyes, D. (2006), “Envisioning e-logistics developments: making spare parts in situ and
on demand – state of the art and guidelines for future developments”, Computers In Industry,
Vol. 57 No. 6, pp. 490-503.
 JMTM Podsakoff, P.M., MacKenzie, S.B., Lee, J. and Podsakoff, N.P. (2003), “Common methods biases in
behavioral research: a critical review of the literature and recommended remedies”, Journal of
Applied Psychology, Vol. 88 No. 5, pp. 879-903.
Pour, M.A. and Zanoni, S. (2017), “Impact of merging components by additive manufacturing in spare
parts management”, Procedia Manufacturing, Vol. 11, pp. 610-618.
Purch (2016a), “Survey software reviews [online]”, available at: www.toptenreviews.com/business/
software/best-survey-software/ (accessed 5 October 2016).
Purch (2016b), “SurveyMonkey gold review”, available at: www.toptenreviews.com/business/software/
best-survey-software/surveymonkey-review/ (accessed 28 December 2016).
Richardson-Harman, N., Phelps, T., Mooney, P. and Ball, R. (1998), “Consumer perceptions of fruit
production technologies”, New Zealand Journal of Crop and Horticultural Science, Vol. 26 No. 3,
pp. 181-192.
Rivard, S. and Huff, S.L. (1988), “Factors of success for end-user computing”, Communications of the
ACM, Vol. 31 No. 5, pp. 552-561.
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

Robson, C. (2011), Real World Research, 3rd ed., John Wiley and Sons, Chichester.
Royal Academy of Engineering (2013), “Additive Manufacturing: opportunities and constraints”, Additive
Manufacturing: Changing the UK Manufacturing Landscape, London, 23 May, available at: www.
raeng.org.uk/publications/reports/additive-manufacturing (accessed 29 September 2016).
Ruan, J., Sparks, T.E. and Fan, Z. (2014), “A review of layer based manufacturing processes for metals”,
available at: http://edge.rit.edu/edge/P10551/public/SFF/SFF%202006%20Proceedings/
Manuscripts/22-Ruan.pdf (accessed 21 September 2016).
Saalmann, P., Zuccolotto, M., da Silva, T.R., Wagner, C., Giacomolli, A., Hellingrath, B. and Pereira, C.E.
(2016), “Application potentials for an ontology-based integration of intelligent maintenance
systems and spare parts supply chain planning”, Procedia CIRP, Vol. 41, pp. 270-275.
Saenz, M., Revilla, E. and Knoppen, D. (2014), “Absorptive capacity in buyer-supplier relationships: empirical
evidence of its mediating role”, Journal of Supply Chain Management, Vol. 50 No. 2, pp. 18-40.
Sandström, C. (2016), “The non-disruptive emergence of an ecosystem for 3D printing – insights from
the hearing aid industry’s transition 1989-2008”, Technological Forecasting and Social Change,
Vol. 102, pp. 160-168.
Santos, J. (1999), “Cronbach’s alpha: a tool for assessing the reliability of scales”, Journal of Extension,
Vol. 37 No. 2, pp. 88-92.
Sasson, A. and Johnson, J.C. (2016), “The 3D printing order: variability, supercenters and supply chain
reconfigurations”, International Journal of Physical Distribution & Logistics Management, Vol. 46
No. 1, pp. 82-94.
Steenhuis, H.J. and Pretorius, L. (2017), “The additive manufacturing innovation: a range of
implications”, Journal of Manufacturing Technology Management, Vol. 28 No. 1, pp. 122-143.
Tatham, P., Loy, J. and Peretti, U. (2015), “Three dimensional printing – a key tool for the humanitarian
logistician?”, Journal Of Humanitarian Logistics & Supply Chain Management, Vol. 5 No. 2,
pp. 188-208.
Teece, D., Pisano, G. and Shuen, A. (1997), “Dynamic capabilities and strategic management”,
Strategic Management Journal, Vol. 18 No. 7, pp. 509-533.
The Financial Times (2016a), “Definition of customer satisfaction”, available at: http://lexicon.ft.com/
Term?term=customer-satisfaction (accessed 1 September 2016).
The Financial Times (2016b), “Definition of price sensitivity”, available at: http://lexicon.ft.com/Term?
term=price-sensitivity (accessed 1 September 2016).
Thomas, D. (2016), “Costs, benefits, and adoption of additive manufacturing: a supply chain
perspective”, International Journal of Advanced Manufacturing Technology, Vol. 85 Nos 5-8,
pp. 1-20.
Thwaites Bee, D. and Murdoch-Eaton, D. (2016), “Questionnaire design: the good, the bad and the pitfalls”,
Archives of Disease In Childhood: Education and Practice Edition, Vol. 101 No. 4, pp. 210-212.
 Tourangeau, R., Rips, L.J. and Rasinski, K. (2000), The Psychology of Survey Response, Additive
Cambridge University Press, Cambridge. manufacturing
Townsend, V. and Urbanic, J. (2012), “Relating additive and subtractive processes in a teleological and
modular approach”, Rapid Prototyping Journal, Vol. 18 No. 4, pp. 324-338.
Tziantopoulos, K., Vlachos, D. and Iakovou, E. (2016), “Additive manufacturing: a decision support
system for spare parts inventory management”, 11th MIBES Conference, 22-24 June, Heraklion,
pp. 521-528.
van Gigch, J.P. (1979), “A methodological comparison of the science, systems and metasystem
paradigms”, International Journal of Man-Machine Studies, Vol. 11 No. 5, pp. 651-663.
Waller, J.L. (2012), How to Perform and Interpret Chi-Square and T-Tests, SAS Institute Inc.,
Durham, NC.
Wernerfelt, B. (1984), “A resource-based view of the firm”, Strategic Management Journal, Vol. 5 No. 2,
pp. 171-180.
Wits, W., García, J. and Becker, J. (2016), “How additive manufacturing enables more sustainable end-user
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

maintenance, repair and overhaul (MRO) strategies”, Procedia CIRP, 40, 13th Global Conference on
Sustainable Manufacturing – Decoupling Growth from Resource Use, pp. 693-698.
Wolbert, J. (2014), “The consequences of a gray market”, available at: www.micross.com/blog/2014/02/
the-consequences-of-a-gray-market/ (accessed 22 September 2016).
Zha, W. and Anand, S. (2015), “Technical paper: geometric approaches to input file modification for
part quality improvement in additive manufacturing”, Journal of Manufacturing Processes,
Vol. 20, pp. 465-477.
Zistl, S. (2014), “3D printing: facts & forecasts”, available at: www.siemens.com/innovation/en/home/
pictures-of-the-future/industry-and-automation/Additive-manufacturing-facts-and-forecasts.
html (accessed 19 September 2016).
Zsidisin, G.A. (2003), “A grounded definition of supply risk”, Journal of Purchasing and Supply
Management, Vol. 9 Nos 5-6, pp. 217-224, doi: 10.1016/j.pursup.2003.07.002.

Further reading
Economic and Social Research Council (2016), “What is ethics?”, available at: www.esrc.ac.uk/funding/
guidance-for-applicants/research-ethics/frequently-raised-questions/what-is-ethics/ (accessed
23 October 2016).
Emerson, R. (2015), “Causation and pearson’s correlation coefficient”, Journal of Visual Impairment &
Blindness, Vol. 36 No. 3, pp. 242-244.
SurveyMonkey Inc. (2016), “Additive manufacturing, customer”, available at: www.surveymonkey.net/
collect/email/manage?sm=HmTERzv6BASdCaEpO_2Bm2JQW_2BebS39pPuB5fJF8qk0
7kOK_2FdA_2BFkpnYRHSQerqna_2F (accessed 28 November 2016).
Veselinova, E. and Samonikov, M. (2012), “SMEs inovation and growth in EU”, Management (1820-0222),
Vol. 64, pp. 87-94.

About the authors

Melanie Muir has gained over ten years’ manufacturing industry experience, within UK branches of
large, publicly listed corporations and smaller private equity backed groups. As a Fellow Chartered
Certified Accountant, experienced finance executive and business partner, Melanie’s career has
predominantly focussed on the finance discipline. However, following role expansion to lead
procurement and demand planning, in addition to finance, Melanie recently completed an MSc in
Operations and Supply Chain Management from the University of Liverpool in the UK. Melanie is
looking forward to completing a PhD in order to contribute to operations and supply chain
management thought leadership, in addition to consulting within the field.
Abubaker Haddud is a Visiting Scholar at the Eastern Michigan University and a Lead Faculty on
the online management programmes at the University of Liverpool in England. He has a PhD in
Engineering Management from Eastern Michigan University, where he was a Fulbright scholar,
 JMTM and an MBA from Coventry University in the UK. His teaching and research interests focus on
operations and supply chain management, technology management, lean manufacturing, and business
performance measurement and analysis. Dr Haddud has several years of university teaching
experience within the technology management, business, and management domains at the
undergraduate and postgraduate levels. Dr Haddud is an active researcher, and his recent research
activities have focussed on the use of disruptive technologies to create unique competitive advantages
and to enhance business performance. Dr Haddud also has more than ten years of industrial work
experience in different sectors. Abubaker Haddud is the corresponding author and can be contacted at:
[email protected]
Downloaded by University of Florida At 19:20 18 January 2018 (PT)

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]
Bottlenecks, Modules, and Dynamic Architectural Capabilities

Oxford Handbooks Online

Bottlenecks, Modules, and Dynamic Architectural
Capabilities  
Carliss Y. Baldwin
The Oxford Handbook of Dynamic Capabilities
Edited by David J. Teece and Sohvi Heaton

Subject: Business and Management, Technology and Knowledge Management

Online Publication Date: Mar 2018 DOI: 10.1093/oxfordhb/9780199678914.013.011

Abstract and Keywords

How do firms create and capture value in large technical systems? In this paper, I argue
that the points of both value creation and value capture are the system’s bottlenecks.
Bottlenecks arise first as important technical problems to be solved. Once the problem is
solved, Then the solution in combination with organizational boundaries and property
rights can be used to capture a stream of rents. The tools a firm can use to manage
bottlenecks are, first, an understanding first of the technical architecture of the system;
and, second, an understanding of the industry architecture in which the technical system
is embedded. Although these tools involve disparate bodies of knowledge, they must be
used in tandem to achieve maximum effect. Dynamic architectural capabilities provide
managers with the ability to see a complex technical system in an abstract way and
change the system’s structure to manage bottlenecks and modules in conjunction with
the firm’s organizational boundaries and property rights.

Keywords: architecture, architectural knowledge, dynamic capabilities, bottleneck, modularity, organization

design, organization boundaries, property rights

1 Introduction
Large technical systems made up of many interacting components are becoming more
common every day. Digital information technology has made it possible to create ever
larger technical systems with faster flows of material and information. At the same time,
digital technology provides a common substrate for design, communication, and control.
Spurred by the commonality of bits, a multitude of previously separate and small

Page 1 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

technical systems are expanding and converging towards one big system (Yoffie, 1997;
Baldwin and Clark, 2000; Weinberger, 2008; Brynjolfsson and McAfee, 2014).

How do firms create and capture value in the increasingly large technical systems
emerging today? In this paper, I argue that the points of both value creation and value
capture are the system’s bottlenecks. Bottlenecks arise first as important technical
problems to be solved. Once the problem is solved, the solution in combination with
organization boundaries and property rights can be used to capture a stream of rents.

The tools a firm can use to manage bottlenecks are, first, an understanding of the
technical architecture of the system; and, second, an understanding of the industry
architecture in which the technical system is embedded. In 1990, Henderson and Clark
(1990) defined architectural knowledge as knowledge about the technical architecture of
the products of a particular firm. However, today large technical systems have expanded
to involve many firms, and thus managers must have a sophisticated understanding of
both technical and industry architectures. Although they are based on disparate bodies of
knowledge, these architectural tools must be used in tandem to achieve maximum effect.

Architectural capabilities are an important subset of dynamic capabilities that provide

managers with the ability to see a complex technical system in an abstract way and
change the system’s structure by rearranging its components (Teece et al., 1997; Pisano
and Teece, 2007). Purposeful architectural change can then be used to create and
capture value at different points in the technical system. In this paper I argue that value-
enhancing architectural change arises through the effective management of bottlenecks
and modules in conjunction with the firm’s organizational boundaries and property rights.

The rest of the paper is organized as follows. Section 2 defines technical and industry
architectures and explains how they are related. In Section 3, building on prior work, I
describe two types of bottlenecks, technical and strategic, and show how bottlenecks are
related to the modular structure of a technical system and to the organizational
boundaries and property rights of firms in the system. Section 4 deals with the
management of technical bottlenecks. Here the most basic question is whether to solve a
particular technical bottleneck by focusing on only a few components or by controlling
many components. Section 5 describes how different forms of modularity can protect or
expose strategic bottlenecks. Section 6 describes how firms can invest to acquire
architectural knowledge and dynamic architectural capabilities. Section 7 concludes.

2 Technical and Industry Architecture

To answer the question, how do firms create and capture value in large technical systems,
it is first necessary to develop ways of describing such systems. A technical system is
made up of conceptually separable components, each of which “does something” to
contribute to the overall performance of the system (Arthur, 2009). The components of

Page 2 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

the system, taken together, are jointly complementary: the value of the system as a whole
is greater than the value of the components disassembled (otherwise there would be no
point in building the system).

