Codification Challenges for Data Science in Construction

Ranjith K. Soman 1 and Jennifer K. Whyte, Ph.D. 2

Abstract: New forms of data science, including machine learning and data analytics, are enabled by machine-readable information but are
not widely deployed in construction. A qualitative study of information flow in three projects using building information modeling (BIM) in
the late design and construction phase is used to identify the challenges of codification that limit the application of data science. Despite
substantial efforts to codify information with common data environment (CDE) platforms to structure and transfer digital information within
and between teams, participants work across multiple media in both structured and unstructured ways. Challenges of codification identified in
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

this paper relate to software usage (interoperability, information loss during conversion, multiple modelling techniques), information sharing
(unstructured information sharing, drawing and file based sharing, document control bottlenecks, lack of process change), and construction
process information (loss of constraints and low level of detail). This paper contributes to the current understanding of data science in
construction by articulating the codification challenges and their implications for data quality dimensions, such as accuracy, completeness,
accessibility, consistency, timeliness, and provenance. It concludes with practical implications for developing and using machine-readable
information and directions for research to extract insight from data and support future automation. DOI: 10.1061/(ASCE)CO.1943-
7862.0001846. © 2020 American Society of Civil Engineers.
Author keywords: Building information modeling (BIM); Codification; Artificial intelligence (AI); Automation; Data science; Machine
readability; Construction.

Introduction planning (Choi et al. 2014), construction defects (Kwon et al.

2014), and lessons learned in a project (Oti et al. 2018). These
Machine-readable information is enabling new forms of data science, studies have advanced knowledge regarding frameworks and
including machine learning and data analytics. These methods offer methods to codify construction information. In addition to these
value to the construction sector through resource and waste optimi- studies, there is a need for work to extend understanding of the
zation, data-driven design, prescriptive analytics for rule checking, issues that prevent codification of information to improve uptake
visual analytics, performance predictions, operational analytics, and of data science in construction.
more (Bilal et al. 2016). Yet the construction sector is not taking ad- By examining information sharing across projects that use
vantage of these developments because data science is not widely de- BIM in the late design and construction phase, this paper aims
ployed. Bilal et al. (2016) have identified poor data management as to identify the codification challenges that arise in practice by ex-
one of the main factors that limit the application of data science in amining information sharing across projects that use BIM. In
construction. The sector is actively trying to overcome this limitation this paper, codification is defined as the process of conversion of
through building information modeling (BIM), an approach to incre- information into a structured and machine-readable format to sup-
mentally building structured (and, hence, machine-readable) informa- port the application of data science. Information refers to the
tion throughout the project cycle (Eastman et al. 2008; Jordani 2010), collection of data contextualized with relevant schema and seman-
and where it is machine readable, such structured information could tics so that insights can be made from the data. The codification
support the use of data science. Recent ideas such as the “digital twin” challenges are the issues that reduce the machine readability and
are also predicated on the availability of such structured information quality of construction information and, in turn, limit the uptake
[both geometries and behaviors; see Bolton et al. (2018)]. Research
of data science in the sector. Inspired by work that frames BIM use
has begun to develop approaches to, and document practices of,
as a complex social activity (Cao et al. 2014; Dossick and Neff
codifying information (to convert it into a structured and machine-
2010), this paper builds on and contributes to strands in the liter-
readable format) to improve delivery practices [e.g., in relation to
ature focused on data quality, machine readability, and BIM adop-
the construction phase (Goedert and Meadati 2008)], workspace
tion and implementation. Data analyses suggest codification
1 challenges that are organizational as well as technical in nature,
Ph.D. Student, Dept. of Civil and Environmental Engineering, Centre
for Systems Engineering and Innovation, Imperial College London, London relating to software use, information sharing, and construction
SW7 2AZ, UK (corresponding author). ORCID: https://orcid.org/0000-0003 process information.
-3967-9121. Email: [email protected] To develop the contribution, the rest of this paper is divided into
2
Fellow of the Institution of Civil Engineers, Laing O’Rourke/Royal four sections. The “Background” section provides a brief back-
Academy of Engineering Professor in Systems Integration and Director, ground on data quality and machine readability in projects and
Dept. of Civil and Environmental Engineering, Centre for Systems Engi- organizational issues associated with BIM adoption and data qual-
neering and Innovation, Imperial College London, London SW7 2AZ, UK. ity. The section “Research Method” describes the cases and method
Email: [email protected]
used in the study. The section “Codification Challenges in Con-
Note. This manuscript was submitted on April 18, 2019; approved on
December 12, 2019; published online on April 27, 2020. Discussion period struction” presents the findings. These findings are discussed fur-
open until September 27, 2020; separate discussions must be submitted for ther in relation to the literature in the “Discussion” section, and the
individual papers. This paper is part of the Journal of Construction En- conclusions are presented in the section “Conclusions and Future
gineering and Management, © ASCE, ISSN 0733-9364. Work.”

© ASCE 04020072-1 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Background it. Data provenance is the description of the origins of data and the
process by which it is manipulated. Jayawardene et al. (2015) have
As the construction sector becomes increasingly digital, most conducted a systematic literature review on the extensive data qual-
information is stored digitally and is accessible through servers or ity dimensions and consolidated overlapping dimensions of quality.
a common data environment (CDE) held in firms or projects (BSI This means that some of the data quality dimensions are inter-
2018; Preidel et al. 2016). However, being digitally accessible does dependent. For example, the data quality dimension related to se-
not mean that the information is machine readable because the se- mantic accuracy might depend on the timeliness dimension
mantics may not be embedded in the data (Hendler and Pardo because the data may be accurate with respect to time. However,
2012). Semantics could be derived from the data using advanced the same value may be inaccurate at a different time. At the outset,
machine learning techniques such as natural language processing a dataset is considered to be of good quality when the measure of
and deep learning (Carrillo et al. 2011; Wang 2017). Nonetheless, these dimensions is high, leading to better inferences.
this is a resource-intensive process that requires training models for What constitutes machine-readable data? Data in a structured
achieving satisfactory accuracy and has costs associated with it. format that could easily be processed by computers are considered
The result of this process may also not be of high quality (Wang as machine readable. Berners-Lee (2006) has stated a set of’ rules
2017). Despite the existence of such data-cleaning algorithms for creating structured data so that it can be connected and inter-
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

across sectors, poor data quality is costing $3.1 trillion in the preted easily by machines. The first one is to index the data to make
United States (Quintero et al. 2015). In addition, the poor quality it digitally accessible by storing it on online servers so that it can be
of data is increasing operational costs, decreasing revenue, and re- easily accessed by computers as well as people. This relates to the
sulting in missed commercial opportunities (Loshin 2010). Within accessibility dimension of data quality. Indexing the data and stor-
construction, Sacks et al. (2017) have described how the quality of ing it online increases the accessibility because it is easy to find.
input information influences semantic enrichment of BIM when The second rule is to structure the data with relevant schemas for
using machine learning. Moreover, Farias et al. (2018) have shown easy interpretation by machines. This step makes the data struc-
the effects of poor quality data resulting in wrong inferences when tured such that semantic relations are embedded within it, resulting
they tried to extract building views using a rule-based method. in better inferencing and thus improving the syntactic accuracy of
Whyte et al. (2016) articulated how managing change in large the data. The third rule is to make the schemas public and machine
datasets becomes a focus in an era of big data in which project readable by using open-source schemas to describe the data model
information is increasingly characterized by volume, velocity, and so that interpretations can be made by computers without any pro-
variety. Recent work has further characterizes such data as also in- prietary data interfaces. Proprietary data formats limit data infer-
cluding characteristics of veracity and value (e.g., Younas 2019). encing as the schema by which data is modeled is only accessible
These issues of data quality occur because construction data is to few applications. Therefore, using open-source schema would
heterogeneous, and its veracity is not always known. Data cleaning increase the measure of accessibility dimension because more ap-
related to the variety (heterogeneity) and veracity characteristics of plications can use the open schema to derive the context of data
big data is especially difficult when compared to data cleaning re- for inferencing. The last rule is to link the data with other datasets
lated to other characteristics such as volume and velocity (Fan 2015; so that better inferences can be made by deriving the context in-
Janssen et al. 2017). Therefore, there is a need to keep the data of the formation. This improves the consistency dimensions of data
highest quality and in a machine-readable format (maintaining the quality because the same data is linked to multiple datasets. Link-
data relationships) as far as possible to have the best inferencing. ing would ensure that there are no conflicts in the data about a
concept stored in multiple databases. Based on these rules, struc-
tured information can be classified into five types in the increasing
Quality and Machine Readability order of machine readability, as shown in Table 1. Increasing the
What constitutes a good quality dataset? According to Wang and machine readability of the data, in turn, increases the data quality
Strong (1996), a good quality dataset is the one that has enough in- because the dimensions relating to accessibility, accuracy, and
formation embedded in it for a particular use by the user. Research- consistency are improved in the process.
ers have set out multiple dimensions to assess the data quality
concerning big data analytics (Batini and Scannapieco 2016; Cai Limits of Existing Research
and Zhu 2015; Delone and McLean 2003; Naumann and Rolker
2000; Wang and Strong 1996). For this paper, the focus is on the What is limiting the generation of good quality machine-readable
following data quality dimensions based on Batini and Scannapieco information within the sector? It does not appear to be technical
(2016) because they best reflect the implications of codification chal- development because novel technical solutions are being developed
lenges. Accuracy is the closeness of the measured/represented data by construction informatics researchers with a focus on the integra-
and reality. There are two kinds of accuracies: semantic and syntac- tion of data in the sector; for example, through the use of data stan-
tic. Semantic accuracy relates to the closeness of the data value to dards (Krijnen and Beetz 2017; Pazlar and Turk 2008), cloud-based
reality, whereas syntactic accuracy refers to the closeness of the data BIM (Beetz et al. 2010; Singh et al. 2011; Zhang et al. 2017a), and
representation to the expected data type/model. Completeness is the linked-data technologies (Kim et al. 2018; Pauwels et al. 2015;
measure of information content present in the data compared to the
extent of information content required to be present in the data to Table 1. Level of machine readability of the data based on linked data
perform a particular task. Temporal dimensions refer to currency, principles set out by Berners-Lee (2006)
volatility, and timeliness. Currency relates to the promptness of data
updates. Volatility refers to the frequency at which the data variance Quality of data Principles for publishing a machine-readable data set
occurs. Timeliness refers to the suitability of the current data to per- 1-star Data is available on the web
form a task. Consistency refers to uniformity and constancy of data 2-star 1-star data structured in a proprietary format
with respect to the semantic rules defined over multiple data items. 3-star 1-star data structured in a nonproprietary format
Accessibility refers to the ability of data to be accessed by a user 4-star 3-star data that is published using open standards
5-star 4-star data with links to other 4-star datasets
(human user or computer program) and generate information from