Architecture is the term given to the abstract description of large technical systems. The
concept of an “architecture” for man-made systems dates back to Herbert Simon. In his
classic paper “The Architecture of Complexity,” Simon proposed a theory of the structure
of complex systems based on the concepts of hierarchy and modularity.1 In brief, the
architecture of a system defines its components, describes interfaces between
components, and specifies ways of testing performance. All man-made systems, including
all products and organizations, have architectures (Simon, 1981; Ulrich, 1995; Blaauw
and Brooks Jr., 1997; Baldwin and Clark, 2000; Whitney et al., 2004; Fixson, 2005; Arthur,
2009).2

A complete architectural description of a man-made thing includes accounts of its design

structure and its task structure. The design structure describes the finished artifact in
terms of its design elements and their relationships. Design structure can be conveyed
using words, pictures, blueprints, and diagrams. The task structure describes the
procedures that must be carried out to bring the artifact into being.3

Engineers and product designers are concerned with the material structure of a product
or process and the technical tasks needed to produce the product or carry out the
process. Their focus is on the technical architecture of the system. However, to bring the
components together and carry out the tasks specified by the technical architecture, a
social structure must be overlaid upon the engineering task structure. Resources must be
brought in, people recruited, tasks assigned, products transported and sold, and money
collected. These actions take place in the social domain, mediated by organizations and
markets.

As distinct from the technical architecture of a given product, Baldwin and Clark (2000)
define the contract structure of an organization to include a description of its internal
departments, compensation practices, boundaries, property rights, contracts, and
transactions. Contract structures are part of the general architecture of a technical
system: they are needed to assemble the resources required to carry out the tasks of
building and maintaining the system. But technical architectures and contract structures
involve different types of expertise and are usually handled by different individuals.
Specifically, the contract structure is generally in the hands of managers and lawyers,
while the technical architecture is in the hands of product designers, developers, and
engineers.

The term industry architecture is now used to refer to the aggregation of contract
structures for firms in a particular industry (Jacobides et al., 2006; Pisano and Teece,
2007). In what follows, I use the term “contract structure” to refer to the social

Page 3 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

architecture of a single firm and “industry architecture” to refer to the collection of

contract structures in a given industry.

Technical architectures and industry architectures are related, but not in a deterministic
way. Industry architectures exist to carry out the tasks specified by a given technical
architecture. Those tasks in turn may be related in different ways: for example, they may
be tightly or loosely coupled (Weick, 1969; Orton and Weick, 1990); or exhibit different
forms of dependency (Thompson, 1967; Baldwin and Clark, 2000; Baldwin, 2008). They
may be carried out by a close-knit team, by different departments within a single firm, or
by different firms. Sometimes the technical recipe is sensitive to the organizational
design, while at other times, a given recipe can be implemented equally well by many
different forms of organization.

Although the mapping of technical architecture to industry architecture is not one-to-one,

changes in the technical architecture of a system often beget changes in the industry
architecture and vice versa. For example, taking advantage of a new product design (a
new technical architecture) may entail setting up a new firm to make and sell the product
(a change in industry architecture). Conversely, when a firm divests a division, it directly
changes its industry architecture. The erstwhile division may then decide to eliminate
some products and introduce others, thereby changing its technical architecture.

As an empirical matter, technical and industry architectures often change in tandem.

When the technical architecture entails high levels of reciprocal task interdependency,
vertically integrated firms become dominant players (Chandler Jr., 1977; Abernathy and
Utterback, 1978; Fine, 1998; Fixson and Park, 2008). Conversely, as the technical
architecture is broken apart into separate modules, firms become more specialized and
the industry takes the form of a cluster or ecosystem (Langlois and Robertson, 1992;
Sanchez and Mahoney, 1996; Fine, 1998; Baldwin and Clark, 2000; Iansiti and Levien,
2004; Jacobides, 2005).

In Section 3, I define bottlenecks, modules, and a firm’s span of control in the context of
technical and industry architectures.

3 Bottlenecks, Modules, and a Firm’s Span of

Control

Page 4 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

In prior work, many scholars have argued that bottlenecks are key to understanding the
direction and pace of technological change and to capturing value in large, complex
technical systems. On the one hand, firms and individuals seeking to create value through
technology are said to look for and resolve the technology’s bottlenecks (Rosenberg, 1969,
1982, 1992; Hughes4 1983, 1987; Langlois and Robertson, 1992; Ethiraj, 2007; Arthur,
2009; Adner and Kapoor, 2010, 2014). On the other hand, firms wishing to capture value
are advised to control bottlenecks, become a bottleneck, and beware of bottlenecks
controlled by others (Teece, 1986; Langlois, 2002, 2007; Jacobides et al., 2006; Pisano and
Teece, 2007; Jacobides et al., 2012; Henkel and Hoffmann, 2014; Jacobides and Tae,
2014).

But what is a bottleneck?

3.1 Bottlenecks Defined

In common usage, a bottleneck is a narrow place that obstructs a flow of water or traffic,
for example. Thus in a road system, if all routes pass through a single stretch (say, a
bridge or a mountain pass), and that part of the system is a source of congestion, then it
is a bottleneck. More generally, a bottleneck is a component in a complex system whose
performance significantly limits the performance of the system as a whole (see, for
example, Goldratt, 1990, 2014).

Consistent with common usage, in what follows, I define a bottleneck as a critical part of
a technical system that has no—or very poor—alternatives at the present time. There may
be one or many bottlenecks in a given system but each has the dual properties that (1) it
is necessary to the functioning of the whole; and (2) there is no good way around it. Thus
to know that something is a bottleneck, the observer must see it in relation to a larger
system, know what constitutes good system-level performance, and understand how the
bottleneck constricts that performance.

In business, there are two types of bottlenecks, technical and strategic. With a technical
bottleneck, the hindrance to performance derives from physical properties of the system.
For example, in a railroad system, if there is no bridge over a river and goods must be
taken onto barges and reloaded on the other side, then the river constitutes a technical
bottleneck. It impedes the performance of the whole system and there is no good way
around it.

Building a bridge can solve the problem of technical performance, but the owner of the
bridge can charge a toll. The bridge plus the ability to control it then constitutes a
strategic bottleneck. The former system of boats and barges is far less efficient, hence
travellers and shippers have no good alternative except to use the bridge and pay the
toll.5

Page 5 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

3.2 Technical Bottlenecks and Modules

There are three basic subtypes of technical bottlenecks in man-made systems. First, some
necessary components of the minimally functional system may not exist at all. Let us call
this a sine qua non bottleneck. Second, in systems involving flows, the capacity of the
system will be constrained by the components with the least capacity relative to their
frequency of use. Call this a flow bottleneck. Third, in systems requiring parts that match
or fit together, the performance of the system as a whole will be constrained by
mismatched components. Call this a matching bottleneck.

Sine qua non and flow bottlenecks both involve a mismatch of elements. In a sine qua non
bottleneck, the mismatch is the non-existence of a critical solution to a sub-problem. In a
flow bottleneck, the mismatch is in flow capacity. Hence these types of bottlenecks can be
viewed as interesting special examples of the general case of matching bottlenecks.
However, the three subtypes have different implications for managerial action, thus it is
useful to distinguish between them. (Box 1 gives further details and examples of the three
bottleneck subtypes.)

Technical bottlenecks can be located within the system and mapped onto specific
components in the system’s technical architecture. An intensive bottleneck is caused by
deficiency in one component or subsystem, whereas extensive bottlenecks are caused by
deficiencies in several. (Because technologies are nested systems, the line between
intensive and extensive bottlenecks is somewhat arbitrary. But the distinction is useful for
strategy, as we shall see in Subsections 4.1 and 4.2.)

It is important to be clear that bottlenecks are theoretically distinct from the modules of a
technical system. Technical bottlenecks are problems that exist whether the designer
wants them or not. (Mostly designers would prefer that bottlenecks not exist.) Technical
bottlenecks are uncovered by identifying functional relationships between the
characteristics of components (their capacity, size, strength, shape, etc.) and the
performance of the system (good or bad).

In contrast, modules are subsets of components that are tightly connected to each other,
but only loosely connected with the rest of the system (Baldwin and Clark, 2000).
Modular structure is revealed by tracking dependencies of the form “if Component A
changes, Component B may have to change as well.” By definition, a change within one
module can be made without triggering changes in the others.

In a given system at a given time, the boundaries of modules may or may not correspond
to location and extent of bottlenecks. However, modular structure is largely under the
control of system designers and thus module boundaries can be drawn or redrawn to suit
the designers’ purposes. By definition, each component in a module is co-specialized to
every other, thus in effect components within modules are subject to very strong
matching requirements. The operative question for designers is, where should these

Page 6 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

strong matching requirements be placed? As we shall see in Sections 4 and 5, the firm’s
strategy toward bottlenecks provides a partial answer to this question.

Box 1 Subtypes of bottlenecks

Below I describe the three bottleneck subtypes in more detail and give examples.

Sine qua non bottlenecks. A complex technical system can generally be broken down
into parts, each of which is necessary to the performance of the whole. For example, a
digital computer at a minimum must have components that perform arithmetic
computations; control the flow of instructions; store instructions and computations in
memory; input instructions and data; and report results. Without all of these components,
the computer cannot perform its basic function, thus each is sine qua non for the system.
Similarly, an airplane at a minimum must have components to provide lift (the wings);
thrust (the engines); a central framework (the fuselage); lateral and vertical stability
(elevators, ailerons, rudder); a steering mechanism (same); and the ability to land (flaps,
wheels). If one of these parts is missing, the flying machine cannot function in a useful
way. Again each is sine qua non for the system.

Arthur (2009) describes the invention of novel technologies as a process of linking

solutions “until each problem and subproblem resolves itself into one that can be
physically dealth with—until the chain is fully in place” (2009: 110). The unsolved
problems Arthur refers to are the sine qua non bottlenecks standing in the way of the
creation of a new technical device. When the last or most difficult sub-problem is solved,
this is generally recognized as a breakthrough, and the event becomes part of the lore of
the technology. For example, the Wright brothers solved the critical sub-problem of
lateral control of a flying machine, and are credited with the invention of the first
successful airplane. Arthur describes several other breakthroughs, including Ernst
Lawrence’s idea to use timed voltages to accelerate particles in a circular tube (the
cyclotron) and Gary Starkweather’s approach to the problems of modulation and pointing
in laser printing (2009: 114–119).

Flow bottlenecks. Many complex systems involve flows. The flows may be water through
an irrigation system, trains through a railroad, goods through a factory, electrons through
a computer, instructions through a software program, messages through a
communication network, transactions through a payments system, customers through a
store, patients through a health-care system, laws through Congress. All systems
involving flow are subject to technical bottlenecks in the form of constraints caused by
lack of capacity in some of the conduits of the flow.

For example, consider a rail system from point A to point E. Assume the system consists
of five segments <a, b, c, d, e>. Segment c is a river for which no bridge exists: goods
must be unloaded, taken on barges, and reloaded on the other side. Shippers are only
interested in going from A to E, thus the value of any subset of segments is zero. (Each

Page 7 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

segment is sine qua non, but since solutions exist for each segment, there is no sine qua
non bottleneck.) Under these assumptions, the overall value of the system is proportional
to the number of trains passing from A to E times shippers’ willingness to pay for such
transportation.

For purposes of illustration, let us assume that shippers value speed: the less time in
transit, the higher their willingness to pay. Obviously, any technology that shortens the
transit time for any segment will increase the value of the system. But on this view, the
river does not stand out as a bottleneck. An hour saved in segment a is as good as an hour
saved in segment c. All opportunities to reduce time-in-transit should be given equal
consideration.

In contrast, the number of trains going from A to E in any given time period depends on
the capacity of the system. Again for purposes of illustration, suppose only one train can
occupy each segment at a given time. Now assume it takes one day for a train to cross
segments a, b, d, and e, and one week to cross the river. The capacity of segments a, b, d,
and e is 365 trains per year, but the capacity of segment c is only 52. Thus if two trains
are dispatched from point A on successive days, the second will wait six extra days and
take 13 days to cross the river. A train dispatched on day 3 will take 19 days to cross.

The capacity of the slowest segment constrains the capacity of the system as a whole.
Mathematically, the capacity of the system equals the minimum capacity of its segments.a
From this perspective, the river segment is a bottleneck. It is necessary to the functioning
of the whole: a train cannot get from A to E without crossing it. Yet it impedes the
performance of the whole. Futhermore, improving the capacity of any other segment has
no effect on the capacity of the system as a whole. Only the bottleneck matters.

All systems involving flow are subject to capacity constraints. And all capacity constraints
take the form of a “minimum” function in the general function determining system value.

Matching bottlenecks. Constraints on “matching” or “fit” are the third source of

technical bottlenecks in man-made systems. For example, the power of the engine in an
automobile must be matched by the power of its brakes. The strength of materials in a jet
engine must match the force of the jets (Arthur, 2009).

Rosenberg (1969) describes a matching bottleneck caused by the introduction of high-

speed steel alloys in the late nineteenth century:

[I]t was impossible to take advantage of higher cutting speeds with machine tools
designed for the older carbon steel cutting tools because they could not withstand
the stresses and strains … As a result, the availability of high-speed steel for the
cutting tool quickly generated a complete redesign in machine tool components—
the structural, transmission, and control elements (1969: 7–8).

Page 8 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

In other words, once high-speed steel was introduced the three other components of a
machine tool—the frame, the transmission, and the controls—became technical
bottlenecks. Until these components were redesigned to match the greater force of the
high-speed cutting tool, machine tools could not take advantage of the value-creating
possibilities inherent in the new steel alloy.b

In the case of machine tools, the inferior capabilities of weaker components made it
impossible to take advantage of the capabilities of stronger components, and this
mismatch constrained the performance of the system as a whole. In other cases involving
matching, there is no question of “weak” or “strong,” but components must be co-
specialized if they are to deliver high levels of functionality. For example, in the design of
a tablet computer or mobile phone handset, the technical challenge is to get components
to fit within a given form factor. In the case of bicycle drivetrains (discussed below), the
challenge is to have the components interact with high precision and minimal friction.
These are different forms of matching bottlenecks.