© ASCE 04020072-2 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Pedro et al. 2017, Zhang and Beetz 2015). It does not appear to be To address the data quality issues emerging due to such
policy interventions either. Standards and public mandates have document-based and model-based information-sharing practices, it
placed BIM at the heart of the information management required is necessary to study information flow across project teams in de-
to coordinate processes in project delivery and operation of infra- tail. By better characterizing the practice, such empirical work can
structure, making BIM central to digital tools and workflows in then inform future technical developments (e.g., Hartmann 2008)
projects (BSI 2018; Sacks et al. 2018). Instead, the literature sug- and address challenges raised in prior work in areas such as auto-
gests that the issues may be both organizational and technical in mated scheduling (e.g., Han and Golparvar-Fard 2017). Previous
nature. research on the use of BIM in organizations articulates antecedents
Prior research has assessed BIM adoption across different mar- to BIM uptake (Taylor 2007) and identifies organizational issues
kets, using a model of diffusion area, macromaturity components, that affect BIM implementation, such as organizational divisions
macrodiffusion dynamics, and so forth, and validated this model by (Dossick and Neff 2010). It describes how practices are always hy-
applying it to assess BIM adoption amongst 21 countries (Kassem brid, overlaying a range of old and new media and processes (Harty
and Succar 2017). This work has determined the BIM project ob- and Whyte 2010; Whyte 2011), with the roles of construction pro-
jectives, critical success factors, and operative critical success fac- fessionals also evolving (Akintola et al. 2017; Jaradat et al. 2013;
Sebastian 2011). Such work draws attention to the organizational
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

tors for effective implementation of BIM (Chegu Badrinath and

Hsieh 2019). It has identified the success factors for adoption of factors associated with information use, whereby technological in-
BIM in a company, selection of projects within the company to tegration cannot be assumed to foster closer collaboration across
implement BIM, and selection of BIM services and software (Won companies (Dossick and Neff 2010). This literature, which under-
et al. 2013). It has surveyed the degree of implementation of BIM stands BIM use as a complex socialized activity (Cao et al. 2014),
statistically by evaluating the level of BIM implementation and provides an approach that can be used to study the codification
quality of collaboration and communication in BIM-enabled proj- challenges and design quality issues that emerge in leading
ects, and linking discussing its influence on uptake of integrated practice.
delivery systems (Chang et al. 2017), and developing strategies for
using BIM to reduce rework in construction (Hwang et al. 2019)
and improve collaboration through the development of BIM-based Research Method
platforms by analyzing requirements and details of elements
To understand the challenges of codification in construction,
needed for a collaborative work model (Zhang et al. 2017b). In ad-
three construction projects are studied qualitatively to investigate
dition, Gu and London (2010) have created a collaborative BIM
the digital tools and workflows used in the projects, structured
decision framework to facilitate BIM adoption through a four-part
and unstructured information flows in these projects, and the
method. The framework first defines the scope, purpose, roles,
problems associated with the information flow. These three lead-
relationships, and project phases, followed by developing a work
ing projects are a multistory residential student apartment block
process roadmap, identifying the technical capabilities and the lim-
in the United Kingdom (Case 1), a metro rail infrastructure
itations of tools, and finally customizing these to suit the capabil-
project in India (Case 2), and a major water infrastructure mega-
ities and skillsets of the project team. Building on this study, Singh
project in the United Kingdom (Case 3). The multistory residen-
et al. (2011) have determined technical requirements for a BIM
tial apartment project (Case 1) is an exemplary project exhibiting
server to serve as a collaboration platform. These studies have ex- the use of BIM in the United Kingdom, constructed by a leading
tended the existing knowledge base, and they give a deeper under- contractor and using state-of-the-art offsite manufacturing ap-
standing of the problems associated with BIM implementation, and proaches in construction. The metro rail project in India (Case
suggested steps for the effective implementation of BIM. However, 2) is pioneering BIM implementation amongst the metro projects
data quality issues emerging from the problems associated with in India, incorporating learning on digital implementation from
BIM implementation is relatively less studied. global megaprojects. The water infrastructure project (Case 3) is
Previous research on data quality within the construction sector one of the biggest construction projects in the United Kingdom,
has studied semantic and syntactic accuracy of BIM, BIM quality using innovative technological solutions to futureproof construc-
assurance processes for the design stage and data quality issues in tion and deliver a physical asset as well as a digital asset for op-
the design model, and completeness of information in BIM for eration. Early in the study, the first author, who collected the
facility management. Solihin et al. (2015) have identified require- data, also visited another infrastructure project and a commercial
ments for good quality for BIM in an industry foundation class retrofit project, and these three projects were then chosen due to
(IFC) format. Building on this study, Lee et al. (2018) have pre- their significance and because they use a level of digital collabo-
sented a semantic rule-checking process to ensure data quality per- ration categorized as BIM level 2. These projects follow the BIM
taining to semantic and syntactic accuracy is maintained while BIM level 2 recommendations set out by the mandate (it is mandated
in an IFC format is exchanged. In another study, the quality of the in Case 3 and seen as best practice in the other cases). Although
information in the design phase was assessed using a structured and the metro rail project (Case 2 is in a country that does not have a
quantifiable process based on a BIM quality assurance by Donato regulatory framework for BIM level 2 recommendation), the
et al. (2017). Mirarchi and Pavan (2019) have analyzed the data owner required the adoption of BIM level 2 as international best
quality issues concerning accuracy, consistency, and completeness practice, which justifies our choice for selecting the case. Infor-
dimensions of the BIM models created during the design. For fa- mation sharing across project teams in the late design and con-
cility management, Zadeh et al. (2017) have proposed a framework struction phases of the three projects is studied qualitatively by
to assess the quality of BIM, focusing on the completeness dimen- visiting the projects, analyzing internal and publicly available
sion. These studies have advanced the knowledge on data quality documents, observing meetings, and conducting informal and
issues associated with BIM models. A limitation of these studies is formal interviews on the use of product and process information
that they do not address the data quality issues emerging due to the during the construction stage (refer to the Appendix).
wider practice of using document-based and model-based informa- Within each project, there was an initial setup meeting to present
tion sharing in the sector. the study and identify interviewees. The interview protocol covered

© ASCE 04020072-3 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 questions of communication, software tools used, BIM, collabora- to an iteration between these steps of data analysis as the com-
tion, and information flow. The data analysis phase overlapped with parison across cases was instructive in directing analytic atten-
the data collection phase. The taped interviews were transcribed, tion within cases. Finally, the third step was a more in-depth
and field notes were typed up. These were read and reread between analysis of the third case study (for which there was more de-
the project meetings. Summaries of interviews were sent back to tailed information).
the interviewees for member checks. All the data was organized A starting assumption of this current study is that there are data
into cases and stored into the qualitative analysis software. These quality issues caused by the way digital tools and workflows are
methods draw on a qualitative case study approach (Eisenhardt and used in late design and construction stages. This research thus ad-
Graebner 2007), building insights across the three cases from dresses the questions: How do codification challenges arise because
multiple sources (site visits, documents, field notes, and interview of the different digital workflows and working practices across
transcripts). projects? How do these lead to data quality issues? Within case
• Multistory residential student apartment block in the United analysis of the multistory student apartment raised issues of data
Kingdom (Case 1): To study this case, the first author visited the interoperability, information loss, use of 2D CAD, and lack of de-
construction site and offices of the projects, had informal con- tail in the schedule. In the metro rail project, the issues of data inter-
versations with the digital and planning engineers, examined operability, use of 2D CAD, and lack of detail in the schedule were
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

construction documents and models, and studied the software also present, but there were also issues of unstructured communi-
tools used to understand the information embedded in the BIMs, cation channels, document control, and lack of skills to adopt dig-
construction schedules, and other reports such as design calcu- ital technologies. In the water infrastructure project, additional
lation, method statements, and so on. These documents were issues were identified concerning problems with CDE, lack of
centrally hosted on a CDE, and the first author had access to it process codification, and long processing times. In the water infra-
while being there at the office. structure project, the design workflow, work package plan, con-
• Metro rail infrastructure project in India (Case 2): To examine struction program, BIM models, and drawings in the CDE were
this case, documents such as the BIM execution plan, presenta- studied to understand the level of detail and machine readability
tion documents for training, and press releases were studied to of the documents. Coding was done on the field notes and interview
understand digital information management practices. The pro- transcripts to identify different issues related to codification and
ject manager, chief site engineer, casting yard engineer, and information sharing. The software was used to track the patterns
BIM consultants, who form a cohort of the key decision makers emerging from these data. These codes were organized to find
during the construction stage, were interviewed informally to themes. The identified themes were then analyzed based on the data
get an insight into the extent of codification in the information quality dimensions to understand their implications about data
flow during their daily work practices. Field notes were taken quality.
during the interview. In addition to these interviews, the casting
yard, viaduct construction site, and a station site were visited to
understand the on-ground practice of various activities. Further Codification Challenges in Construction
insight into this case was obtained through a workshop, co-
Table 2 summarizes the codification challenges observed from
organized by the authors, with 40 participants, including client
studying the projects. Low machine readability of data is a signifi-
representatives of six major Indian metro-rail projects along
cant challenge for codification, which was observed across the proj-
with technology providers and delivery teams. The workshop
ects. Product information is well codified through BIM, CAD
provided a perspective on the digitization of this project in drawings, analysis models, and so on in all the projects. However,
the broader landscape of Indian metro rail construction. the codified information is distributed amongst different formats
• Water infrastructure project in the United Kingdom (Case 3): To and databases, limiting the application of analytics. In addition,
understand the codification challenges in this case, eight semi- multiple modes of communication, multiple CDEs, and lack of pro-
structured interviews were conducted. All eight interviewees cess change also limit the codification of information in the projects
had more than 10 years of experience in the construction sector studied. Different codification challenges observed in the cases
and had worked with different major projects in the United have been mapped in Table 2. These topics are discussed in detail
Kingdom and abroad. The interviewees’ areas of expertise cov- in this section.
ered design, planning, project engineering, digital engineering,
prefabricated construction, and information management. All
interviewees had teams working with them on their areas of spe-
Table 2. Codification challenges across the cases
cialization and also interacted with the other stakeholders in the
project. These characteristics make them ideal for case-based The The The
research. Following the semistructured approach, ensured par- Codification student metro water
ticipants would talk broadly on their experiences with informa- challenges Observations apartment project project
tion flow using digital collaboration tools. Seven out of the eight Software Interoperability X X X
interviews were recorded, and transcripts were made from the usage Information loss during X — X
recordings. In addition to the interviews, the first author con- conversions
ducted multiple visits over 2 weeks to the project office, obser- Modeling technique — — X
ving meetings and the work practice. The first author also had Information Unstructured information X X X
access to CDE and documents such as a construction program, sharing sharing
look-ahead schedule, and method statements. Drawing and file-based sharing X X X
Data analysis took place in three steps. First, each of the cases Document control bottleneck — X X
was separately analyzed. Second, the cases were compared and Lack of process change X X X
contrasted. The initial analyses were conducted during data collec- Process Loss of constraints X X X
tion, so early analyses focused and directed later data collection. information Low level of detail X X X
The within case analyses and the cross-case comparisons also led