(a) This statement holds when there is only one path through the system. More
complicated functions are needed when there are multiple paths, but the minimum
function is prominent in all of them. The minimum function may also constrain
performance in systems that do not involve flow. For example, a chain is only as strong as
its weakest link; a security system is only as good as its weakest portal.

(b) There is no generally accepted term for a component like high-speed steel that
generates a new set of technical possibilities but only if the related technical bottlenecks
are solved. Yet such components are common in the history of technology. We might call
them generative components because they generate both opportunities and bottlenecks.
In effect, generative components open up new design spaces, which can then be fruitfully
explored. General purpose technologies (GPTs) are a fortiori generative components, but
smaller innovations may be generative as well.

Page 9 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

3.3 Strategic Bottlenecks and a Firm’s Span of Control

Strategic bottlenecks are points of value capture and thus a source of rent in a technical
system. A strategic bottleneck needs two things: (1) a unique solution to an underlying
technical bottleneck; plus (2) control over access to the solution. For example, if a river is
a technical bottleneck in a railroad system, a firm seeking to capture a strategic
bottleneck must first build a bridge (the solution) and then prevent others from using the
bridge unless they pay a toll (control).

In economics, the ability to exclude others from using a given resource is the classic
definition of a property right (Alchian, 1965; Demsetz, 1967, 1988; Alchian and Demsetz,
1973; Grossman and Hart, 1986; Libecap, 1989, 1993; Hart and Moore, 1990; Hart, 1995;
Mahoney, 2004; Kim and Mahoney, 2005).6 Property rights in turn can be de facto based
on power (my army controls the bridge; the chemical formula is a secret) or de jure based
on the law (I own the bridge and police will arrest any trespassers; the chemical formula
is patented). Property rights over the solutions to technical bottlenecks, whether de facto
or de jure form the basis of strategic bottlenecks. Property rights establish boundaries
hence they are part of the contract structure of firms.

Teece (1986) called the state of property rights, particularly intellectual property (IP)
rights, the “appropriability regime” pertaining to a resource, and noted that the regime
might be weak or strong. In strong appropriability regimes, it is easy to exclude others
from using a particular resource. In weak appropriability regimes, it is hard.

Under the resource-based view of the firm, a strategic bottleneck is a VRIN asset—
valuable, rare, inimitable, and non-substitutable (Wernerfelt, 1984; Barney, 1991). VRIN
status follows directly from the definition of a strategic bottleneck. First, it is valuable
because it is a solution to a problem affecting the performance of the system, and the
next best solution is far inferior. As the far-superior solution, the strategic bottleneck is
more than simpy rare, it is unique. It must be inimitable, for if it can be copied, the copies
will also solve the problem, and the bottleneck will cease to exist. By the same token, it
must be non-substitutable, for if substitutes exist, they will also constitute solutions, and
the bottleneck again will disappear.7 The technical, legal, and organizational setting
surrounding the bottleneck will determine whether it is easy or difficult to maintain the
VRIN status of the technical solution and thus protect the strategic bottleneck.8

Property rights—the ability to determine who has access to superior solutions to technical
bottlenecks—are thus critical to protecting a strategic bottleneck and claiming the
associated rents. I define the span of control of a given firm to be the totality of its
property rights over the components of a technical architecture. A firm can exercise
control through a combination of physical control, secrecy, contracts, patents, and
copyrights. The components it controls by these means are deemed to be within its span
of control.

Page 10 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

In general, a firm’s span of control coincides with its organizational boundaries. As long
as it obeys the law, a firm can determine what goes on within its span of control. Its
senior managers and board of directors can set policies, establish procedures, and
delegate authority as they see fit (Freeland and Zuckerman, 2014). Both the technical
architecture and the contract structure of the firm lie within their purview. In contrast, a
firm may influence, but does not control what happens outside its span of control.

Some firms have a small or narrow span of control. Such firms perform few tasks in-
house, have few organization-specific skills or secrets, and little or no formal IP rights.
Others have a large or wide span of control. They perform many tasks, have many
organization-specific skills and secrets, and have large IP portfolios.

Spans of control thus constitute a third dimension of architecture after bottlenecks and
modules. While bottlenecks and modules are aspects of the technical architecture, spans
of control, which reflect organizational boundaries and property rights, are key parts of
the industry architecture. Thus a given component might be an intensive technical
bottleneck located within module A in the span of control of firm X. Firm X might then
split module A into several subsidiary modules in order to accelerate the search for
solutions to the technical bottleneck.9 Alternatively, a group of components might form an
extensive technical bottleneck, located in modules B, C, D, and E, in the spans of control
of firms W, X, and Y. Firm Z might unite modules B, C, D, E within its own span of control,
dissolve the modular boundaries, solve the technical bottleneck, and control a strategic
bottleneck as a result.

To create and hold a strategic bottleneck, a firm must understand the system’s
architecture at several levels—legal and organizational as well as technical. First, what is
the solution to the technical bottleneck? Second, what property rights to solutions does
the legal system grant, and how can these be secured? Third, how should an organization
be formed to deliver the solution, protect it, and obtain payment for it?

The capacity to see bottlenecks, modules, and spans of control in a large technical system
and the ability to shift them by changing the underlying modular structure and/or
organization boundaries and property rights are an important subset of dynamic
capabilities. To distinguish them from other dynamic capabilities, for example, human
resource management or mergers and acquisitions (M&A) capabilities, we may call them
architectural capabilities. Combining architectural capabilities with good strategy and
resources can lead to competitive advantage and associated rents in the short run (Teece,
2014). However, bottlenecks, modules, and spans of control are not cast in stone:
technical bottlenecks can be solved, strategic bottlenecks can be seized, and modules and
spans of control can be redrawn. When all things are shifting, competitive advantage can
only be sustained by continually investing to renew the firm’s architectural knowledge
and hone its dynamic architectural capabilities.

Page 11 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

In Sections 4 and 5, I describe and give examples of ways in which a firm can change the
modular structure of a system in conjunction with its property rights and organizational
boundaries (its span of control) to gain competitive advantage. I first consider how to
solve technical bottlenecks and then how to protect strategic bottlenecks.

4 Solving Technical Bottlenecks

A technical system can be envisioned as a network of components connected by
dependencies or links (Steward, 1981; Eppinger, 1991; Baldwin and Clark, 2000). The
dependencies and links involve transfers of material, energy, or information plus
payments for goods and services (Pahl and Beitz, 1995; Baldwin, 2008).

Dependencies among components can be managed in two different ways. The first is for
the designers to communicate in real time and work out by mutual adjustment how to
handle each one. The second is for an architect to specify ex ante what dependencies will
exist. This specification transforms a negotiated set of dependencies into a rule—a design
rule—that is binding on all designers. In the presence of a mutually binding rule, the
designers do not have to communicate with one another (Parnas, 1972; Parnas et al.,
1985; Baldwin and Clark, 2000). In effect negotiated dependencies require lateral
coordination, while design rules provide hierarchical coordination.

Lateral dependencies are dense within modules and absent (or at least sparse) across
modules. Design rules place more restrictions on the final architecture of the system than
dependencies managed through real-time communication and mutual adjustment. With a
rule, there is less room for give-and-take, and thus some new ideas may not be
accommodated. Notwithstanding this fact, putting design rules in place does not
necessarily hurt system performance. Architectural knowledge can be used to determine
which design rules offer little or no harm to the overall system.

Corresponding to these two methods of coordination, a firm can manage technical

bottlenecks in two different ways. In some cases, it will be advantageous to use design
rules to split a module in order to isolate bottleneck components and maintain a narrow
span of control in the underlying technical architecture. This move reduces the need for
lateral coordination. In other cases, it will be advantageous to reduce the separation
imposed by modular boundaries in order to integrate several components of a bottleneck
within a single module. This move increases the need for lateral coordination and the firm
will need to maintain a correspondingly wide span of control in the architecture.

Different types of firms will obtain different benefits from pursuing each approach (Teece,
1996). A strategy of combining modules might be feasible and attractive for a large firm
but out of reach for a smaller one. Conversely, small firms, with no organizational
hierarchy or vested interests, can often move swiftly to isolate technical bottlenecks and

Page 12 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

thus claim strategic bottlenecks in promising new markets. Figuring out which approach
is better requires deep knowledge and fine judgment—the bedrock of architectural
capabilities.

In Subsections 4.1, 4.2, and 4.3, I discuss each of these strategies and give supporting
examples.

4.1 Isolating Bottlenecks within a Small Span of Control

Page 13 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

When a technical bottleneck involves only a few components it is often possible to change
the dependencies via design rules to isolate the bottleneck components in a separate
module. Isolating the bottleneck has a number of benefits, both technological and
strategic. From a technological perspective, the bottleneck component can be redesigned
to achieve higher performance without triggering negative interactions with components
outside the bottleneck. As a result, the cost and time needed to instigate new rounds of
technical improvement will decrease (Baldwin and Clark, 2000). And because it is a
bottleneck, the performance of the whole system will improve as the performance of this
part improves. The firm that owns the superior solution to the bottleneck can then charge
users based on the incremental boost to system performance.

Isolating the bottleneck has strategic benefits, especially to (1) small firms with limited
resources; and (2) second movers trying to overtake an incumbent with a more integrated
architecture. Transaction costs are low at module boundaries (Baldwin, 2008) and non-
bottleneck components are (by definition) substitutable. Thus the firm that owns the
bottleneck solution can outsource design and production of non-bottleneck components.
There will (again by definition) be several adequate alternatives for each non-bottleneck
component hence the focal firm should be able to source these components on
competitive terms. The firm will operate with a “small span of control.”

Two financial advantages arise from a small span of control. First, because it needs
proportionately less capital to scale its operations, the firm with a small span of control
will have a “sustainable growth” advantage (Higgins, 1977; Donaldson, 1984; Baldwin
and Clark, 2006). This means that using only internally generated funds, the firm can
grow more rapidly than rivals with larger spans of control. Second, because its rate of
return on capital is higher than that of its rivals, the firm can issue new equity on better
terms, with less dilution to existing investors. This further increases the growth rates the
firm can achieve without harm to its owners. In dynamic markets with network effects,
which are subject to “tipping,”10 the ability to grow fast and profitably is a key
determinant of success (Katz and Shapiro, 1985, 1992, 1994; Farrell and Saloner, 1986;
Farrell and Shapiro, 1988).

In summary, the isolating strategy depends on achieving a superior, unique solution and
exclusive control of an intensive technical bottleneck. The controlled part constitutes the
firm’s “span of control” in the overall system hence this is a “small span of control”
strategy (Baldwin and Clark, 2006). Architectural knowledge and capabilities are
required to locate and solve the technical bottleneck, to modularize the technical
architecture, and to set up the contract structure needed to obtain non-bottleneck
components (Fine and Whitney, 1996; Mayer and Argyres, 2004; Argyres and Mayer, 2007;
Blair et al., 2011).

In the 1990s, Sun Microsystems and Dell Computer both pursued small span of control
strategies with success. Highlights of these two cases are given below. Further details
may be found in Baldwin and Clark (1997a, 2006).

Page 14 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

4.1.1 Sun Microsystems

Sun’s architectural capabilities were based on new methods measuring flows of
instructions in computer hardware and software systems.11 Using these methods, Sun’s
engineers identified memory management as a critical constraint limiting the calculating
capacity of engineering workstations. In effect, the principal bottleneck in the technical
architecture of engineering workstations was a flow bottleneck from memory to the CPU.
In the second generation of Sun’s products, the engineers addressed this constraint with
a specially designed and patented chip and associated code in the operating system (OS)
(Baldwin and Clark, 1997a).

This technical strategy in turn allowed Sun to adopt a small span of control contract
structure. Most of the components in Sun’s systems were “cheap, off-the-shelf” parts. Sun
also used Unix, a low-cost OS, and Ethernet, a freely available set of network standards,
instead of developing its own proprietary OS or network software.

Sun’s chief rival Apollo Computer adopted a very different technical strategy. Its
engineers combined specific hardware designs with a proprietary OS plus firm-specific
network protocols and software. An Apollo system operated with Apollo components or
not at all. All parts were co-specialized and produced by Apollo alone. Apollo had to
design and manufacture all parts, thus it had a relatively large span of control in the
technical architecture.

Events proved Sun’s strategy to be superior. By isolating the memory management

bottleneck, Sun was able to build machines that were as fast as Apollo’s but cost less.
And because it manufactured fewer components in-house, Sun inventories turned over
more quickly, reducing its need for capital. Sun used the financial advantages associated
with its small span of control to drive down product prices and to grow very fast. Apollo
was unable to keep pace, and avoided bankruptcy only by being acquired by HP in 1989.

4.1.2 Dell Computer

Dell’s architectural capabilities were rooted in the superior management of working
capital. In the manufacturing of personal computers, Dell’s managers saw that order
taking and assembly were a technical bottleneck, and sought to maximize throughput for
these stages of the production process. Dell became formidably efficient in this narrow
set of activities. It could produce similar goods at lower cost, in less time, with less
working capital than its rivals. (Much of the time, Dell had negative working capital, thus
the more it grew, the more cash came into the business.)

Like Sun, Dell used its financial advantage to attack its rivals by driving prices down. As
it grew, moreover, Dell exercised more power over its non-bottleneck suppliers. It
demanded and received generous trade credit, which gave it more cash and allowed the
company to grow even faster. The other first-generation PC manufacturers could not
match Dell’s prices or growth and were forced to exit.

Page 15 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Interestingly, however, Dell never converted its original architectural capabilities into
dynamic capabilities to manage architecture more broadly. Instead it defined innovation
as a quest for higher return on invested capital (ROIC). When demand for PCs leveled off
in 2009, and Chinese manufacturers, like Lenovo, became low-cost suppliers, Dell
struggled to find new products and new distribution channels, but to no avail. In 2013,
Dell became a private company through a leveraged buyout.

4.2 Integrating Bottlenecks within a Large Span of Control

Sun and Dell each succeeded by isolating a strategic bottleneck within a small span of
control. They were able to employ this strategy because the underlying technical
bottlenecks were intensive, that is, relatively narrow in scope. In contrast, some
bottlenecks are extensive, involving a wide range of components all of which must be
redesigned to solve the bottleneck (Iansiti and Clark, 1993). Addressing extensive
bottlenecks requires coordinated change in all the components in the bottleneck. One
way to achieve such coordination is to break down modular boundaries and integrate the
bottleneck components within a single large module.

In comparison to the isolating strategy described above, the integrating strategy requires
more resources. Improving the joint performance of many interdependent components is
harder than working on just a few, but the rewards in terms of better performance can be
commensurately high. Also because the technical architecture is not modular, there will
be few external points of attachment to the new system (Fixson and Park, 2008). Rivals
must then recreate the whole system in order to compete.

Thus as long as the performance of the integrated system is superior to that of

comparable modular systems, the integrated firm’s strategic bottleneck will be strongly
protected and virtually unassailable. However, when performance is equal in the eyes of
customers, the modular strategy, which is more adaptable and less resource-intensive,
will win, as was the case for Sun and Dell.

Firms pursuing an integrative strategy must design many components and perform many
tasks within their boundaries. Necessarily, such firms will have large spans of control.
When the bottleneck in question is related to the flow of materials in production, such
firms will become vertically integrated. Indeed, as documented by Alfred Chandler, the
“modern corporations” that arose in the 1880s succeeded by solving extensive flow
bottlenecks via integrative, large-span of control strategies. More recently, the Japanese
bike drivetrain manufacturer Shimano addressed a matching bottleneck related to gear
shifting also by using an integrative, large-span of control strategy. Highlights of these
examples are discussed in Sections 4.2.1 and 4.2.2. Further details can be found in
Chandler Jr. (1977, 1986) and Fixson and Park (2008).

4.2.1 Chandler’s “Modern Corporations”

Page 16 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

The firms Alfred Chandler labeled “modern corporations” grew up in response to

opportunities made possible by high speed, mechanized production systems, rapid
reliable transportation via railroads, and fast communication via telegraph (Chandler Jr.,
1986).12 Given these new technologies, the previous systems of production and
distribution were quickly shown to be shot through with bottlenecks. Chandler (1962)
quotes Charles R. Flint, an organizer of US Rubber Company, on the advantages of
vertical integration in this context:

[T]he specialization of manufacture on a large scale … permits the fullest

utilization of special machinery and processes, thus decreasing costs; the standard
of quality is fixed and raised; those plants which are best equipped and most
advantageously situated are run continuously … there is no multiplication of the
means of distribution … terms and conditions of sales become more uniform, and
credits through comparison are more safely granted; the aggregate of stocks [i.e.,
inventories] is greatly reduced … greater skill in management accrues to the
benefit of the whole, instead of the part; and large advantages are realized
through comparative accounting and comparative administration. (1962: 34–35)

In other words, with the new machine technologies, the capacity of the manufacturing
stages of production increased dramatically. But those advantages could only be realized
if the plants were run continuously. Hence flow through the system had to be both high
(close to capacity) and consistent.

The old methods of purchasing raw materials and distributing finished goods were
capacity constrained, thus immediately became technical bottlenecks in the general
system of mass production. Addressing these constraints uncovered subsidiary technical
bottlenecks in inventory, transactions processing, and the management of trade credit.
Managers, a new professional group, were specifically trained to identify and address
such bottlenecks and the entire system became more efficient as the technical
bottlenecks were systematically removed.

In industries where the cost advantages of mechanized plants depended on maintaining a

constant, high level of throughput, the techniques of flow management created such large
cost savings that only vertically integrated enterprises remained competitive. In such
industries, corporations with large spans of control administered by professional
managers became the dominant form of organization. These companies are a prime
example of the potential advantages of pursuing a large span of control strategy that
integrates many components in a non-modular fashion within a single corporation.

Indeed any technical system whose performance rests on balanced flows will generate
shifting bottlenecks that need constant oversight and coordination in real time. Today,
however, the first line of monitoring and adjustment is usually done by computers such as
numerical control systems, scheduling systems, and automated trading systems. Flows
through computer mediated systems have increased accordingly. Transactions are also

Page 17 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

cheaper today and thus many high-flow, automated systems cross firm boundaries. Thus,
in many industries, intermediate markets and firms with “vertically permeable”
boundaries have replaced purely vertically integrated firms (Jacobides, 2005; Jacobides
and Billinger, 2006; Luo et al., 2012; Kapoor, 2013; Atalay et al., 2014).

4.2.2 Shimano
The Japanese bike drivetrain manufacturer, Shimano, illustrates the solution of an
extensive matching bottleneck via integration of component designs within a large span
of control. A standard bicycle drivetrain is made up of six components. In the early 1980s
the standard drivetrain architecture was highly modular, and components made by
different manufacturers could be mixed and matched at will. However, the standard
drivetrain had two deficiencies with respect to shifting. First, the rider had to carefully
adjust the chain after shifting (to center the chain on the sprocket). This took time (a
problem for racing) and generally required taking a hand off the handlebars (a problem
for mountain biking). Second, the rider had to be pedaling during the shift.

Shimano addressed these functional deficiencies by redesigning five of the six

components to be co-specialized to very tight tolerances. No longer would any chain work
with any derailleur, freewheel, or hub: each component had to be Shimano-made. (Today,
the Shimano Hyperglide™ cassette body has become the de facto industry standard and
several manufacturers make compatible parts. However, it is claimed that suboptimal
shifting results from mixing chains and hubs from different companies.13)

The functional superiority of the new technical architecture turned the drivetrain from a
technical bottleneck into a strategic bottleneck. Over the course of six years (1984–1990)
following the introduction of the new drivetrain, Shimano’s market share increased from
approximately 20 percent to 50 percent in road bicycles and almost 80 percent in
mountain bicycles. Most of Shimano’s competitors were forced to exit. Shimano continues
to dominate the industry today.

4.3 Do All Extensive Bottlenecks Require Integration?

Given these examples, it is natural to ask whether addressing extensive technical

bottlenecks requires integration of components in a single module and a correspondingly
large span of control. The answer is “no.” As indicated, coordination in the design of
components can be achieved in two ways: through integration, which facilitates lateral
communication; or through modularization, which achieves hierarchical coordination via
design rules. Intensive bottlenecks almost always benefit from modularization, but
extensive bottlenecks may go either way depending on circumstances.

The transition from traditional freight to containerized shipping is an example of an

extensive bottleneck that was addressed via modularization and hierarchical
coordination. To handle containers efficiently, new docking facilities had to be built at
every major port. But the ports themselves operated independently and thus each one

Page 18 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

could be handled as a separate project (a module). Ships and railcars also needed to be
redesigned but they too could be produced independently. All that was needed for the
ships, railcars, and ports to work together was a common standard for containers.

In effect, the common standard for containers served as parsimonious design rule,
providing hierarchical coordination throughout the system. Once the industry arrived at a
common definition of the standard container, work on other parts of the system could
proceed in a modular fashion. Furthermore, because it was impossible to deny anyone
access to the standard container design, it was hard to convert the separate technical
bottlenecks in ports, ships, and railroads into a system-wide strategic bottleneck. As a
result, profits from containerization generally flowed to the module innovators and not to
the original system innovators (see Levinson, 2006, for additional details).

In general, the less that is known about a given technical architecture and the location of
its technical bottlenecks, the more advantageous is lateral communication and
coordination (Monteverde and Teece, 1982; Monteverde, 1995). And, because transaction
costs are high within modules, a large span of control is often the most cost-effective way
to organize the resulting non-routine problem-solving process (Nickerson and Zenger,
2004; Baldwin, 2008; Colfer and Baldwin, 2016). However, lateral coordination is time-
consuming and may not converge on a solution (Steward, 1981; Eppinger, 1991; Baldwin
and Clark, 2000). In contrast, a modular architecture restricts alternatives but often
saves time by limiting the degree to which changes in the design of one component can
propagate through the system (Ethiraj and Levinthal, 2004).

It follows that integrating all technical bottlenecks within a single module is a good way
to develop a truly novel technical system. However, if the system is large, it will take a
very long time to complete. For example, the ATLAS particle detector at CERN, a very
large and novel project, was designed in an integrated fashion but took eighteen years to
go from concept to working system (Tuertscher et al., 2014).

5 Protecting Strategic Bottlenecks

The value of a strategic bottleneck is proportional to the difference it makes in the value
of the system relative to the next best solution. If the difference is large, then the rent
stream to the bottleneck owner will be commensurately high. For instance, in the railway
example above, suppose building a bridge over the river cuts transit time on this segment
from one week to one day. The capacity of the system would then increase from 52 to 365
trains per year. The owner of the bridge should be able to capture some fraction of the
value thus created.14

The firm that owns a strategic bottleneck must protect it to ensure that its value remains
high. The main technical threat is the arrival of almost-as-good or better solutions to the
underlying technical bottleneck. The main legal threat is loss of control of the solution. IP

Page 19 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

rights over the design of the solution are one way to defend a strategic bottleneck against
both threats. Indeed, in strong appropriability regimes, a good patent or copyright may
be all that is needed (Teece, 1986). But intellectual property rights, especially patents,
always exist in the shadow of litigation and thus are uncertain (Hagiu and Yoffie, 2013).
For example, of all patents that are litigated, approximately half are invalidated. (For this
reason, Lemley and Shapiro (2005) call patents “probabilistic property rights.”) What is
certain is that the higher the value of the strategic bottleneck, the higher the legal costs
of defending it and the greater the legal risks.

The modular structure of the system can also be used to protect a strategic bottleneck. To
understand how modularity works in this context, we must first distinguish between
modularity in production and modularity in use. Modularity in production refers to the
way tasks of production are divided among separate groups. If the tasks and knowledge
of different groups do not overlap and the groups do not communicate, the system is
modular in production. (The groups may be business units in a single firm or different
firms in a supply network.) In contrast, modularity in use refers to way the system is
perceived by users. A system that can be configured by mixing and matching modules and
adding or replacing modules is modular in use. A single, indivisible item is not modular in
use (Baldwin and Clark, 1997b; Sako and Murray, 1999; Baldwin and Clark, 2000;
Baldwin and Henkel, 2015).

In general, modularity in production protects strategic bottlenecks while modularity in

use exposes them to competition. Subsections 5.1, 5.2, and 5.3 explain this apparent
conundrum.

5.1 Modularity in Production Protects the Bottleneck

One of the features of modules is that they “hide information” from other modules
(Parnas, 1972; Parnas et al., 1985; Baldwin and Clark, 2000). People working on module A
don’t have to know what’s going on inside module B. Thus through modularity, it is
possible to divide up knowledge about a bottleneck solution so that no single employee or
supplier can replicate the whole. Modularity in production can thus protect a strategic
bottleneck from being replicated by suppliers and employees of the focal firm (Baldwin
and Henkel, 2015).

Liebeskind (1997), Henkel et al. (2013), and Baldwin and Henkel (2015) show how firms
have used modularity in production to protect organizational secrets. Highlights are
given below: the original papers provide further examples and details.

5.1.1 Michelin Radial Tires

In the 1960s, Michelin used modularity in production to protect its strategic bottleneck in
radial tires. According to Liebeskind (1997):

Page 20 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

During the 1960s, Michelin had a monopoly on knowledge relating to the

production of high quality steel-belted radial tire manufacturing. In order to
preserve this monopoly, manufacturing was divided into two separate processes:
steel belt manufacturing, and tire production. Employees were not rotated
between these manufacturing processes in a deliberate effort to restrict the
number of employees that had knowledge about both processes. As a result, only a
handful of very senior managers within Michelin were knowledgeable about the
entire manufacturing process (1997: 645).

5.1.2 Production and R&D in Countries with Weak IP Protection

In interviews conducted in Brazil and Mexico, Sherwood (1990) reports that, in many
instances, only the founder of a company would know the details of all steps in the
production process, though key employees would know parts. In a more recent example,
Schotter and Teagarden (2014) quote a former general manager of a Western-owned
factory in China:

Processes were broken down into those elements critical to the business so that
no one single individual had access to all of them. As general manager, I was the
only one who knew the entire manufacturing process. Extrusion dies, for example,
were critical to the manufacturing process. These were kept overnight in a
separate vault that could only be opened by one person, me, using a lock and key.
Other staff members were never allowed to put the dies into the extrusion
machine. (2014: 45)

5.1.3 Costs of Modularity in Production

Using modularity in production to protect a strategic bottleneck can be costly, however.
Modularizing production replaces lateral coordination with hierarchical coordination via
design rules. As indicated above, this can be fatal for the production of novel systems,
where unknown dependencies can result in conflicts that cause the whole system to
crash. Also in systems where throughput is a key determinant of cost, excessive
modularization can slow down the flow and increase inventories needed for adequate
buffering (Zhou and Wan, 2012). Finally, some technologies are more amenable to
modularization than others: for example, electronic chips are highly modular, whereas
automobile parts must often be co-designed in order to work together in the same vehicle
(Whitney, 2004; Luo et al., 2012).

5.2 Modularity in Use Exposes the Bottleneck

Page 21 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

In contrast to modularity in production, modularity in use exposes a strategic bottleneck

to a greater threat of imitation or substitution by third parties (Rivkin, 2000; Pil and
Cohen, 2006; Ethiraj et al., 2008). To make a system modular in use, the system sponsor
must offer modules as separate products and provide ways for new modules to be
attached to the system. Rivals then do not have to replicate the system as a whole, but
can concentrate on a single module. In contrast, integral, i.e., non-modular, products have
no separate components and no external points of attachment. Rivals must then recreate
the whole system or stay out of the game.