© ASCE 04020072-4 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Software Usage Information Loss during the Conversion
For structured information flow through a CDE, the files are con-
This section presents the codification challenges related to software
verted to a PDF format. The original file may be uploaded as a
usages such as interoperability, information loss during conver-
supporting document, but this is not a necessary requirement.
sions, and multiple modeling techniques during the late design and
construction phase. The implication of these challenges on the data If we’re conveying CAD information, it’s being uploaded as a
quality with respect to dimension accuracy, completeness, acces- supporting file. There’s a facility in [CDE1] that when you
sibility, and data provenance is explained in this section. upload a PDF, you can also upload a secondary file. (Infor-
mation manager, C3I4)
Interoperability
Interoperability was raised as a central problem by the interview- The student apartment and the water project followed a work-
ees, especially when working with multiple CAD tools in projects, flow that required documents to be uploaded as a PDF to the CDE,
resulting in data loss during format conversions. Even within the resulting in the loss of information during conversions (access to
same software environment, there are problems related to data com- CDE C1D4, access to CDE1, C3D2). However, the metro project
patibility while working between different software versions. used a workflow without this requirement, resulting in retaining the
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

information (digital project management, C2D2).

Sometimes the drawings where I am using this MicroStation, The conversion from native formats to a PDF format results in
but sometimes they were drafted from the client, let’s say, in the loss of semantic relationships embedded in the file. The loss of
Autodesk. And transferring things from Autodesk to Micro- the semantic relationship between datasets results in data silos and
Station, you lose data [ : : : ] the 2019 will open the 2018, limits machine readability. This aspect of the loss of information
2017, and 2016. But when the 2020 comes into play, then results in incomplete information and lowers the data quality re-
you cannot open it anymore with the 2019 files that are gen- lated to the completeness dimension. Furthermore, during the con-
erated with the 2020. (Technical manager, C3I5) versions, the metadata related to the original file is lost. This lowers
the data quality dimension associated with data provenance.
Here, the drawings are made using different CAD tools such as
AutoCAD and MicroStation. However, the data may not be opened Multiple Modeling Techniques
(or edited) using the same tools with which they were created. This Software tools allow different methods for creation of information.
creates problems of interoperability and loss of information when However, not all methods lead to the information being reusable.
data in one format is converted to another. Similar problems also The modeled information would have multiple uses, which may not
occur when using multiple versions of the same tool. be known to the creator of information.
This issue was observed in all three cases but at different scales.
Uneven distribution of tools would result in issues of data interop- So, if they use the wrong tool to model something, you don’t
erability. The student apartment (Case 1) and the metro project have the appropriate dataset to it [ : : : ] when you go into your
(Case 2) had predominantly used tools from a single software ven- authoring tool, do I use a slab tool or do I go and use a generic
dor (Autodesk for the student apartment and Bentley for the metro solid modelling tools then try and attach a dataset to it [ : : : ] if
project) for executing most of their tasks, resulting in better inter- they’re not, therefore we have to go in there and say, well I
operability when compared to the water project (Case 3), which can’t just say there’s a slab now, that’s just a piece of geometry
uses tools from different software vendors. The scale of the project (Digital engineer, C3I3)
has an influence on this diverse distribution of software use. The
scale of the student apartment was smaller than that of the metro For example, a slab could be modeled as a generic solid model
and water projects, with a leading firm involved in both design and with a dataset attached to it or as a slab component. From the hu-
construction, resulting in evenness of software usage (contractor man point of view, the information contained in both models is the
office visit, C1S1). Even though the metro project had different same. However, during automated quantity take-off, the slab mod-
firms engaged in design and construction, the information manage- eled as a generic solid would not be considered because the com-
ment was handled by an owner support organization, resulting in puter cannot classify it as a slab.
evenness in the data (BIM consultant 1, C2I4). The water project, For tunneling purposes, when you try to extract 2D drawings
on the other hand, had multiple firms working on different phases from 3D BIM models, those drawings are not as correct and as
of the projects with their own sets of tools, resulting in issues of detailed as they used to be. They have glitches, they have
data interoperability (digital engineer, C3I3). errors (Technical manager, C3I5)
Data interoperability is a significant problem when it comes to
data quality and machine readability. If the data is locked to a pro- This issue was observed predominantly in the water project.
prietary format, it limits the application of data science. Software Limited observation of this issue in the student apartment and
vendors provide a proprietary application program interface (API) the metro project can be attributed to the absence of multiple firms
to access the data. However, access to the data through APIs is in modeling the data.
limited, and information access is limited to proprietary domain- Organizational divisions in large projects lead to lower machine
specific programs. This limits data science as different systems can- readability and data quality because of differences in modeling ap-
not talk to each other and derive the context from the information. proaches and software tools used. Software tools offer different ap-
Although there are open-data formats available for the exchange of proaches to model the same information at the same level of detail.
information, these are not often used in the construction phase and Moreover, modeling approaches used by the firms are guided by
are only submitted at specific data drops. Evidence also shows that the norms and practices followed in the firm. These norms may be
there is a loss of information when converting between formats be- different for the firms who use the data. While examining the cases
cause of inefficient exporters and importers. This problem reflects of the student apartment and metro project, where the information
data quality related to the accessibility dimension. As long as this modeling is performed by a single firm, the issues such as inter-
data remains inaccessible, algorithms make inferences with limited operability and improper modeling of information were limited.
data, resulting in incomplete inferences. However, in the water project, where the information modeling

© ASCE 04020072-5 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 spanned over different firms, there was evidence of issues of inter- that at all. So, at the very end of the day, on the eleventh hour
operability of data and improper information modeling. Therefore, when I have finished everything, by the way, we have this
the information created could be used as a digital submission but spread sheet, and it’s exactly what I wanted to do (Technical
limits further use. Even when the information is present in the manager, C3I5)
model in the correct format, the level of detail of the modeled in-
formation is less than desired, making the information not fit for In this case, the information required for a task was already
further use. This problem reduces the quality of the data concerning available. However, it was not accessible for the person who needed
accuracy and completeness. The fact that the data exists but not in it, who was not part of the group within which the information was
the way it was supposed to be a case of syntactic inaccuracy. shared. This led to the recreation of the information and loss of
productive time. The unstructured information-sharing issue was
observed in all cases. The office visits in the student apartment
Information Sharing (C1S1), meetings with project participants in the metro project
This section presents the codification challenges in construction, (C2I1, C2I2, C2I3), and interviews with participants in the water
such as unstructured information sharing, drawing and file-based project (Table 3) revealed the problem of unstructured information
in the projects.
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

sharing, document control issues, and lack of process change.

Even when there are structured workflows for information shar-
The implications of these challenges on the data quality dimensions
ing, project participants find it easier to use the unstructured chan-
are presented in this section.
nels of communication. They often find structured information
Unstructured Information Sharing flow through the CDE slow and complicated. Despite being more
Information shared over modes such as meetings, reports, e-mails, traceable and accountable compared to unstructured information-
etc., contains relevant data for decision making. This information is exchange practices, the complexity of the new structured methods
embedded in documents, is shared in a human-readable format for information sharing and the poor understanding of workflows
(documents, PowerPoint presentations, drawings, etc.), and is in across the teams lead to the use of a combination of structured and
formats that are both human and machine readable (spreadsheet, unstructured channels for information sharing. This issue, however,
BIM, etc.). The main limitation of the information shared through has an implication on data quality, lowering it with respect to the
these unstructured channels is its accessibility, which is limited to accessibility dimension because data is not available in a common
people involved in the meeting or e-mail conversation. repository. Instead, it is distributed in different silos and e-mail da-
tabases, and the access is limited. In addition, it introduces incon-
A guy made a design on a spreadsheet for quantities. Some sistency as the same information is distributed amongst different
people knew about it; he logged it as well. And I didn’t know databases which are not connected or synchronized. Tracing the