The case of the IBM PC illustrates the risks modularity in use poses to a strategic
bottleneck. The case of the Apple Macintosh illustrates the offsetting risks of restricting
users’ options by making only non-modular systems.

5.2.1 The IBM PC and the Risks of Modularity in Use

The original IBM PC had a highly modular technical architecture. Because IBM was a late
entrant to the PC market and the PC division was resource constrained, PC’s managers
adopted a small span of control strategy. They focused on controlling a small number of
components, which they hoped would serve as strategic bottlenecks. They outsourced all
other components of the PC system, including the central processor (from Intel) and the
OS (from Microsoft) (Ferguson and Morris, 1993).

A key point of control for IBM was the so-called BIOS, the Basic Input Output System. A
BIOS is very much like a bridge connecting peripheral devices to the central processor.
Without using the BIOS, software written for the PC could not access peripheral devices,
such as disk drives, printers, keyboards, and monitors. The IBM BIOS was protected by
copyright, and also etched in the circuits of a read-only memory (ROM) chip. However,
the chip itself was relatively small (35KB of memory).

Shortly after the IBM PC was introduced, Compaq and Phoenix Technologies both legally
copied the PC BIOS using a practice known as “clean room reverse engineering.”15 With
the reverse-engineered BIOS in hand, it was relatively easy for other manufacturers to
create clones of the IBM PC. (Clones were substitutes that were fully compatible with
IBM PCs, hence could run software and use peripheral devices developed for IBM
machines.) By 1986, Compaq and numerous Taiwanese clone makers were flooding the
market with IBM-compatible machines.

In this way, IBM lost control of a major strategic bottleneck in the PC’s architecture. The
BIOS was still an essential component, but, using reverse-engineered chips, Compaq and
the clone manufacturers could make machines that worked equally well with the vast
number of peripherals and software applications designed for IBM PCs. The BIOS was no
longer rare, inimitable, and non-substitutable, and thus it ceased to be very valuable.
Furthermore, as the clone makers grew, IBM lost the advantage of scale, and ceased
being the low-cost producer of PCs. The PC division struggled to remain profitable until it
was sold to Lenovo in 2004.

Page 22 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

5.2.2 Apple Macintosh and the Risks of Integrality

Unlike IBM, Apple opted to make the Macintosh BIOS an integral part of its OS.
According to one observer:

[T]he Mac BIOS is very large and complex and is essentially part of the OS, unlike
the much simpler and more easily duplicated BIOS found on PCs. The greater
complexity and integration has allowed both the Mac BIOS and OS to escape any
clean-room duplication efforts. This means that without Apple’s blessing (in the
form of licensing), no Mac clones are likely ever to exist.

(Mueller, 2003: 28)

Apple had a correspondingly large span of control in the Macintosh’s technical

architecture. Users not only had to purchase their computers from Apple, but also their
printers, disk drives, keyboards, and mice. Apple also wrote most of the application
software. Apple hardware and software were offered for sale at high markups, thus user
options for the system were largely limited to high-cost items that Apple was able to
supply.

In contrast, IBM managers adopted what is now called an “open” platform strategy
(Ferguson and Morris, 1993; Gawer and Cusumano, 2002; West, 2003; Boudreau, 2010;
Baldwin and Woodard, 2011; Eisenmann et al., 2011; Gawer, 2014). They actively
recruited hardware manufacturers to build complementary peripheral devices and
software developers to write programs that would run on the PC. The PC’s expandability
via third-party hardware and software was prominently featured in its early marketing
campaigns.

In this fashion, perhaps only half consciously, IBM tapped into a powerful set of “indirect
network effects” (Church and Gandal, 1992, 1993; Rochet and Tirole, 2003). Indirect
network effects arise when a particular institution (the platform) permits interactions
between two or more types of participants, each of which has reason to value the
presence of the other. In the case of computer platforms like the IBM PC, computer users
have reason to value large numbers of compatible applications and peripheral devices
because these increase their options and extend the useful life of their machines. At the
same time, software developers and peripheral manufacturers have reason to value large
numbers of users, because they face economies of scale in development and production.
Thus computer users will be attracted to platforms with more peripherals and software,
while peripheral manufacturers and software developers will seek out platforms with
more users.

Platforms with indirect network effects often enjoy the dynamics of increasing returns to
scale: as the numbers on one side grow, the numbers on the other side grow apace
(Eisenmann et al., 2006; Evans et al., 2006; Zhu and Iansiti, 2012). The result can be a
winner-take-all outcome where the market “tips” and most participants converge on a
single platform (Katz and Shapiro, 1994). IBM managers set loose this dynamic when

Page 23 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

they introduced the open, modular IBM PC in 1981. By 1987, IBM-compatible computers
accounted for over 60 percent of the personal computer market; this figure rose to over
90 percent in the early 1990s.16

The upshot is that, whereas the Apple Macintosh was better protected from imitators by
virtue of its non-modular BIOS and OS, the IBM PC platform offered a more attractive
value proposition to users and hardware and software developers. Users could purchase
many compatible, low-cost hardware devices and software applications. Hardware
manufacturers and software developers had the corresponding lure of many users. The
market overwhelmingly voted for the IBM PC, and this standard dominated the personal
computer industry long after IBM bowed out.

5.3 Best of Both Worlds? Apple’s iPhone and iPad Hardware Devices

It is clear from this analysis that the best way to protect a strategic bottleneck is to make
it modular in production but not modular in use. The Apple iPhone and iPad hardware
devices are designed this way. Each device contains components made by many different
manufacturers to Apple specifications, thus the designs are modular in production.17 But
each device is also sold as a single, indivisible, non-upgradable unit, thus the designs are
not modular in use.18 (In contrast, the software and music libraries are completely
modular in use.)

Apple is thus engaged in a delicate balancing act. As long as the iPhone and iPad are
perceived to be superior to more modular and configurable devices (such as most Android
phones and other tablet and laptop computers), Apple can enjoy a strongly protected
strategic bottleneck together with strong market demand. These are the “commanding
heights” of a large technical system, and Apple’s current profits are a testament to the
value of such a position.

However, as Apple learned from the Macintosh, it must give users the options they truly
value, or they will abandon the system en masse for one that is more open and flexible. A
more open system in turn might unleash the power of indirect network effects, thus
further undercutting Apple’s strategic bottleneck. Apple thus gives users a wide range of
software and content options, even as it restricts their hardware options. So far this
hybrid strategy seems to be working.

6 Acquiring Architectural Knowledge and

Capabilities
How does a firm acquire architectural knowledge and capabilities that allow it to solve
technical bottlenecks and protect strategic bottlenecks in large and evolving technical
systems? Large technical systems are generally not easy to understand (Perrow, 1984).

Page 24 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Inconveniently, they tend to change via local adaptation and grow by accretion. Changes
are often undocumented. The end result is a system no one understands, burdened with
“technical debt” that stands in the way of constructive change. Such systems are often
labeled “hairballs,” or “great balls of mud,” by the technicians who must coax them to
perform.

Architectural knowledge of a technical system begins with seeing the system as a

working whole, but also seeing that is made up of distinct components. The system as a
whole has a purpose (Arthur, 2009) and its performance against its purpose can be
measured in terms of effectiveness and efficiency. Each component in turn contributes in
some way to the performance of the system as a whole. Observers gain knowledge about
technical bottlenecks by studying each component in relation to the whole and assessing
how the properties of a given component enable or impede the performance of the
system.

In flow systems, technical bottlenecks give rise to delays and/or inventory buildup ahead
of the bottleneck and slack capacity below it. However, given even moderately
complicated flows, local adaptations and work-arounds often mask the true location of the
bottlenecks (Goldratt, 1990). Generally it is necessary to track the progress of
representative flow units through the system, recording the times spent in each
processing stage or buffer. The observer must then aggregate over a typical mixture of
paths to identify the points that place the greatest constraints on overall system capacity.
Goldratt (1990) describes this process for manufacturing, retail services, and project
management. Patterson and Hennessy (1994) describe how to track instructions flowing
through computer systems.

Matching bottlenecks, by definition, involve several components, and thus are generally
harder to identify than flow bottlenecks. Learning about the bottleneck and solving it
tend to merge into the same trial-and-error process. For example, in the case of bicycle
drivetrains, discussed above, Shimano engineers had to form the conjecture that tighter
coupling (a better match) between the derailleur, the sprockets, and the chain would
make shifting more precise and faster. They then had to test that conjecture by
redesigning and testing various configurations. Today, computer simulation and 3-D CAD
modeling can shorten the experimental generate-test cycle and thus greatly accelerate
the process of finding a better match.

Solving a technical bottleneck also requires knowledge of how modular structure

can be changed to isolate or integrate the components in the bottleneck.
One builds knowledge of modular structure by asking the question, “If this component’s
design were changed, what other components might have to change as a result?” (Parnas,
1972; Baldwin and Clark, 2000). Once direct linkages of this type have been identified,
they can be aggregated into a network graph and summarized in terms of a Design
Structure Matrix (DSM) (Steward, 1981; Eppinger, 1991; Eppinger et al. 1994; Browning,
2001).

Page 25 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Unfortunately, direct linkages by themselves do not reveal the modular structure of the
system. The DSM must be organized in particular ways to reveal hierarchy (chains of
dependencies) and modules (cyclic groups of dependencies). Baldwin et al. (2014) provide
an algorithm for discovering and displaying the modular structure of large technical
systems. In a study of over 1,000 software codebases (in 17 families), they found that the
great majority of these systems had a “core-periphery” structure, that is, there was one
dominant cyclic group (the “Core”). However, the existence and composition of the Core
and the location of other components in relation to the Core was not evident from the
systems’ documentation. Thus the modular structure of these software systems was
hidden, even to the systems’ own developers.

Finally, protecting strategic bottlenecks requires knowledge of organizational boundaries

and property rights. In developed countries, property rights to physical goods—products,
machinery and equipment, buildings and land—are usually fairly easy to ascertain and
enforce. However, to an increasing degree, control of strategic bottlenecks rests on
control of knowledge through secrecy and/or IP rights. Thus to occupy a secure position
with respect to a strategic bottleneck, a firm must know what parts of its knowledge are
critical to maintaining exclusive control of the bottleneck solution. It must also know who
might threaten the bottleneck by introducing a superior technical solution or by
contesting the firm’s claims to the underlying IP.

7 Conclusion
Large technical systems made up of many interacting components are becoming more
common every day. Many of these systems, like tablet computers, smartphones, and the
internet, are based on digital information technologies. Others, like the electrical grid,
the financial payments system, and modern factories rely on digital technologies. As
indicated, digital information technology has made it possible to create ever larger
technical systems and ever faster flows. This paper asks the question: how do firms
create and capture value in large technical systems comprising billions of components,
millions of users, and thousands of firms?

To answer this question, it was first necessary to develop ways of describing such
systems. The architecture of a technical system is an abstract description of what it does
and how it works. The technical architecture includes both a design and a task structure
and shows, from a technological perspective, how to make the system. A contract
structure is a social architecture that is overlaid on the technical architecture to
assemble resources, assign tasks, transfer goods, and collect compensation. Industry
architecture is the aggregation of contract structures within a given sector of the
economy.

Page 26 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Within a large technical system, I’ve argued that the points of both value creation and
value capture are the system’s bottlenecks. In prior work, a number of scholars, including
Rosenberg (1969), Teece (1986), Langlois and Robertson (1992), Langlois (2002, 2007),
Baldwin and Clark (2006), Jacobides et al. (2006), Ethiraj (2007), Pisano and Teece (2007),
and Henkel and Hoffmann (2014), have similarly argued that bottlenecks are key drivers
of technical change and sources of competitive advantage.

In light of this prior work, perhaps the most important contribution of this paper is to
point out that bottlenecks come in two flavors. Technical bottlenecks are important
technical problems to be solved. Once a solution is found, it may be combined with
organization boundaries and property rights and become a strategic bottleneck and a
source of rents.

A second contribution is to show how bottlenecks are related to modules, organizational

boundaries, and property rights. Bottlenecks, both technical and strategic, exist within
the modular structure of a given technical system. The system’s modular structure in turn
can be modified to better find and solve technical bottlenecks and protect strategic
bottlenecks. Finally, a firm’s span of control determines its organizational boundaries and
property rights, and thus the extent of its control over various bottlenecks in the system.

Dynamic architectural capabilities give firms the ability to understand a large technical
system coherently and change it in ways that are competitively advantageous. In essence,
these capabilities amount to the ability to find and solve technical bottlenecks and hold
and protect strategic bottlenecks. Modular structure, organizational boundaries, and
property rights are all tools the firm can use for this purpose. For example, modular
boundaries can be shifted to isolate or integrate the components of technical bottlenecks.
Modularity in production can be used to protect a strategic bottleneck: each agent will
know part of the technical recipe, but only a few trusted agents will know the whole. In
contrast, modularity in use exposes a strategic bottleneck to competition from third
parties. However, modularity in use also creates user options, which may be critical to the
firm’s value proposition.

I close the paper by suggesting two areas of opportunity for future work.

First, this paper has only considered how a single firm can manage a single technical or
strategic bottleneck. However, large technical systems have many bottlenecks at different
levels of the technical design hierarchy. Also, it is now rare for large technical systems to
be contained within the boundaries of a single firm. Instead, technically related activities
are often distributed across many firms and involve other institutions, including standard-
setting organizations and open source development communities. Further work is needed
to understand how bottlenecks can be managed when knowledge and property rights are
distributed among many independent agents.

A related theoretical problem is to develop tractable ways of modeling bottlenecks in

large technical systems. Functions and functionality, which determine a user’s inclination

Page 27 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

to buy and willingness to pay, are key to this characterization. In essence, what is needed
is a mapping from system components to system functionality to system value.

The functional decomposition of systems is an important topic in many engineering

disciplines, but it is difficult to translate the engineering models into concepts meaningful
for economics and strategy (Suh, 1990; Hatchuel and Weil, 2009). In economics, Milgrom
and Roberts (1990) developed a theory of complementarity based on the mathematical
concept of “supermodularity” (Topkis, 1998).19 Some value functions fit nicely into this
framework, but others do not. Another approach is the construction of so-called NK value
landscapes, but these are randomly generated structures, and have not to this point
proved capable of representing real systems (Kauffman, 1993; Levinthal, 1997; Rivkin,
2000; Ethiraj and Levinthal, 2004; Ethiraj et al., 2008).

What is needed, I believe, is a representation that is powerful enough to capture the

existence, location, and nature of bottlenecks in real technical systems, but simple
enough to support direct intuition and reasoning. Arthur’s (2009) theory of technologies
is one possible starting point. He conceives of technologies as hierarchies of interacting
methods, each of which solves a problem and thus contributes to the value of the whole.
This approach echoes earlier work on industry evolution and dominant designs by
Tushman and Rosenkopf (1992), Tushman and Murmann (1998), and Murmann and
Frenken (2006).

Five decades ago, the economy contained many separate and incompatible technical
systems, managed independently by different firms. Today, through the evolution of
digital technology, we are advancing towards a world comprising one large technical
system with many interoperating and evolving parts. Architectural knowledge provides a
way of understanding large technical systems and dividing them up into coherent units of
analysis. Dynamic architectural capabilities provide a comprehensive framework in which
to manage bottlenecks, modules, organizational boundaries, and property rights in
pursuit of value and competitive advantage.

Acknowledgements
I thank David Teece and Sunyoung Leih for soliciting this paper and suggesting that it
should focus on bottlenecks. I also thank my collaborator and co-author, Joachim Henkel:
my thinking on strategic bottlenecks and modularity cannot be separated from our joint
work. Likewise I thank Michael Jacobides for discussions over many years on the topic of
industry architecture and value migration. Finally I thank Nuno Gil and his students for
very insightful comments on an earlier draft. Errors and omissions are mine alone.

References

Page 28 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Abernathy, William J. and Utterback, James (1978). “Patterns of Industrial Innovation.”

Technology Review 80: 41–47.

Adner, R. and Kapoor, R. (2010). “Value Creation in Innovation Ecosystems: How the
Structure of Technological Interdependence Affects Firm Performance in New Technology
Generations.” Strategic Management Journal 31(3): 306–333.

Adner, R. and Kapoor, R. (2014). Innovation Ecosystems and the Pace of Substitution: Re-
examining Technology S-curves. Mimeo.

Alchian, A. A. (1965). “Some Economics of Property Rights.” In Economic Forces at Work

(Indianapolis, IN: Liberty Press): 294–307.

Alchian, A. A. and Demsetz, H. (1973). “The Property Rights Paradigm.” Journal of

Economic History 33(1): 16–27.

Alexander, Christopher (1964). Notes on the Synthesis of Form. Cambridge, MA: Harvard
University Press.

Argyres, N., and Mayer, K. J. (2007). “Contract Design as a Firm Capability: An

Integration of Learning and Transaction Cost Perspectives.” Academy of Management
Review 32(4): 1060–1077.

Arthur, W. Brian (2009). The Nature of Technology: What It Is and How It Evolves. New
York, NY: Free Press.

Atalay, E., Hortaçsu, A., and Syverson, C. (2014). “Vertical Integration and Input Flows.”
American Economic Review 104(4): 1120–1148.

Augier, M. and Simon, H. A. (2003). “The Architecture of Complexity: Background and

Central Idea.” In R. Garud, A. Kumaraswamy, and R. N. Langlois (Eds), Managing in the
Modular Age: Architectures, Networks, and Organizations (Hoboken, NJ: Blackwell): 38–
44.

Baker, G., Gibbons, R., and Murphy, K. J. (1999). “Informal Authority in Organizations.”
Journal of Law, Economics, and Organization 15(1): 56–73.

Baldwin, Carliss Y. (2008). “Where Do Transactions Come From? Modularity, Transactions

and the Boundaries of Firms.” Industrial and Corporate Change 17(1): 155–195.

Baldwin, Carliss Y. and Clark, Kim B. (1997a). “Sun Wars: Competition within a Modular
Cluster.” In D. B. Yoffie (Ed.), Competing in the Age of Digital Convergence (Boston, MA:
Harvard Business School Press).

Baldwin, Carliss Y. and Clark, Kim B. (1997b). “Managing in the Age of Modularity.”
Harvard Business Review Sept/Oct: 81–93.

Page 29 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Baldwin, Carliss Y. and Clark, Kim B. (2000). Design Rules, Volume 1, The Power of
Modularity. Cambridge, MA: MIT Press.

Baldwin, Carliss Y. and Clark, Kim B. (2006). “Architectural Innovation and Dynamic
Competition: The Smaller ‘Footprint’ Strategy.” Harvard Business School Working Paper
No. 07-014, available at http://www.hbs.edu/faculty/Publication%20Files/07-014.pdf,
accessed December 2017.

Baldwin, C. Y. and Henkel, J. (2015). “Modularity and Intellectual Property Protection.”

Strategic Management Journal 36(11): 1637–1655.

Baldwin, Carliss Y. and Woodard, C. Jason (2011). “The Architecture of Platforms: A

Unified View.” In Annabelle Gawer (Ed.), Platforms, Markets and Innovation (London:
Edward Elgar).

Baldwin, C. Y., MacCormack, A. D., and Rusnak, J. (2014). “Hidden Structure: Using
Network Methods to Map Product Architecture.” Research Policy 43: 1381–1397.

Barney, Jay (1991). “Firm Resources and Sustained Competitive Advantage.” Journal of
Management 17(1): 99–120.

Blaauw, G. A. and Brooks, F. P. Jr. (1997). Computer Architecture. Reading, MA: Addison-
Wesley.

Blair, M. M., O’Connor, E. O., and Kirchhoefer, G. (2011). “Outsourcing, Modularity, and
the Theory of the Firm.” Brigham Young University Law Review 2011(2): 263–314.
Available at http://digitalcommons.law.byu.edu/lawreview/vol2011/iss2/1, accessed
December 2017.

Boudreau, K. (2010). “Open Platform Strategies and Innovation: Granting Access vs.
Devolving Control.” Management Science 56(10): 1849–1872.

Browning, Tyson R. (2001). “Applying the Design Structure Matrix to System

Decomposition and Integration Problems: A Review and New Directions.” IEEE
Transactions in Engineering Management 48(3): 292–306.

Brynjolfsson, E. and McAfee, A. (2014). The Second Machine Age: Work, Progress, and
Prosperity in a Time of Brilliant Technologies. New York, NY: WW Norton.

Chandler, A. D. Jr. (1986). “The Beginnings of the Modern Industrial Corporation.”

Proceedings of the American Philosophical Society 130(4): 382–389.

Chandler, Alfred D. (1962). Strategy and Structure. Cambridge, MA: MIT Press.

Chandler, Alfred D. Jr. (1977). The Visible Hand: The Managerial Revolution in American
Business. Cambridge, MA: Harvard University Press.

Page 30 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Church, J. and Gandal, N. (1992). “Network Effects, Software Provision, and

Standardization.” Journal of Industrial Economics 40(1): 85–103.

Church, J. and Gandal, N. (1993). “Complementary Network Externalities and

Technological Adoption.” International Journal of Industrial Organization 11(2): 239–260.

Colfer, Lyra J. and Baldwin, Carliss Y. (2016). “The Mirroring Hypothesis: Theory,
Evidence and Exceptions.” Industrial and Corporate Change 25(5): 709–738.

Demsetz, Harold (1967). “Toward a Theory of Property Rights.” American Economic

Review 57: 253–257.

Demsetz, Harold (1988). “The Theory of the Firm Revisited. ” Journal of Law, Economics
and Organization 4(1): 141–161.

Donaldson, Gordon (1984). Managing Corporate Wealth. New York, NY: Praeger.

Eisenmann, Thomas, Parker, Geoffrey, and Alstyne, Marshall van (2006). “Strategies for
Two-Sided Markets.” Harvard Business Review 84(10): 92–101.

Eisenmann, Thomas, Parker, Geoffrey, and Alstyne, Marshall van (2011). “Opening
Platforms: How, When and Why?” In Annabelle Gawer (Ed.), Platforms, Markets and
Innovation (London: Edward Elgar).

Eppinger, S. D., Whitney, D. E., Smith, R. P., and Gebala, D. A. (1994). “A model-based
method for organizing tasks in product development.” Research in Engineering Design
6(1): 1–13.

Eppinger, Steven D. (1991). “Model-based Approaches to Managing Concurrent

Engineering.” Journal of Engineering Design 2: 283–290.

Ethiraj, Sendil K. (2007). “Allocation of Inventive Effort in Complex Product Systems.”

Strategic Management Journal 28(6): 563–584.

Ethiraj, Sendil and Levinthal, Daniel (2004). “Modularity and Innovation in Complex
Systems.” Management Science 50(2): 159–174.

Ethiraj, Sendil K., Levinthal, Daniel, and Roy, Rishi (2008). “The Dual Role of Modularity:
Innovation and Imitation.” Management Science 54(5): 939–955.

Evans, D. S., Hagiu, A., and Schmalensee, R. (2006). Invisible Engines: How Software
Platforms Drive Innovation and Transform Industries. Cambridge: MIT Press.

Farrell, Joseph and Saloner, Garth (1986). “Installed Base and Compatibility: Innovation,
Product Preannouncement and Predation.” American Economic Review 76(4): 940–955.

Farrell, Joseph and Shapiro, Carl (1988). “Dynamic Competition with Switching Costs.”
Rand Journal of Economics 19(1): 123–137.

Page 31 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Ferguson, Charles H. and Morris, Charles R. (1993). Computer Wars: How the West Can
Win in a Post-IBM World. New York, NY: Times Books.

Fine, Charles H. (1998). Clockspeed: Winning Industry Control in the Age of Temporary
Advantage. Reading, MA: Perseus Press.

Fine, Charles H. and Whitney, Daniel E. (1996). “Is the Make-Buy Decision a Core
Competence?” MIT Center for Technology, Policy, and Industrial Development Discussion
Paper, available at http://web.mit.edu/ctpid/www/Whitney/morepapers/
make_ab.html, accessed Aug 2007.

Fixson, Sebastian K. (2005). “Product Architecture and Assessment: A Tool to Link

Product, Process, and Supply Chain Design Decisions.” Journal of Operations
Management 23(2005): 345–369.

Fixson, Sebastian K. and Park, Jin-Kyu (2008). “The Power of Integrality: Linkages
between Product Architecture, Innovation and Industry Structure.” Research Policy 37(8):
1296–1316.

Freeland, Robert F. and Zuckerman, Ezra W. (2014) “The Problems and Promise of
Hierarchy: Voice Rights and the Firm.” Available at SSRN: https://ssrn.com/
abstract=2523245

Gawer, A. (2014). “Bridging Differing Perspectives on Technological Platforms: Toward an

Integrative Framework.” Research Policy 43(7): 1239–1249.

Gawer, Annabelle and Cusumano, Michael A. (2002). Platform Leadership: How Intel,
Microsoft and Cisco Drive Industry Innovation. Boston, MA: Harvard Business School
Press.

Gibbons, R. (2003). “Team Theory, Garbage Cans and Real Organizations: Some History
and Prospects of Economic Research on Decision‐Making in Organizations.” Industrial
and Corporate Change 12(4): 753–787.

Gibbons, R. and Henderson, R. (2012). “Relational Contracts and Organizational

Capabilities.” Organization Science 23(5): 1350–1364.

Goldratt, E. M. (1990). Theory of Constraints. Croton-on-Hudson, NY: North River.

Goldratt, E. M. (2014). The Goal: A Process of Ongoing Improvement (30th Anniversary

Edition). Croton-on-Hudson, NY: North River.

Grossman, Sanford J. and Hart, Oliver D. (1986). “The Costs and Benefits of Ownership: A
Theory of Vertical and Lateral Integration.” Journal of Political Economy 94(4): 691–719.

Gulati, R. and Puranam, P. (2009). “Renewal through Reorganization: The Value of

Inconsistencies between Formal and Informal Organization.” Organization Science 20(2):
422–440.

Page 32 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Hagiu, A. and Yoffie, D. B. (2013). “The New Patent Intermediaries: Platforms, Defensive
Aggregators, and Super-Aggregators.” Journal of Economic Perspectives 27(1): 45–65.

Hart, O. (1995). Firms, Contracts, and Financial Structure. Oxford: Oxford University
Press.

Hart, Oliver and Moore, John (1990). “Property Rights and the Nature of the Firm.”
Journal of Political Economy 98(6): 1119–1158.

Hatchuel, A. and Weil, B. (2009). “CK Design Theory: An Advanced Formulation.”

Research in Engineering Design 19(4): 181–192.

Henderson, Rebecca M. and Clark, Kim B. (1990). “Generational Innovation: The

Reconfiguration of Existing Systems and the Failure of Established Firms.” Administrative
Science Quarterly 35: 9–30.

Henkel, J. and Hoffmann, A. (2014). “Value Capture in Hierarchically Organized

Industries: The Hierarchy Strategy.” Mimeo.

Henkel, J., Baldwin, C., and Shih, W. C. (2013). “IP Modularity: Profiting from Innovation
by Aligning Product Architecture with Intellectual Property.” California Management
Review 55(4): 65–82.

Hennessy, John L. and Patterson, David A. (1990). Computer Architecture: A Quantitative

Approach. San Mateo, CA: Morgan Kaufmann.