Table 3. Information sharing between different teams

Media Example evidence from the dataset
Common data Formal submissions, drawing receipts, design and temporary works: “if it’s formal document submission, we do it through
environments (CDE1, [CDE1] [ : : : ] When I receive drawings from [design consultant], I get them through [CDE2]. And quite a lot of the designers use
CDE2 from different [CDE2] and [ : : : ] Well, we try get all the design functions, including temporary works to use [CDE2].” (Project engineer, C3I1)
vendors)
Reports Spreadsheets and documents: “There’s lots of reporting on the project [ : : : ] And then that gets out into various outputs, so that
could be just a schedule in Excel. Lots of Excel outputs as well, huge amount of Excel outputs. And, if it’s a commercial discussion
there may need to be some narrative around it, so using Microsoft Word to develop a narrative.” (Project Planner lead, C3I2)
Meetings Design review meetings: “You could do a design meeting, review something and then say, write comment on that [ : : : ] to
understand what information, they’re going to require at a particular stage. So that may consist of meetings; that might consist of
face-to-face conversations, emails, etc.” (Digital engineering lead, C3I3)
Buildability meetings: “attendance to buildability meetings and trying to get out of them what sort of temporary works may be
needed to build something” (Project engineer, C3I1)
Client meetings: “I will be going off-site to attend meetings with the client” (Information manager, C3I4)
Design for Manufacture and Assembly (DfMA) input: “I’m trying to attend meetings and troubleshoot and try to help and provide
technical input into the design and assisting the designer and support team” (Technical manager/DfMA coordinator, C3I5)
Email Highly used: “So, obviously we do use emails a lot.” (Project Planner lead, C3I2)
A normal type of communication: “Then emails, meetings, usually types of communication.” (Project engineer, C3I1)
“Stakeholders, yes, it’s certainly meetings and emails. Most of our stakeholders don’t want to use [CDE1], because of the admin
that comes with that” (Principal engineer, C3I8)
Provides remote precision: “If I’m communicating over longer distances or if I think to myself, I’d better make a precise request,
then it’ll be emails. We don’t use a communicator-type facility in [CDE1]”. (Information manager, C3I4)
“Stakeholders, yes, it’s certainly meetings and emails. Most of our stakeholders don’t want to use [CDE1], because of the admin
that comes with that” (Principal engineer, C3I8)
Remote conversations Online meetings: “Generally, like a Skype, conference calls, linked meetings” (Senior digital engineer, C3I6)
Online communications and records of design logs, discussion points, online forums: “Yes, so all the design data is held within a
common data environment, which was [CDE2]. And so, I managed that area and access to that area. Then all communications
were stored on SharePoint on Microsoft online, so everyone had access to registers or design logs or discussion points, almost
used as an online forum where anyone could ask questions” (Senior digital engineer, C3I6)
Telephone calls with design consultants: “So, I’ll start with between us and design consultants: there’s meetings, emails, and
phone calls. I prefer meetings and phone calls” (Principal engineer, C3I8)

© ASCE 04020072-6 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 source of the data and its history is also difficult when using un- must be approved by the relevant authority (depending on the docu-
structured channels for information sharing, thereby reducing the ment). The quantity of documents uploaded to CDEs in the projects
quality of data associated with the provenance dimension. These studied is enormous, and in each project, processing information
issues limit the data science as the datasets for drawing inferences becomes a significant task, with bottlenecks in the process leading
are siloed and disconnected. to workarounds and data quality issues. As the authorization of
documents is limited to specific individuals, they tend to get asked
Drawings and File-Based Sharing for a huge volume of authorizations, and this slows down the in-
Even though the projects follow BIM level 2, the engineers inter- formation flow.
viewed are more comfortable performing submissions and appro-
vals using drawings rather than model-based information sharing. So, there’s certain people who are responsible for issuing in-
This is mainly because they find it intuitive to use drawings. formation or authorize certain communications, and if they’re
not available then things can stop. Or they may have a high
I use, I’m not very good at, but I use all the navigator tools that volume of these authorizations to do that it takes them a lot of
we’ve got here. But I prefer to use AutoCAD because I find it time to get through (Project planner, C3I2)
a lot easier (Project engineer, C3I1)
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

This process of authorization means that multiple versions of

Despite having a BIM coordination tool, Bentley Navigator, the designs can be circulating in different parts of the project. For ex-
project engineers resort to using the CAD tool because they find it ample, whilst one design is in use by the construction team, in the
easier. In the contractor’s offices of the projects studied, the engi- meantime, the designers could have progressed the design, and the
neers had drawings on their desks and the CAD software opened on
latest design is not uploaded as it is in the queue to get approved.
the monitors. If they find errors in the drawings, they modify and
edit them first on paper and then on the computer. Design teams and checkers and approvers could be progress-
ing designs but then it’d be held up when someone say high up
Because I have to open up every drawing individually and
needed to actually approve the whole design. (Digital engi-
print them all, or even if you do it the other way around,
neer, C3I3)
it’s very slow anyway. And then, once I’ve printed them, re-
viewed them all, you could do it on there [computer] but
it’s not the best way because we haven’t got the technology. It’s just making sure that I’m getting the latest information and
I haven’t got a big screen (Project engineer, C3I1) no-one’s updating it in the background and then the right ver-
sions are going onto [CDE1] [ : : : ] it’s no longer the most
The engineers use laptops with small screens. Some of the hot current version anymore by the time I’m reviewing it. (Project
desks have an additional screen; however, these were also small engineer, C3I1)
[less than 58.42 cm (23 in.)]. This is highly inconvenient when re-
viewing large drawings as they pan and zoom to detect mistakes.
This issue was also observed in the student apartment and metro [ : : : ]when this revision has been updated from the designer to
projects. revision 10 and I sit here on my desk checking the revision 1
During the visit to the contractor’s office (C1S1) in the student and the designer has the revision ten, then that revision 10 is
apartment, the first author observed multiple discussions between internal [ : : : ] because he keeps on updating but he hasn’t he
engineers using drawings as a common representation medium. On hasn’t put it on the [CDE1].(Technical manager, C3I5)
the site, a tablet-based application was used to open the drawing.
This shows that the information available in the CDE is not the
Similarly, in the metro project, the workflow for design, review, and
latest version, thereby reducing data quality associated with the
approval (C2D2) presents how drawings are reviewed and proc-
timeliness dimension. The CDE has provisions for labeling status
essed using the CDE. Such evidence points towards the drawing
of a document as work-in-progress. However, even with that, it is
and file-based information sharing in the construction phase.
hard to ascertain whether the information at hand is the latest
The file-based sharing impacts the data quality dimensions as-
because the work-in-progress documents can only be accessed by
sociated with temporality and consistency. Most of the file-based
internal teams. This accessibility becomes a dimension of data
manipulations happen within the computer and are uploaded only
quality, which such approval processes make challenging. Addi-
when complete. Hence, there is a mismatch between the rate of the
tionally, even when the information is internally approved, it does
volatility of data and currency of data. The volatility is high because
not get stored on the CDE. Another approval is necessary to share
the data is manipulated on the users’ desks (for example, they are
manipulating the information in a printed drawing). However, the the information with other stakeholders, thus limiting other stake-
currency of the data is low because it is uploaded as a batch. That holders and algorithms from accessing this information for further
means the data is updated at a lesser speed than it is varied. This has decision making. In the metro project, an innovative approach to
an implication on the data quality associated with the timeliness address this issue for drawings was implemented, placing a quick
dimension because the data in the CDE is not the latest version. response (QR) code (a matrix barcode) in the document and a mo-
File-based sharing impacts consistency too. The files act as indi- bile app to scan the QR code and inform the user whether the draw-
vidual entities and have information from related files in them. ing is the latest version or not. However, the document status must
Unlike a model, this information is not connected. Therefore, when be continuously updated to make this useful and is limited to the
the source is updated, the information in the file may not be up- issued drawings and not the work-in-progress drawings.
dated, which introduces inconsistency problems. Although it seems to be straightforward from the outset, many
users find document control frustrating because submission ends up
Document Control Bottlenecks being a long process even when all the attributes are correct. For
Document control plays a vital role in the flow of structured infor- example, engineers submit the packages to the document controller
mation in the projects, and document control professionals are in their firm, who sends it to the document controller in the other
tasked with managing the access, version control, and availability firm (receiving end), which is then sent to the team lead and, finally,
of documents. Before being uploaded to the CDE, such documents to the user who would get the useful information out of that package.

© ASCE 04020072-7 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 This is a long process with checks and iterative cycles involved in the metro project as well. Conversations with the project manager
each stage. The document controller makes sure that the files in the (C2I1), chief site engineer (C2I2), and BIM consultants (C2I4,
CDE have the relevant attributes before they are published in the C254, C2I6) revealed that the project team weren’t exposed to
CDE. If there are missing attributes, the submission is rejected. structured information flow used with digital technologies in
In the case of a request for information (RFI), if the document con- the past. This made the implementation of CDE-based structured
troller does not understand the request, it gets rejected. At times, it workflows difficult despite the training given to the participants,
takes more than 2–3 weeks for the document to reach the recipient resulting in a combination of structured and unstructured work-
while following the document control workflow. This is essentially flows in the project.
slowing down the whole information flow by implementing a system
that was supposed to speed up the process. The complexity of workflows and document control measures has
implications for data quality. There are shared norms, values, and ex-
It took us three weeks to actually get the package with all the pectations for the users regarding the tools, such as speed, easy com-
right documents and revisions in there, and that’s a long time; munication, transparency, and so on, which were developed based on
again [ : : : ] I gave it to my document control and my docu- their previous experiences of collaboration. When the new tool does
ment control sent it to the other company’s document control, not meet the expected qualities, it reduces their productivity, and users
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