Higgins, R. C. (1977). “How Much Growth Can A Firm Afford?” Financial Management
6(3): 7–16.

Hughes, Thomas P. (1983). Networks of Power: Electrification in Western Society, 1880–

1930 (Baltimore, MD: Johns Hopkins University Press).

Hughes, Thomas P. (1987). “The Evolution of Large Technological Systems.” In W. E.

Bijker, T. P. Hughes, and T. Pinch (Eds), The Social Construction of Technological Systems:
New Directions in the Sociology and History of Technology (Cambridge, MA: MIT Press).

Iansiti, Marco and Clark, Kim B. (1993). “Integration and Dynamic Capability: Evidence
from Product Development in Automobiles and Mainframe Computers.” Industrial and
Corporate Change 3(3): 557–605.

Iansiti, Marco and Levien, Roy (2004). The Keystone Advantage: What the New Dynamics
of Business Ecosystems Mean for Strategy, Innovation, and Sustainability. Boston, MA:
Harvard Business School Press.

Jacobides, Michael G. (2005). “Industry Change Through Vertical Dis-Integration: How

and Why Markets Emerged in Mortgage Banking.” Academy of Management Journal
48(3): 465–498.

Page 33 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Jacobides, M. G. and Billinger, S. (2006). “Designing the Boundaries of the Firm: From
‘Make, Buy, Or Ally’ to the Dynamic Benefits of Vertical Architecture.” Organization
Sscience 17(2): 249–261.

Jacobides, M. G. and Tae, C. J. (2014). “Kingpins, Bottlenecks and Value Dynamics within
a Sector.” Organization Science 15(3): 259–275.

Jacobides, Michael G., Knudsen, Thorbjorn, and Augier, Mie (2006). “Benefiting from
Innovation: Value Creation, Value Appropriation and the Role of Industry Architecture”
Research Policy 35(8): 1200–1221.

Jacobides, M., MacDuffie, J. P., and Tae, C. J. (2012, May). “When Value Sticks Around:
Why Automobile OEMs Still Rule Their Sector,” paper presented at Industry Studies
Association Conference, pp. 29–31.

Jacobides, M. G., Veloso, F., and Wolter, C. (2014). “Ripples through the Value Chain and
Positional Bottlenecks: Innovation, Profit, and Scope Evolution in a Competitive Setting.”
Mimeo.

Kapoor, R. (2013). “Persistence of Integration in the Face of Specialization: How Firms

Navigated the Winds of Disintegration and Shaped the Architecture of the Semiconductor
Industry.” Organization Science 24(4): 1195–1213.

Katz, M. L. and Shapiro, C. (1994). “Systems Competition and Network Effects.” Journal
of Economic Perspectives 8(2): 93–115.

Katz, Michael L. and Shapiro, Carl (1985). “Network Externalities, Competition and
Compatibility.” American Economic Review 75(3): 424–440.

Katz, Michael L. and Shapiro, Carl (1992). “Product Introduction with Network
Externalities.” Journal of Industrial Economics 40(1): 55–83.

Kauffman, Stuart A. (1993). The Origins of Order. New York, NY: Oxford University Press.

Kim, J. and Mahoney, J. T. (2005). “Property Rights Theory, Transaction Costs Theory, and
Agency Theory: An Organizational Economics Approach to Strategic Management.”
Managerial and Decision Economics 26(4): 223–242.

Langlois, R. N. (2007). “Organizing the Electronic Century.” University of Connecticut

Working Paper, March.

Langlois, Richard N. (2002). “Modularity in Technology and Organization.” Journal of

Economic Behavior and Organization 49(1): 19–37.

Langlois, Richard N. and Paul L. Robertson (1992). “Networks and Innovation in a

Modular System: Lessons from the Microcomputer and Stereo Component Industries.”
Research Policy 21(4): 297–313.

Page 34 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Lemley, M. A. and Shapiro, C. (2005). “Probabilistic Patents.” Journal of Economic

Perspectives 19(2): 75–98.

Levinson, M. (2006). The Box: How the Shipping Container Made the World Smaller and
the World Economy Bigger. Princeton, NJ: Princeton University Press.

Levinthal, Daniel A. (1997). “Adaptation on Rugged Landscapes.” Management Science

43: 934–950.

Libecap, G. D. (1989). “Distributional Issues in Contracting for Property Rights.” Journal

of Institutional and Theoretical Economics (JITE)/Zeitschrift für die gesamte
Staatswissenschaft 6–24.

Libecap, G. D. (1993). Contracting for Property Rights. Cambridge: Cambridge University

Press.

Liebeskind, J. P. (1997). “Keeping Organizational Secrets: Protective Institutional

Mechanisms and Their Costs.” Industrial and Corporate Change 6: 623–663.

Luo, J., Baldwin, C. Y., Whitney, D. E., and Magee, C. L. (2012). “The Architecture of
Transaction Networks: A Comparative Analysis of Hierarchy in Two Sectors.” Industrial
and Corporate Change 21(6): 1307–1335.

Mahoney, J. T. (2004). Economic Foundations of Strategy. Thousand Oaks, CA: Sage

Publications.

Mayer, Kyle J. and Argyres, Nicholas S. (2004). “Learning to Contract: Evidence from the
Personal Computer Industry.” Organization Science 15(4): 394–410.

Milgrom, Paul and Roberts, John (1990). “The Economics of Manufacturing: Technology,
Strategy and Organization.” American Economic Review 80(3): 511–528.

Monteverde, Kirk (1995). “Technical Dialog as an Incentive for Vertical Integration in the
Semiconductor Industry.” Management Science 41(10): 1624–1638.

Monteverde, K., and Teece, D. J. (1982). “Supplier Switching Costs and Vertical
Integration in the Automobile Industry.” Bell Journal of Economics 13(1): 206–213.

Mueller, Scott (2003). Upgrading and Repairing PCs, 15th ed. Indianapolis, IN: Que
Publishing.

Murmann, Johann Peter and Frenken, Koen (2006). “Toward a Systematic Framework for
Research on Dominant Designs, Technological Innovations, and Industrial Change.”
Research Policy 35: 925–952.

Nadler, D. and Tushman, M. (1997). Competing by Design: The Power of Organizational

Architecture. Oxford: Oxford University Press.

Page 35 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Nickerson, Jack A. and Zenger, Todd R. (2004). “A Knowledge-Based Theory of the Firm—
The Problem-Solving Perspective.” Organization Science 15(6): 617–632.

Orton, J. D. and Weick, K. E. (1990). “Loosely Coupled Systems: A Reconceptualization.”

Academy of Management Review 15(2): 203–223.

Pahl, G. and Beitz, W. (1995). Engineering Design: A Systematic Approach, 2nd English
ed., trans. K. Wallace. London: Springer-Verlag.

Parnas, D. L., Clements, P. C., and Weiss, D. M. (1985). “The Modular Structure of
Complex Systems,” IEEE Transactions on Software Engineering, SE-11: 259–266;
reprinted in Hoffman and Weiss (Eds), Software Fundamentals: Collected Papers of David
Parnas, Boston MA: Addison-Wesley.

Parnas, David L. (1972). “On the Criteria to Be Used in Decomposing Systems into
Modules.” Communications of the ACM 15: 1053–1058; reprinted in Hoffman and Weiss
(Eds), Software Fundamentals: Collected Papers of David Parnas. Boston, MA: Addison-
Wesley.

Patterson, David A. and Hennessy, John L. (1994). Computer Organization and Design:
The Hardware/Software Interface. San Mateo, CA: Morgan Kaufmann.

Perrow, Charles (1984). Normal Accidents: Living with High-Risk Technologies. Princeton,
NJ: Princeton University Press.

Pil, Frits K. and Cohen, Susan K. (2006). “Modularity: Implications for Imitation,
Innovation, and Sustained Competitive Advantage.” Academy of Management Review
31(4): 995–1011.

Pisano, Gary P. and Teece, David J. (2007). “How to Capture Value from Innovation:
Shaping Intellectual Property and Industry Architecture.” California Management Review
50(1): 278–296.

Rivkin, Jan W. (2000). “Imitation of Complex Strategies.” Management Science 46: 824–
844.

Rochet, Jean-Charles and Tirole, Jean (2003). “Platform Competition in Two-sided

Markets.” Journal of the European Economic Association 1(4): 990–1029.

Rosenberg, N. (1969). “The Direction of Technological Change: Inducement Mechanisms

and Focusing Devices.” Economic Development and Cultural Change 18(1): 1–24.

Rosenberg, N. (1982). Inside the Black Box: Technology and Economics. Cambridge:
Cambridge University Press.

Rosenberg, N. (1992). “Scientific instrumentation and university research.” Research

Policy 21(4): 381–390.

Page 36 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Sako, Mari and Murray, Fiona (1999). “Modules in Design, Production and Use:
Implications for the Global Automotive Industry,” paper presented at the Fall Meeting of
the International Motor Vehicle Program (IVMP) Annual Sponsors Meeting, 5–7 October
1999, Cambridge MA.

Sanchez, Ronald A. and Mahoney, Joseph T. (1996). “Modularity, Flexibility and

Knowledge Management in Product and Organizational Design.” Strategic Management
Journal 17: 63–76.

Schotter, A. and Teagarden, M. (2014). “Protecting Intellectual Property in China.” Sloan

Management Review Summer: 41–50.

Scott, W. R. (1987). Organizations. Englewood Cliffs, NJ: Prentice Hall.

Sherwood, R. M. (1990). Intellectual Property and Economic Development. Boulder, CO:

Westview Press.

Simon, Herbert A. (1981). The Sciences of the Artificial, 2nd ed. Cambridge, MA: MIT
Press.

Steward, Donald V. (1981). “The Design Structure System: A Method for Managing the
Design of Complex Systems.” IEEE Transactions on Engineering Management 28(3): 71–
74.

Suh, Nam (1990). The Principles of Design Oxford: Oxford University Press.

Teece, D. (2014). “The Foundations of Enterprise Performance: Dynamic and Ordinary

Capabilities in an (Economic) Theory of Firms.” The Academy of Management
Perspectives 28(4): 368–380.

Teece, David J. (1986). “Profiting from Technological Innovation: Implications for

Integration, Collaboration, Licensing and Public Policy.” Research Policy 15(6): 285–305.

Teece, David J. (1996). “Firm Organization, Industrial Structure, and Technological

Innovation.” Journal of Economic Behavior and Organization 31(2): 193–224.

Teece, David J., Pisano, Gary, and Shuen, Amy (1997). “Dynamic Capabilities and
Strategic Management.” Strategic Management Journal 18(7): 509–533.

Thompson, James D. (1967). Organizations in Action: Social Science Bases of

Administrative Theory. New York, NY: McGraw-Hill.

Topkis, D. M. (1998). Supermodularity and complementarity. Princeton, NJ: Princeton

University Press.

Tuertscher, P., Garud, R., and Kumaraswamy, A. (2014). “Justification and Interlaced
Knowledge at ATLAS, CERN.” Organization Science 25(6): 1579–1608.

Page 37 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Tushman, Michael L. and Murmann, J. Peter (1998). “Dominant Designs, Technological

Cycles and Organizational Outcomes.” In B. Staw and L. L. Cummings (Eds), Research in
Organizational Behavior (Stamford, CT: JAI Press, Vol. 20).

Tushman, Michael L. and Rosenkopf, L. (1992). “On the Organizational Determinants of

Technological Change: Toward a Sociology of Technological Evolution.” Research in
Organizational Behavior 114: 311–347.

Ulrich, Karl (1995). “The Role of Product Architecture in the Manufacturing Firm.”
Research Policy 24: 419–440.

Weick, Karl E. (1969). The Social Psychology of Organizing, 2nd ed. New York, NY:
McGraw-Hill.

Weinberger, D. (2008). Small Pieces Loosely Joined: A Unified Theory of the Web. New
York, NY: Basic Books.

Wernerfelt, Birger (1984). “A Resource-Based View of the Firm.” Strategic Management

Journal 5(2): 171–180.

West, Joel (2003). “How Open Is Open Enough? Melding Proprietary and Open Source
Platform Strategies.” Research Policy 32(7): 1259–1285.

Whitney, Daniel E. (2004). “Physical Limits to Modularity.” Available at http://

esd.mit.edu/symposium/pdfs/papers/whitney.pdf, accessed July 2005.

Whitney, Daniel E., Crawley, Edward, de Weck, Olivier, Eppinger, Steven, Magee,
Christopher, Moses, Joel, Seering, Warren, Schindall, Joel, Wallace, David (2004). “The
Influence of Architecture in Engineering Systems,” Engineering Systems Monograph,
MIT, Cambridge, MA (March). Available at http://esd.mit.edu/symposium/
monograph/, accessed July 2007.

Zhou, Y. M. and Wan, X. (2012). “The Bottleneck: Product Variety and Coordination
Failures at a Major Soft Drink Bottling Company.” Mimeo.

Zhu, F. and Iansiti, M. (2012). “Entry into Platform‐Based Markets.” Strategic

Management Journal 33(1): 88–106.

Notes:

(1) Simon used the term “near-decomposability” instead of modularity. Both concepts
refer to an abstract pattern of linkages between components that can be arranged as a
(nearly) block diagonal matrix. See Simon (1981: 212); Baldwin and Clark (2000: 63–76);
Augier and Simon (2003).

Page 38 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

(2) Many large technical systems are not entirely planned or designed. Such systems have
architectures in the sense of “abstract descriptions,” but some features of the system are
not the result of planning or conscious design: instead they arise out of chance, history,
mistaken beliefs, and serendipity. For example, cities, legal systems, and corporations all
have both designed and undesigned features (Alexander, 1964; Arthur, 2009). In
organization theory, these roughly correspond to the “formal” and “informal” aspects of
the organization (Scott, 1987; Nadler and Tushman, 1997; Baker et al., 1999; Gibbons,
2003; Gulati and Puranam, 2009; Gibbons and Henderson, 2012).