the document control there sends it to whoever the lead is, and move back to the older ways of information sharing to expedite the
the lead then sends it on to whoever’s doing the work–and that task. When users find it difficult to utilize the CDE for structured in-
may take a week or two. And that’s completely wasted time, formation flow, they bypass the workflows to get the work expedited.
and no one in the middle of that process has done any work This leads to the loss of metadata, document trails, and information
[ : : : ] And actually, by the time it gets to the people who are dependencies because these unstructured communication chan-
reviewing the actual technical data, it may be three or four nels offer limited or no codification. In addition, the document
weeks later [ : : : ] In fact, I did it yesterday, I sent a load control workflow itself makes the process slow. Document control
of RFIs through to [designer] informally, five minutes after bottlenecks have multiple implications for data quality. Firstly, the
I’d sent it through my document control. (Principal engineer, value for the timeliness dimension is lowered as the data that is
C3I8) published might not be the latest. Hence, the inferences are based
on old data, which leads to false interpretations. Secondly, this
To bypass this obstacle, workers send the information through lowers the semantic accuracy of the data because its attributes
an unstructured channel in addition to the structured workflows. might no longer be true. This also introduces the problem of con-
This is because of poor understanding of document control work- sistency. Depending on which database employees look at, they
flows amongst the project participants regarding the requirement of see different values. For instance, one CDE to which the informa-
these structured workflows and the CDE. The slow processes and tion was packaged would have the latest value, while the one that
the need for completing the task before deadlines force employees must go through another document controller would have a differ-
not to follow document control workflows. ent value. This tampers with the idea of a “single source of truth.”
In addition, when users circumnavigate the workflow, there are
There is generally a poor understanding of document control
more data quality issues related to unstructured information shar-
requirements, certification requirements [ : : : ] we’re finding
ing such as accessibility and provenance.
that general good practice that people should have brought
with them from other projects is being conveniently put to
one side for the purposes of expediting the work that people Lack of Process Change
are being asked to do. (Information manager, C3I4) Even though structured information flow is digitized through the
introduction of a CDE, the process enabling information flow re-
In addition, the workflows in the CDE are complex and not in- mains unchanged. For example, for a piece of information to be
tuitive, making it difficult for users to follow the protocols for docu- approved, it must be printed, associated with a cover sheet, and
ment control. signed.
Just because of the way they need to store it in certain places We’re actually going to export that out of the CDE, we’re
and stuff like that, and it can’t be done : : : The way that going to print it out, we’re going to staple it together, we’re
[CDE1] is set up here, I believe it is not easy to use (Project going to put our own cover sheet on the front of it, with the
engineer, C3I1) exact same details on the back and we’re going to go off and
go and get three signatures, scan it back in, put it back into
It seems very complex, you open them up, there’s lots of [CDE1] and submit it. (Project engineer, C3I1)
things going on [ : : : ] I just want to know where I can get Printing and scanning the document results in loss of metadata.
my latest drawing (Digital engineer, C3I3) A scanned document in a PDF format has little machine-readable
This has resulted in employees bypassing the workflow, which information in it. Inferring the contents from a scanned document is
leads to system conflicts and further delays in the processes and, at also resource intensive when compared to its original form. In the
times, in the information being stored on the CDE. “I think some- process of printing and scanning, the content becomes digitally
one within the doc management system had obviously circumnav- accessible but not machine readable, thereby limiting the applica-
igated it somehow, to get the drawings out. And then when we tion of data science.
were trying to get the said revisions for our set out, the system What I’m finding now is it’s not actually speeding everything
wouldn’t allow it because directory hadn’t been properly created. up, it’s sort of making everything a lot slower; which I find
(Technical manager, C3I5).” Not following the document control very frustrating (Principal engineer, C3I8)
workflows leads to information loss in the CDE. This is a major
setback to codification because the data is stored in an unstruc- The lack of change in the processes reduces the value in the
tured way that is difficult to access. This problem was found in adoption of a CDE as it increases the time to do these tasks rather

© ASCE 04020072-8 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 than a total reduction in time. In the water project, three CDEs were information is relatively less structured and detailed when com-
used for the project, which created issues such as double handling, pared to product information. Process information is codified as
data inconsistencies, and so on. The presence of multiple CDEs in a Gantt chart models in scheduling software and then linked to
project is another example of the lack of process change. Multiple the BIM model. The detailed process information is not structured
CDEs resembles the paper-based workflows such as document flows into a model. Instead, it is shared as method statements in less-
between the designer and contractor, another set of document flows structured PowerPoint presentations and PDF documents. Sharing
within the contractor’s office, and another set of document flows to information in these unstructured formats has implications for data
the clients for approvals. quality, which are presented in this section.

Our client prefers [CDE1], and we have the designer who Loss of Constraint Information
stores things in [CDE2], so we have both of those tools, In construction, the constraints for any activity execution are dis-
and we have to balance between those two. That can be very cussed during team meetings because the constraints span between
confusing when we have two platforms (Technical man- different teams. For example, logistics constraints span amongst pre-
ager, C3I5). fabrication, logistics, and site teams. These discussions lead to the
removal of constraints by rearranging the start and end times for the
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

For the information submissions to the clients, CDE1 was used; activities. These are then translated to Gantt charts as an output.
for the information from the design consultants to the contractor,
CDE2 was used; and for internal file handling and sharing with the So, from an engineering perspective, we have to interpret en-
contractor, CDE3 was used. CDE1 and CDE3 came from the same gineering information, whether it be drawing or written con-
vendor. However, CDE2 was from a different vendor. straints, written narratives and interpret those into a Gantt
chart. So, we physically need that information to know what
Everything had to be taken out of one data environment and we’re building and what the constraints in building it are.
pushed into another. One of the issues with that is the consis- (Project planner, C3I2).
tency or the compliance or knowing the latest versions of in-
formation (Digital engineer, C3I6) The above statement from the project planner provides evidence
on the processes used to convert the constraints into a Gantt chart
The existence of multiple CDEs within a project introduces the for communication. However, during this process, many of the con-
problem of data inconsistencies. Documents must be taken out of straints themselves, and thus context information for rearranging
one CDE and placed in another. When the volume of information is the activities, are lost. This is because Gantt charts can hold only
huge, with each file having multiple versions, it is difficult to main- precedence constraints. Other constraints, such as disjunctive
tain consistency of documents across multiple CDEs. This means (where activities cannot overlap) and logical constraints, are not
that information in a CDE might not be accurate and up to date, embedded into the Gantt model. Instead, they are retained only
leading to incorrect interpretations when data analytics is per- as tacit information by project participants involved in team meet-
formed on it. ings. This is a case of incomplete information within the dataset
because this information is only accessible to the meeting partic-
As the contractor, then we have to deliver it to a completely
ipants. For example, one of the meetings in the water project had an
separate, disconnected CDE [ : : : ] we’re double-handling
issue with piling, where the pile-driving equipment did not have
(Digital engineer, C3I3)
access to the site for a specific date because there was another ac-
When the CDEs are disconnected, the document trail is lost tivity going on which limited the width of the site access road. At
when a document is moved from one CDE to another, leading the meeting, this was raised:
to the loss of traceability.
Access chamber works will conflict with access road for pile
Lack of process change has multiple implications on data qual-
work, piling work package has to be moved back 2 weeks.
ity as well. Printing and scanning remove metadata and data re-
(Progress review meeting, C3M3)
lationships from the files. A scanned version of the file would also
have very little machine-readable information embedded in it and Here, there is a dependency between access chamber works, and
would require resource-intensive methods to extract insights from the piling work package as the access chamber work would reduce
it. This reduces the data quality dimensions such as accessibility the road width. Therefore, the piling activity was delayed to a later
(i.e., the metadata and data relationships are removed), complete- date. The constraint was removed. However, the knowledge that
ness (information is not complete), and provenance (document there was a constraint is not recorded, and thus the presence of that
trail is lost in the process). In addition, having multiple CDEs in- constraint is not codified. This means such constraints are not ma-
troduces the data quality issue associated with provenance be- chine readable because the access to this information was limited to
cause files residing in multiple CDEs are disconnected, and the the participants of a particular meeting. If an automatic scheduler is
relationships of that particular file with another file are lost in used to reschedule these activities, this rescheduling activity would
the transition process. There are further issues with synchroniza- not have access to this information, resulting in an unrealistic
tion of the information when it is distributed in multiple CDEs. schedule.
This introduces the data quality issues associated with consis- Similar issues were observed in all cases. This issue introduces
tency, which limits the quality of findings made by inference data quality problems associated with accessibility (information is
algorithms. limited to people who attended the meeting) and data completeness
(the model does not include any constraint information; hence, it is
Construction Process Information incomplete).

This section presents the codification challenges related to con- Low Level of Detail
struction process information. The data analysis shows that product The precedence information codified into Gantt charts is linked
information is relatively well structured as BIM models, analysis with BIM to create 4D BIM simulations. However, the lack of
models, and CAD drawings. However, construction process detail in work packaging and associated information (such as

© ASCE 04020072-9 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Deck is reported as completed

Supporting pier is yet to be completed

Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Visual representation showing the effect of low level of detail in process information. Deck is completed before the completion of the
supporting pier.

constraints and resources) results in misinterpretation from the 4D understanding of codification challenges. Furthermore, building on
models. The metro project follows a 5D BIM workflow (digital and extending Batini and Scannapieco (2016), it shows how these
project management PPT—C2D2), where the schedule is linked codification challenges are then mapped to their data quality di-
to a BIM model, bill of quantities (BOQ), and an enterprise re- mensions, such as accuracy, completeness, timeliness, consistency,
source planning (ERP) system to compare the cost based on the accessibility, and data provenance.
quantities versus the cost stated in the work orders from the sub-
contractors. The progress information is also linked to this model to
ensure that the work is done before sanctioning the bills for the Software Usage
work orders. The findings on software usage show that, despite significant digi-
tization of work processes, data remains fragmented into different
Work package for three spans were linked to a work order.
domains and formats because of the multiple software tools in use
Model showed the deck for a span was completed before the
across the organizations involved in construction. Work in the con-
pier supporting it was completed because the work package
struction information technology community is pioneering new
for the first span was reported as completed. (Field notes, BIM
data management solutions to improve interoperability (Hu et al.
Consultant 2, C2I5)
2016; Pauwels et al. 2010; Pazlar and Turk 2008; Redmond et al.
The deck of the metro can be completed only when the pier 2012), and it is disappointing to find that construction projects still
supporting it is completed, as shown in Fig. 1. However, the system suffer from poor quality data as a result of problems of interoper-
recorded the deck assembly to be completed when the pier was not ability caused by the existence of multiple domain-specific tools
completed. This was because work packages for the deck and pier and modeling practices. In their work, Dossick and Neff (2010)
were different because they were done by different subcontractors, have previously shown how the organizational and cultural divi-
and the level of detail of the work package is low. A subcontractor sions between the designers and builders, contractors, and subcon-
who dealt with deck assembly had a part of the work package com- tractors stifle collaborative work. This paper shows these issues are
pleted, but the lack of detail in work packaging triggered the com- not resolved. In the projects studied, organizational and cultural
puter to record the whole work package as completed, resulting in divisions between the firms involved in the late design and construc-
the error. This is a clear case of lack of detail in the model leading to tion stages of projects cause software usage problems (interoperability,
wrong inferencing. information loss during format conversion, and multiple modeling
Similar issues were observed in the student apartment (Case 1) techniques). While there may be shared norms and tools within a
by examining the schedule data in the construction program firm for modeling information, these norms differ across the firms
(C1D1) and the water project (Case 3) by examining the construc- that modeled project information. Multiple modeling techniques
tion program update (C3D4) and observing the review meeting and data created using different software result in datasets that are
(C3M3). These issues are caused by low data quality due to incom- not interoperable and require format conversions, resulting in loss
plete information related to the completeness dimension. of information and low machine readability. The water project
(Case 3) had multiple firms working on the data over different
phases of the project, with interoperability problems more prevalent
Discussion in this case in comparison with the student apartment (Case 1), in
which a single leading firm was involved with the creation and use
This section discusses the software usage, information sharing, and of the model. Although the metro project (Case 2) had different
construction process information codification challenges that limit firms over the different phases of the project, digital data creation
the uptake of data science in construction, drawing on the evidence was handled through a single owner support organization, limiting
from the empirical study. The discussion relates the findings to the the impact of this problem. Software usage problems are found to
literature on BIM use in practice (e.g., Dossick and Neff 2010; lead to challenges of codification for data science; hence, this work
Harty and Whyte 2010) and other strands of research on data qual- extends prior insights by Dossick and Neff (2010) to show how
ity, machine readability, and BIM adoption and implementation organizational and cultural divisions between designers and build-
to articulate how these new analyses contribute by extending ers not only stifle collaborative work and joint problem-solving but

© ASCE 04020072-10 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 also result in fragmented datasets in construction, leading to data Construction Process Information
silos and data loss and, thus, resulting in poor data quality which is
Regarding construction process information, this paper shows that
more difficult to use in data science. the process codification is limited to the master planning or phase
As it is relatively unusual and potentially undesirable to have
planning level and lacks the level of detail and linkage required for
one firm or owner with overall control of the model, to enable more the application of data science tools. While product information is
distributed working, developers of new tools or digitally enabled relatively well codified in BIM, process information is less well
processes should consider the implication of organizational sepa- detailed, and there is a lack of constraint information. These chal-
ration in the sector in addition to the technical requirements. In their lenges are identified by researchers implementing 4D BIM. For ex-
work, Dossick and Neff (2010) have described the influence of ample, Han and Golparvar-Fard (2017) have stated that the process
strong leadership to hold people together on a project to improve modeling methods fail to document field issues to be made avail-
collaboration despite professional segregation. Similarly, a set of able for further analysis; for instance, the 4D BIM’s “Model Break-
common practices and a larger vision for the data creation and man- down Structure typically does not match operational details or
agement should be laid out in the project to ensure the data meets require creating complicated namespaces which, without visual
the necessary quality to enable its use without loss of information in representations, are difficult to communicate” (p. 1733). Giretti
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

between. To achieve better-quality data in projects, practitioners et al. (2012) have further reported the lack of correlation between
must focus beyond the individual scope of their multiple firms to- the resources employed hourly and work progress. This led to the
wards the common goals of the project. decomposition of tasks into subtasks to determine causal relation-
ships between the involved variables so the whole progress could
Information Sharing be determined. This study suggests that to overcome such reported
issues, the methods for codifying construction process information
The analyses suggest that the construction sector has not yet made must be more detailed. The institutionalized practice of planning
the transition from document-based to model-based ways of organ- being limited to master planning and phase planning, without the
izing digital data. The use of drawings and file-based sharing, focus on granular planning such as look-ahead planning and
unstructured information sharing, printing and scanning of docu- weekly planning, is causing this codification challenge, with the
ments, multiple CDEs, and so forth in information sharing has a lack of semantic relationships embedded in the model limiting the
significant impact on the machine readability of data. Paper-based application of automatic schedulers. These issues suggest both a
practices are institutionalized in the sector, and while they are being change in the modeling of process information in construction, with
replaced by digital ways of working, this change is slow, with users the need to develop tools that support modeling of complex con-
of construction information still conditioned to work with drawings straint information, and also a change in the practice to codify the
and PDFs and unstructured information sharing. Even in projects process information in greater detail so that data science could be
that are championing newer BIM-based workflows using CDEs, employed to augment decision making in construction.
this work finds it is difficult to replace these practices, as evidenced
by the problems associated with information sharing (i.e., the sec-
tion “Information Sharing”). The complexity and long processing Machine Readability of Construction Datasets
times involved in these workflows force users to shift back to This section discusses the machine readability of the construction
existing practices and workarounds to expedite their work. The datasets. Common construction datasets are classified based on the
findings from this paper also support the previous characterization set of rules for creating structured data described by Berners-Lee
of users in construction combining new structured methods of (2006) in Table 4.
information sharing along with the prior practice of unstructured Most of the construction information observed from the cases
information sharing when they were hindered by bottlenecks in satisfies the requirement for a 1-star category. The observed proj-
processes, such as document control. Thus, we can characterize the ects use a CDE for storing and managing project data, resulting in
project participants use a range of new and existing practices to- indexing the data and storing it on online servers, resulting in 1-star
gether as “hybrid practices” (Harty and Whyte 2010), and their data. CDE makes the data easier to retrieve for the computers to
circumnavigation of workflows results in unstructured information make inferences on them. However, the complexity of the new
sharing [as shown in a previous discussion in Whyte et al. (2016)]. structured methods for information sharing using CDE, and the
However, this paper goes beyond such studies to characterize poor understanding of workflows across the teams, leads to the
the implications for data quality and to highlight, building on use of a combination of structured and unstructured channels for
Hartmann (2008), the potential to develop newer workflows and information sharing, as discussed in the section “Information Shar-
digitally enabled processes which address the challenges faced by ing.” This aspect reduces the machine readability of the information
practitioners. distributed over unstructured channels because the information is

Table 4. Construction data sets categorized based on levels of machine readability

Quality Principles for publishing
of data a machine-readable data set Construction data sets
1-star Data is available on the web Files and models uploaded in the common data environment
2-star 1-star data structured in a proprietary format BIM files in proprietary formats (Revit files, Microstation files, etc.), project management
information (Asta power project, Primavera P6, Microsoft project), design rationale, and
associated information (in Microsoft Excel), etc.
3-star 1-star data structured in a non-proprietary format BIM files in IFC format, CSV data etc.
4-star 3-star data that is published using open standards BIM files published using open standards such as ifcOWL, BOT ontology etc.
5-star 4-star data with links to other 4 star datasets BIM files published using open standards linked to other such files (BIM files, GIS data, etc.).

© ASCE 04020072-11 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Table 5. Mapping between findings and data quality dimensions
Data quality
dimension Codes and data
Accuracy Multiple modelling techniques: “use the wrong tool to model something [ : : : ] I can’t just say there’s a slab now, that’s just a
piece of geometry” (Digital engineer, C3I3); “when you try to extract 2D drawings from 3D BIM models, those drawings are
not as correct and as detailed as they used to be” (Technical manager, C3I5)
Document control bottlenecks: “it’s no longer the most current version anymore by the time I’m reviewing it” (Project
engineer, C3I1) “he keeps on updating but he hasn’t he hasn’t put it on the [CDE1].” (Technical manager, C3I5) “I just
want to know where I can get my latest drawing” (Digital engineer, C3I3) “I think someone within the doc management
system had obviously circumnavigated it somehow, to get the drawings out. And then when we were trying to get the
said revisions for our set out, the system wouldn’t allow it because directory hadn’t been properly created.” (Technical
manager, C3I5)
Completeness Interoperability: “transferring things [ : : : ], you lose data” (Technical manager, C3I5) Information loss during conversion:
“when you upload a PDF.” (Information manager, C3I4) Multiple modelling techniques: “use the wrong tool to model
something [ : : : ] I can’t just say there’s a slab now, that’s just a piece of geometry” (Digital engineer, C3I3); “when you try to
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

extract 2D drawings from 3D BIM models, those drawings are not as correct and as detailed as they used to be” (Technical
manager, C3I5)
Lack of process change: “We’re going to print it out, we’re going to staple it together [ : : : ] get three signatures, scan it back in,
put it back into [CDE1] and submit it.”(Project engineer, C3I1) “it’s not actually speeding everything up, it’s sort of making
everything a lot slower; which I find very frustrating” (Principal engineer, C3I8) “can be very confusing when we have two
platforms” (Technical manager, C3I5) “Everything had to be taken out of one data environment and pushed into another. One
of the issues with that is the consistency or the compliance or knowing the latest versions of information” (Digital engineer,
C3I6) “As the contractor, then we have to deliver it to a completely separate, disconnected CDE [ : : : ] we’re double-handling”
(Digital engineer, C3I3)
Loss of constraint information: “we physically need that information to know what we’re building and what the constraints in
building it are.” (Project planner, C3I2). “Access chamber works will conflict with access road for pile work, piling work
package has to be moved back 2 weeks.” (Progress review meeting, C3M3); Low level of detail: “Work package for three
spans were linked to a work order. Model showed the deck for a span was completed before the pier supporting it was
completed because the work package for the first span was reported as completed.” (Field notes- BIM Consultant 2, C2I5)
Timeliness Unstructured information sharing: “when I have finished everything-by the way, we have this spread sheet” (Technical
manager, C3I5); Document control bottlenecks: “it’s no longer the most current version anymore by the time I’m reviewing it”
(Project engineer, C3I1) “he keeps on updating but he hasn’t he hasn’t put it on the [CDE1].” (Technical manager, C3I5) “I just
want to know where I can get my latest drawing” (Digital engineer, C3I3) “I think someone within the doc management
system had obviously circumnavigated it somehow, to get the drawings out. And then when we were trying to get the said
revisions for our set out, the system wouldn’t allow it because directory hadn’t been properly created.” (Technical manager,
C3I5)
Lack of process change: “We’re going to print it out, we’re going to staple it together [ : : : ] get three signatures, scan it back in,
put it back into [CDE1] and submit it.”(Project engineer, C3I1) “it’s not actually speeding everything up, it’s sort of making
everything a lot slower; which I find very frustrating” (Principal engineer, C3I8) “can be very confusing when we have two
platforms” (Technical manager, C3I5) “Everything had to be taken out of one data environment and pushed into another. One
of the issues with that is the consistency or the compliance or knowing the latest versions of information” (Digital engineer,
C3I6) “As the contractor, then we have to deliver it to a completely separate, disconnected CDE [ : : : ] we’re double-handling”
(Digital engineer, C3I3)
Consistency Document control bottlenecks: “it’s no longer the most current version anymore by the time I’m reviewing it” (Project
engineer, C3I1) “he keeps on updating but he hasn’t he hasn’t put it on the [CDE1].” (Technical manager, C3I5) “I just want to
know where I can get my latest drawing” (Digital engineer, C3I3) “I think someone within the doc management system had
obviously circumnavigated it somehow, to get the drawings out. And then when we were trying to get the said revisions for our
set out, the system wouldn’t allow it because directory hadn’t been properly created.” (Technical manager, C3I5)
Accessibility Interoperability: “transferring things [ : : : ], you lose data” (Technical manager, C3I5); Unstructured information sharing:
“when I have finished everything-by the way, we have this spread sheet” (Technical manager, C3I5);
Drawings and file-based sharing: “I use all the navigator tools that we’ve got here. But I prefer to use AutoCAD because I find
it a lot easier” (Project engineer, C3I1) “It’s not the best way because we haven’t got the technology. I haven’t got a big screen”
(Project engineer, C3I1)
Lack of process change: “We’re going to print it out, we’re going to staple it together [ : : : ] get three signatures, scan it back in,
put it back into [CDE1] and submit it.” (Project engineer, C3I1) “it’s not actually speeding everything up, it’s sort of making
everything a lot slower; which I find very frustrating” (Principal engineer, C3I8) “can be very confusing when we have two
platforms” (Technical manager, C3I5) “Everything had to be taken out of one data environment and pushed into another. One
of the issues with that is the consistency or the compliance or knowing the latest versions of information” (Digital engineer,
C3I6) “As the contractor, then we have to deliver it to a completely separate, disconnected CDE [ : : : ] we’re double-handling”
(Digital engineer, C3I3)
Loss of constraint information: “we physically need that information to know what we’re building and what the constraints in
building it are.” (Project planner, C3I2). “Access chamber works will conflict with access road for pile work, piling work
package has to be moved back 2 weeks.” (Progress review meeting, C3M3)
Low level of detail: “Work package for three spans were linked to a work order. Model showed the deck for a span was
completed before the pier supporting it was completed because the work package for the first span was reported as completed.”
(Field notes- BIM Consultant 2, C2I5)