(3) The design structure and the task structure correspond respectively to what Simon
called the “state description” and the “process description” of a complex system (Simon,
1981: 222).

(4) Hughes (1983, 1987) uses the quasi-military term “reverse salient” to mean something
very similar to “bottleneck.”

(5) This assumes that no one builds a second bridge. However, if traffic only warrants one
bridge, the owner of the first bridge can either set the toll or credibly threaten to lower
the toll so as to make a second bridge unprofitable.

(6) In law and philosophy, “rights” refer to entitlements conferred by the state and/or
natural rights recognized under some ethical system. In contrast, economic usage of the
term “property right” focuses on the ability to exclude and the locus of effective control.
See, for example, Alchian, A. A., Concise Encyclopedia of Economics, http://
www.econlib.org/library/Enc/PropertyRights.html: “A property right is the exclusive
authority to determine how a resource is used.” Whether control over the resource
derives from power, community norms, or a legal system is secondary to the fact of
control.

(7) Thus strategic bottlenecks have low mobility as defined by Jacobides et al. (2006).

(8) In the resource-based view, “isolating mechanisms” prevent third parties from
imitating a firm’s technical and operational methods, thus maintaining the VRIN status of
its resource base. Isolating mechanisms are ways of exercising control over access to
technical and operational knowledge, thus are ways of effectively creating de facto
property rights over the solutions to technical bottlenecks.

(9) The methodology and consequences of “splitting” are discussed in Baldwin and Clark
(2000: Chs. 10, 11).

(10) Tipping is “the tendency of one system to pull away from its rivals in terms of
popularity once it has gained an initial edge” (Katz and Shapiro, 1994: 106).

(11) In their 1990 and 1994 textbooks, Hennessy and Patterson describe these new
architectural methods in detail. Hennessy and Patterson’s first textbook was published in

Page 39 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

1990. Sun’s founders, Andreas Bechtolscheim and Bill Joy learned them as Ph.D students
in the early 1980s.

(12) In 2017, it seems ironic to call firms started in the late nineteenth and early twentieth
centuries “modern,” but this is the term that Chandler used.

(13) http://sheldonbrown.com/k7.html#sram (accessed August 2014).

(14) In general, given a significant capacity change, price changes will ripple through the
system (Jacobides et al., 2014). There is no simple way to calculate precisely the value
that will be captured by the bottleneck owner. The difference in system value using the
old prices is an upper bound.

(15) Clean room reverse engineering is a method of copying a design without infringing
copyrights or trade secrets associated with it. Typically one group of engineers will
examine the behavior of an artifact and write up a detailed specification. Another group
with no knowledge of the original design will then implement the specification. The
implementation constitutes an independent invention, though its behavior mimics that of
the original design.

(16) http://arstechnica.com/business/2012/08/from-altair-to-ipad-35-years-of-personal-
computer-market-share (accessed October 2014).

(17) See https://www.ifixit.com/Teardown/iPhone+5+Teardown/10525 (accessed

September 2014) for a list of components of the iPhone 5 components and manufacturers.

(18) In contrast the application software is highly modular in use. See the discussion of
indirect network effects and users’ options in Subsection 5.2.2.

(19) “Supermodularity” is a functional property of components wherein increasing

contributions from one increases the positive impact of another. It is not related to
modularity, which is a structural property of components.

Carliss Y. Baldwin

Carliss Y. Baldwin, Harvard Business School

Page 40 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

Bottlenecks, Modules, and Dynamic Architectural Capabilities

Page 41 of 41

PRINTED FROM OXFORD HANDBOOKS ONLINE (www.oxfordhandbooks.com). © Oxford University Press, 2018. All Rights
Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a title in
Oxford Handbooks Online for personal use (for details see Privacy Policy and Legal Notice).

Subscriber: Freie Universitaet Berlin; date: 22 October 2018

J PROD INNOV MANAG 2004;21:401–415
r 2004 Product Development & Management Association

Value Creation by Toolkits for User Innovation and Design:

The Case of the Watch Market

Nikolaus Franke and Frank Piller

This study analyzes the value created by so-called ‘‘toolkits for user innovation and
design,’’ a new method of integrating customers into new product development
and design. Toolkits allow customers to create their own product, which in turn is
produced by the manufacturer. In the present study, questions asked were (1) if
customers actually make use of the solution space offered by toolkits, and, if so,
(2) how much value the self-design actually creates. In this study, a relatively sim-
ple, design-focused toolkit was used for a set of four experiments with a total of 717
participants, 267 of whom actually created their own watches. The heterogeneity of
the resulting design solutions was calculated using the entropy concept, and will-
ingness to pay (WTP) was measured by the contingent valuation method and
Vickrey auctions. Entropy coefficients showed that self-designed watches vary quite
widely. On the other hand, significant patterns still are visible despite this high level
of entropy, meaning that customer preferences are highly heterogeneous and diverse
in style but not completely random. It also was found that consumers are willing to
pay a considerable price premium. Their WTP for a self-designed watch exceeds the
WTP for standard watches by far, even for the best-selling standard watches of the
same technical quality. On average, a 100% value increment was found for watches
designed by users with the help of the toolkit. Taken together, these findings suggest
that the toolkit’s ability to allow customers to customize products to suit their in-
dividual preferences creates value for them in a business-to-consumer (B2C) setting
even when only a simple toolkit is employed. Alternative explanations, implications,
and necessary future research are discussed.

Introduction

T
he advent of the Internet has facilitated new
forms of producer-customer interaction in
Address correspondence to Nikolaus Franke, Vienna University of product development (Sharma and Sheth,
Business Administration and Economics, Department of Entrepre- 2004). One promising new form of interaction is out-
neurship, Augasse 2-6, 1090 Vienna, Austria; Tel: þ 43-1-31336-4582;
E-mail: [email protected]. lined in the concept of toolkits for user innovation and
 The authors would like to thank Steffen Wiedemann, Martin
design (Thomke and von Hippel, 2002; von Hippel,
Schreier, Marion Pötz, and the students of the E&I Research course
(2002–2003) for their valuable assistance. We also are indebted to 2001), or user design (Dahan and Hauser, 2002). Both
Helmut Strasser (chair of experimental mathematics and statistics, ideas are based on the proven ability of customers to
Vienna University of Economics and Business Administration) for his
contributions regarding the concept of entropy. Also, we would like to design their own products (von Hippel, 1988).
express our gratitude to Abbie Griffin and the two anonymous review- A toolkit is a design interface that enables trial-
ers, whose comments helped us to improve the article significantly.
Finally, the authors would like to thank the Jubilaeumsfonds der
and-error experimentation and gives simulated feed-
Oesterreichischen Nationalbank for funding this research project. back on the outcome. In this way, users are enabled to
402 J PROD INNOV MANAG N. FRANKE AND F. PILLER
2004;21:401–415

learn their preferences iteratively until the optimum the other hand, a considerable number of successful
product design is achieved (von Hippel and Katz, applications have been reported anecdotally. A recent
2002). The manufacturer, in turn, produces the prod- literature review, however, shows that only very few
uct to the customer’s specifications. Toolkits exist in academic studies have dealt with this new phenome-
various fields, ranging from computer chips (ASICS) non. Among them, merely anecdotal case studies are
to individualized athletic shoes. Depending on the the rule, and none have attempted a quantitative anal-
type of toolkit, the outcome is an individualized prod- ysis of the value such toolkits deliver to customers
uct (Park et al., 2000) or even an innovation (Thomke (Franke and Piller, 2003).
and von Hippel, 2002). The rationale underlying the The purpose of this study is to advance the under-
toolkit, however, is the same: it allows the customer to standing of the new concept of toolkits for user inno-
take an active part in product development. vation and design by analyzing their actual value
Any new concept must be analyzed thoroughly for creation from the customer’s perspective. In a quan-
its actual value. While several authors advocate the titative study of a design-focused watch toolkit in a
merits of the toolkit concept, others recently have B2C setting, the questions posed were (1) if customers
discussed its limitations (e.g., Agrawal et al., 2001; actually make use of the solution space offered by
Zipkin, 2001). They argue that the broad solution toolkits, and, if so, (2) how much value the process of
space offered by toolkits is of limited value because self-design actually creates.
for most users the cost of actively designing might This article is organized as follows: section 2
exceed the benefits of getting an individualized prod- presents a literature review of the concept of toolkits
uct. Particularly in business-to-consumer (B2C) ap- for user innovation and design, and section 3 de-
plications, the value creation potential of toolkits is scribes the methods applied. The findings then are
questioned, and this notion has received some en- summarized in section 4. The final section discusses
dorsement by the recent shutdown of Mattel’s ‘‘My- the implications of the study as well as alternative ex-
Design Barbie’’ and Levi’s ‘‘Original Spin’’ site.1 On planations and future research.

BIOGRAPHICAL SKETCHES
Nikolaus Franke is professor of entrepreneurship at the Vienna
Literature Review
University of Economics and Business Administration and is direc-
tor of the Research Center for Entrepreneurship and Innovation Toolkits for User Innovation and Design
Management at the same institution. His research interests include
innovation management, toolkits for user innovation, communities, Von Hippel (2001) defines toolkits for user innovation
and entrepreneurship. as a technology that (1) allows users to design a novel
Frank Piller is senior lecturer in the M.B.A. program at the Tech- product by trial-and-error experimentation and (2)
nische Universitaet Muenchen (TUM), Germany, and is director of delivers immediate (simulated) feedback on the po-
the TUM Research Group Mass Customization. His research in-
terests include customer integration, mass customization and per- tential outcome of their design ideas. This idea of
sonalization, innovation and technology management, as well as outsourcing design-related tasks in product develop-
strategic management. ment to customers stands in sharp contrast to the tra-
ditional practice of new product market research. The
1
traditional method of obtaining customer input is to
Levi’s closed its ‘‘Original Spin’’ (mass customization) operations
in October 2003, after being in this business for almost 10 years. Cus- gather data meticulously from representative custom-
tomers received the program quite well and happily paid the premium ers in a chosen market sector and then to use this
of about 10 to 20% compared to the standard products. Analyses show
that Levi’s lacked, among other problems, a functioning toolkit to help
(need-related) information in order to create ideas for
users to capture the entire value for themselves, and also for the com- new products (e.g., Lonsdale et al., 1996; Ranga-
pany. Thus, costly and error-prone interactions between the company swamy and Lilien, 1997). In order to reduce the risk
and its customers were rather the norm than the exception, leading to a
rather unstable business model (Pilter, 2004). Mattel, another pioneer- of failure, need-related information from customers
ing company in the field, abandoned its customized ‘‘MyDesign Barb- is integrated iteratively at many points in the new
ie’’ as well, though the company had a rather sophisticated toolkit for
children users on the Internet. In an interview conducted by the au-
thors, one manager said that the reason for stopping the program was creating in its manufacturing system. Given the relatively small volume
indeed too much user feedback. The 39-dollar customized doll (a pre- of sales of the customized dolls compared to overall sales volumes,
mium of about 100%) attracted so many orders that the supply chain Mattel decided not to invest in its manufacturing and logistics capa-
and fulfillment system was not able to handle all orders in the promised bilities but merely to keep the toolkit online without the order button.
time, leading to dissatisfied customers due to long delivery times. The Today, users still can configure and reconfigure dolls, but just for the
company was not prepared to capture the customer value it was fun of doing it.
VALUE CREATION BY TOOLKITS J PROD INNOV MANAG 403
2004;21:401–415

product development (NPD) process (e.g., Cusumano preferences; and (2) the problems associated with
and Selby, 1995; Dahan and Hauser, 2002; Holmes, shifting preference information from the customer to
1999). After many time-consuming iterations between the manufacturer.
the customer and manufacturer, a new or adapted It is common knowledge that customer preferences
product is found, usually at high cost. However, al- are heterogeneous and change quickly in many mar-
though enormous market research expenditure is the kets. The need for economies of scale, however, has
norm, the flop rates for new products are still rela- forced manufacturers either to satisfy the general
tively high (Cooper, 1999; Crawford, 1979; Griffin needs and preferences of a customer segment with a
et al., 2002). standard product (thus leaving many customers or
Toolkits offer potential advantages compared to potential customers dissatisfied, if only to a certain
the traditional method of new product development in degree) or to offer a custom-made product at a very
that they enable an individual user to state or specify high price. Recently, new production technologies
his or her preferences precisely. Moreover, the inter- dramatically have reduced the fixed costs of tooling
action between user and toolkit can be easier than the in manufacturing. These ‘‘mass customization’’ meth-
alternative of costly interaction between the user and ods have enabled custom goods to be produced with
manufacturer in the process of market research. Most near mass production efficiency (Pine, 1993; Tseng
notably, the information obtained with the help of a and Jiao, 2001; Tseng and Piller, 2003; Wind and
toolkit is located at the individual level: the manufac- Rangaswamy, 2001).
turer then can produce and can deliver a designated To date, only few studies have attempted to quan-
product to suit the individual user. The resulting— tify the heterogeneity of user preferences. In an em-
potentially far closer—fit between user preferences pirical study on Apache’s security software, Franke
and the product itself should yield a higher level of and von Hippel (2003a) show that users in fact do
satisfaction with the new product and subsequently have very unique needs, leaving many displeased with
should increase the customer’s willingness to pay standard products. Users even claimed that they were
(WTP). willing to pay a considerable premium for improve-
Obviously, there are variations in the types of avail- ments that satisfy their individual needs. In a meta-
able toolkits. Some very complex toolkits offer a large analysis of published cluster analyses, Franke and
solution space and cannot be employed without a Reisinger (2003) find evidence that this dissatisfaction
pr