© ASCE 04020072-12 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Table 5. (Continued.)
Data quality
dimension Codes and data
Data provenance Information loss during conversion: “when you upload a PDF.” (Information manager, C3I4); Drawings and file-based
sharing: “I use all the navigator tools that we’ve got here. But I prefer to use AutoCAD because I find it a lot easier” (Project
engineer, C3I1) “It’s not the best way because we haven’t got the technology. I haven’t got a big screen” (Project engineer,
C3I1)
Lack of process change: “We’re going to print it out, we’re going to staple it together [ : : : ] get three signatures, scan it back in,
put it back into [CDE1] and submit it.”(Project engineer, C3I1) “it’s not actually speeding everything up, it’s sort of making
everything a lot slower; which I find very frustrating” (Principal engineer, C3I8) “can be very confusing when we have two
platforms” (Technical manager, C3I5) “Everything had to be taken out of one data environment and pushed into another. One
of the issues with that is the consistency or the compliance or knowing the latest versions of information” (Digital engineer,
C3I6) “As the contractor, then we have to deliver it to a completely separate, disconnected CDE [ : : : ] we’re double-handling”
(Digital engineer, C3I3)
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

not indexed nor available on a common server. The same issue oc- to another. Storing the information in proprietary formats also re-
curs when the users circumnavigate the workflows to get the work duces the accessibility dimension for data quality (Table 5).
expedited. Similarly, codification challenges associated with con-
struction process information also lower the machine readability
because the information is not recorded (lack of detail and loss of Implications for Data Quality
construction information) and, hence, not indexed or stored in on- The codification challenges discussed earlier have many significant
line servers. These issues make the information inaccessible for implications for data quality (refer to Table 5). To unpack these in
inferencing. this section, they are mapped onto the different quality dimensions.
With regard to the structure of the construction information, Accuracy: The organizational and cultural divisions between
construction datasets in the form of BIM, project management in- different teams results in problems associated with multiple mod-
formation in project management software (Primavera P6, Asta eling techniques, leading to data quality issues concerning the syn-
powerproject, etc.), outputs from Microsoft tools such as Excel, tactic accuracy of the data. This was evident from the dataset when
and so forth, are structured, satisfying the requirements for 2-star different people had different perceptions of the model, as in the
data. However, construction data are also unstructured in the forms case of the preceding slab example. Similarly, the hybrid practices
of PDF documents, drawings, and other file-based formats, as de- associated with information sharing lead to lowering the semantic
scribed in the sections “Drawings and File-Based Sharing” and accuracy of the data because the data with which inferences are
“Lack of Process Change.” The lack of structure in the datasets made are not accurate due to inefficiencies in information sharing.
makes inferencing from them difficult, leading to the need for com- When there are syntactic errors in the data, this leads to incorrect
plex algorithms. Where the construction data is structured, the data
insights (for example, if a slab is modeled as a geometric object
structure is in proprietary formats, which require the APIs to access
with attributes attached to it and when a software tool is used to
the semantic relationships within the data. The observed projects
compute quantity take-off from the model for all the slabs). The
do not use open formats or open standards for publishing the data.
output would be zero as quantity required because the program fails
Proprietary tools for the authorship of construction data are far
to identify the geometric object as a slab. If this software is inte-
more advanced and easier to use than the open-source tools. Hence,
grated with a costing tool used for cash flow analytics, this error
construction projects resort to using the tested and robust propri-
etary tools, resulting in issues associated with interoperability and gets propagated into that tool. These errors can be removed to an
loss of information presented in the section “Software Usage.” extent using semantic enrichment programs. However, even the ac-
Thus, the construction information rarely achieves 3-star classifi- curacy of inferences of semantic enrichment programs is dependent
cation, as mentioned by Berners-Lee (2006). In conclusion, the on the quality of input datasets (Sacks et al. 2017).
maximum level of machine readability of the construction datasets Completeness: Concerning the completeness of the data, organi-
in the observed projects is 2-star, with most of the information with zational and cultural divisions between the teams resulting in prob-
a 1-star rating. lems such as interoperability, format conversions, multiple modeling
Low machine readability of the data has implications on data techniques, and the implication of hybrid practices such as printing
quality. When the construction information is not stored on servers and scanning the documents play a role in reducing the data quality.
due to information sharing issues or lack of detail in the models, the For example, the software usage issues caused by the organiza-
accessibility of that data is affected, thus reducing the accessibility tional divisions leads to format conversion resulting data loss lead-
dimension of the data quality. When information is stored in CDE ing incomplete dataset. Lack of process change also leads to similar
(satisfying conditions for 1-star data) as PDF documents, the struc- problems such as loss of metadata when documents are printed and
ture of the data is not maintained, resulting in data quality issues scanned. The institutionalized practices of process modeling with
associated with syntactic accuracy and consistency. Lack of a data low levels of detail and the practice of not codifying constraints in
structure removes the context from the information, thus resulting the model result in incomplete data. Inferring insights from incom-
in the need for complex algorithms to infer the contexts and infer plete datasets reduces the quality of the output. For example, if the
from data. This issue also introduces problems associated with con- constraints are not codified in a schedule, an automatic scheduler
sistency as the data value for a field might be different in different would create an unrealistic schedule. This leads to further problems
files, and the lack of context limits the computer programs to detect down the line. In the case of a product model, an incomplete dataset
it. This problem is further worsened because the datasets are not used for a structural capacity prediction would give incorrect
linked since the links are lost when a file is moved from one CDE results.

© ASCE 04020072-13 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Timeliness: This dimension of data quality is mostly affected Conclusions and Future Work
by the information-sharing practices in construction using hybrid
practices by using old and new practices simultaneously. File-based Codification challenges for data science in construction are found
sharing (resulting in slowing down the data updates compared to to be related to: (1) software usage, i.e., interoperability, informa-
model-based sharing), document control bottlenecks delaying the tion loss during conversion, and multiple modeling techniques;
submission of data into the system, and the use of multiple CDEs (2) information sharing, i.e., unstructured information sharing,
requiring data transfer from one to another (resulting in a delay in drawing and file-based sharing document control, and lack of pro-
fetching the data) all constitute outdated data-skewing analytics. cess change; and (3) construction process information, i.e., loss of
When outdated data is used for analytics, the resulting inference constraints and low levels of detail. The implication of these chal-
is not time appropriate, and decisions taken with those inferences lenges was discussed by mapping them to data quality dimensions
lead to problems. For example, if the construction plan is made such as accuracy, completeness, timeliness, consistency, and prov-
using resource availability data, during the actual construction date, enance. Through the identification of the codification challenges
the assigned resource might not be available. When this is managed in the late design and construction phase of the projects and their
manually, the planner makes sure this does not happen. However, mapping to the data quality dimensions, this paper extends the
knowledge on data quality issues in construction. It shows how data
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

when performed automatically, it is necessary that the datasets are

updated close to real time. quality arises from organizational as well as technical practices.
Consistency: Hybrid practices in information sharing, resulting The persistence of organizational and cultural divisions, paper-
in document control bottlenecks, multiple CDEs, and circumnav- and document-based (as well as digital- and model-based) ways
igating workflows, induce inconsistency in the data. The data in of working, and institutionalized practices of construction process
modeling are challenges for the uptake of data science as they lead
one CDE might be different from that in another CDE. Similarly,
to the partial codification of information in machine-readable for-
the document control bottlenecks in publishing the data result in
mats. While the fragmented nature of the sector is well understood,
different teams and their databases having different versions of the
this work shows how codification challenges arise because of the
data. This creates a problem during the data analysis phase. For ex-
different digital workflows and working practices across projects,
ample, for the same content, different values might exist, of which
and how these lead to fragmented data as a result of the use of
only one is true. This induces problems of semantic accuracy if the
multiple software packages, poor information sharing, and only
computer takes the incorrect value for analysis.
partially captured construction process information.
Accessibility: This dimension of data quality is affected by
While all of the projects studied were BIM enabled, within these
organizational and cultural divisions between teams, hybrid prac-
projects, information sharing issues and software usage issues
tices, and institutionalized practices in process modeling. Organi-
emerge as a result of their document-based, and sometimes paper-
zational and cultural divisions between teams, resulting in the use
based, practices. Project-wide standards and policies are created to
of multiple software, multiple modeling techniques, and so forth,
streamline information sharing through structured workflows, and
affect the accessibility of data because employees must convert
these workflows are aimed at improving collaboration, but the pro-
the data to access it, and there are instances that show conversions cess of document control raises issues. In addition, this study iden-
result in loss of data. Hence, the data is only accessible in full tifies inefficiencies (such as long processing times, complexity,
content to the team, which created it. Similarly, the hybrid etc.) in these workflows, which pushes the users to bypass the struc-
practices, particularly printing and subsequent scanning of the tured methods in the current workflows. This forces the users to
documents, remove the metadata as well as semantic information revert to older methods or use a combination of old and new meth-
within those data, resulting in a less-efficient analysis. In addition, ods, resulting in the generation of unstructured information. These
the lack of constraint codification restricts the information regard- cause data quality issues relating to inaccessibility of data, timeli-
ing constraints, making it accessible to the few people who at- ness of data, and data consistency.
tended the meeting. It is not embedded in the model and, hence, Limited codification in construction process information is a
not accessible to any analytic algorithms. This reduces the result of the institutionalized practice of scheduling focusing on
capability of such algorithms to infer accurate information and codifying just precedence relationships into a Gantt chart. The ad-
determine causalities. vancements in 4D and 5D BIM have attempted to address this prob-
Provenance: This dimension of the data quality is affected by lem to an extent by integrating the resource and cost information to
both organizational and cultural divisions in the industry as well as a single platform. However, the data herein suggest that the level of
hybrid practices. Fragmentation in the industry leads to many con- detail of the scheduling is still at a macro-planning level. In addi-
versions and much manipulation of data to suit purposes, in the tion, the constraint relationships between resources and processes,
process removing the traceability and origins of the data. Similarly, resources and the site conditions, and site conditions and the proc-
multiple CDEs, printing and scanning of documents and unstruc- esses discussed in weekly meetings are not codified into a model.
tured information sharing also lead to loss of metadata and ends up This limits the data science due to data quality issues associated
removing information regarding origins of the data. This limits al- with completeness and accessibility.
gorithms from making inferences based on data origins and their How might codification challenges be overcome to enable
incremental manipulation. greater uptake of data science in construction? The evidence from
This section described the challenges of codification as causes this paper points towards a need for change in the policy and prac-
of data quality issues, which have implications for different data tice to ensure machine readability and better-quality construction
quality dimensions and the output of data analytics. While big data information. The current policies on information modeling are
techniques suggest future opportunities to use data science with an aimed at improving collaboration amongst the project participants.
increasing variety of data of different levels of veracity, data clean- Building on the existing policies, newer policies should ensure
ing is a resource-intensive process, and significant training models that the machine readability and quality of the data are maintained
are required. To improve the uptake of data science in construction, during information sharing. Care should be taken to ensure that
high-quality data is necessary, and achieving this requires overcom- newer policies suit the existing work practices in construction so
ing the codification challenges identified here. that the workflows recommended by such policies are not bypassed

© ASCE 04020072-14 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 by the users. Newer policies also ensure the transition from file- and modeling different constraints within the software environ-
based information sharing to model-based information sharing, ment to capture it from the discussions.
where the model data is shared based on the principles stated by The current study has identified the codification challenges by
Berners-Lee (2006) to ensure machine readability. Standards under examining information sharing in the late design and construction
development, such as ISO/DIS 21597—information container for phase and mapped it to the data quality dimensions. The findings
data drop, exchange specification parts 1 and 2—are moving in this from this study inform the researchers, who are developing frame-
direction, where linked data technologies are used to define the re- works and methods to codify construction information, of the
lationships between documents and datasets, thereby automating organizational issues to be considered in their work. This paper also
the document control processes (ISO 2019). These standards would provides the data quality implications of issues associated with
have a positive impact on data quality pertaining to dimensions, BIM implementation, motivating the researchers focusing on im-
timeliness, accessibility, and consistency. Even though policy inter- plementation studies to widen their scope from collaboration to
ventions would have a positive effect on data quality, the construc- include data quality and machine readability. Future research can
tion sector is known for institutionalized practices and resistance to build on this study to develop recommendations for ensuring the
change. Thus, policies must influence practices to enable greater machine readability of construction information generated through-
uptake of data science.
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

out different stages of the lifecycle of the project. To achieve this,

Thus, in addition to the policies, it is, therefore, necessary to future research should focus on different aspects. Firstly, research-
adopt newer methods in practice to make construction informa- ers should elaborate on the organizational issues identified in the
tion machine readable. This study examined a student apartment
current study across different phases of the project and amongst
building project constructed by a leading contractor in the United
different stakeholders involved in the project. Secondly, accounting
Kingdom using state-of-the-art offsite manufacturing techniques
for the complexity of information-use trends in construction, there
(Case 1), a metro project pioneering digital transformation in
is a need for research in fundamental data science to pave the way
India (Case 2), and an innovative water megaproject in the United
for the integration of the disconnected information in the construc-
Kingdom (Case 3). Despite these projects having advanced work-
flows and innovative approaches to ensure structured informa- tion sector. This includes developing newer information modeling
tion, lack of process change after digitization of workflows was approaches that adapt to the current work processes as well as sup-
evident in the datasets, which included multiple CDEs, printing port codification of information, such as algorithms for crawling
and scanning of documents, need for wet signatures, etc. These through the disconnected information to draw insights and learning
practices continue to occur due to the lack of trust in digital work- algorithms to detect discrepancies in data (such as with accuracy,
flows; thus, there must be a change in this mindset, and trust in completeness, timeliness, consistency, and provenance) and predict
digital data must be developed in the construction practice. There their consequences. Finally, researchers and managers of construc-
is also a need for a shift to model-based information sharing from tion projects should work together towards developing workflows
the file-based information sharing between teams. To address this and information sharing practices that ensure the machine readabil-
issue, related to a lack of detail in process information, there is a ity of construction information while considering the issues of frag-
need for change in the look-ahead scheduling practice to incor- mentation, hybrid practices, and institutionalized practices. This
porate the codification of different constraint relationships into paper has provided a foundation for such future research by extend-
information. This might involve a significant change in the plan- ing the knowledge on data quality issues in construction through
ning of a progress review process by integrating BIM-based the identification of codification challenges, considering the wider
scheduling tools within the planning and progress meetings practice of model and document-based information sharing.

Appendix. Data Collection Record

Sources Source ID Details Time in the field; documentation/research records; and/or dates
Case 1: Multistory residential student apartment block
Document C1D1 Construction program 3 years of planning data
C1D2 BIM model 1.5 gigabytes
C1D3 BIM training document 5 pages
C1D4 Access to common data 2 weeks of access
environment
Interviews C1I1 Digital engineer (August 15, 2017) 2 h, Field note (3 pages)
C1I2 Planner (August 15, 2017) 1 h, Field note (2 pages)
Office visit C1S1 Contractor’s office Multiple visits over 2 weeks, Field notes (7 pages) August 14,
2017–August 18, 2017
Case 2: Metro rail project
Documents C2D1 BIM execution plan 17 Pages
C2D2 Digital project management (PPT) 26 slides
C2D3 Employee information 32 pages
requirements
Informal interviews C2I1 Project manager October 1, 2018, 30 min, Field notes (4 pages)
C2I2 Chief site engineer October 1, 2018, 30 min, Field notes (3 pages)
C2I3 Casting yard engineer October 1, 2018, 20 min, Field notes (3 pages)
C2I4 BIM consultant 1 November 1, 2018, 20 min, Field notes (2 pages)
C2I5 BIM consultant 2 November 1, 2018, 20 min, Field notes (2 pages)
C2I6 BIM consultant 3 November 1, 2018, 20 min, Field notes (2 pages)

© ASCE 04020072-15 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Appendix. (Continued.)
Sources Source ID Details Time in the field; documentation/research records; and/or dates
Site visit C2S1 Casting yard October 1, 2018, 1 h Field notes (5 pages); 22 Photographs
C2S2 Station site October 1, 2018, 2 h
C2S3 Viaduct site October 1, 2018, 1.5 h
Workshop C2W1 Workshop with 6 metros, Nagpur, August 3, 2018, 8 h, 22 Photographs; 6 flipcharts; written summary
Ahmedabad, Kochi, Chennai, (30 pages)
Bombay, Hyderabad, Delhi.
Case 3: Complex water project
Set-up meeting C3M1 Introducing the study May 31, 2018, 1 h, 20 slides
Documents C3D1 Design workflow 2 pages
C3D2 Access to CDE 1 2 months
C3D3 Summary construction program 4 year data
C3D4 Construction program update 1 month data
C3D5 Work package plan 45 pages
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

Semistructured interviews C3I1 Project engineer July 9, 2018 1 h, taped and transcribed (T&T) (10 pages)
C3I2 Project planner lead July 10, 2018, 1 h, T&T (10 pages)
C3I3 Digital engineer lead July 11, 2018, 1 h, T&T (11 pages)
C3I4 Information manager July 11, 2018, 1 h, T&T (15 pages)
C3I5 Technical manager/DfMA design July 12, 2018, 1 h, T&T (15 pages)
lead
C3I6 Senior digital engineer July 12, 2018, 1 h, T&T (8 pages)
C3I7 Design manager July 13, 2018, 1 h, Fieldnotes (4 pages)
C3I8 Principal engineer July 17, 2018, 1 h, T&T (12 pages)
Office visit C3S1 Contractor’s office July 5, 2018–July 19, 2018, multiple visits over 2 weeks, Fieldnotes
(5 pages)
Meetings C3M2 Temporary works design meeting July 13, 2018, 2 h, Fieldnotes (3 pages)
C3M3 Progress review meeting July 16, 2018, 4 h, Fieldnotes (4 pages)

Data Availability Statement Berners-Lee, T. 2006. “Linked data—Design issues.” Accessed March 19,
2019. https://www.w3.org/DesignIssues/LinkedData.html.
Bilal, M., L. O. Oyedele, J. Qadir, K. Munir, S. O. Ajayi, O. O. Akinade, H. A.
Data generated or analyzed during the study are available from the
Owolabi, H. A. Alaka, and M. Pasha. 2016. “Big data in the construction
corresponding author by request. Information about the Journal’s
industry: A review of present status, opportunities, and future trends.” Adv.
data-sharing policy can be found here: http://ascelibrary.org/doi/10 Eng. Inf. 30 (3): 500–521. https://doi.org/10.1016/j.aei.2016.07.001.
.1061/(ASCE)CO.1943-7862.0001263. Bolton, A., et al. 2018. The Gemini principles: Guiding values for
the national digital twin and information management framework.
Cambridge, UK: Centre for Digital Built Britain and Digital Frame-
Acknowledgments work Task Group.
BSI (British Standards Institution). 2018. “ISO19650—1 & 2: Organization
The authors are grateful to the research participants in the three and digitization of information about buildings and civil engineering
case studies. The Ph.D. research of the first author is cofunded by works, including building information modelling (BIM). Information man-
Bentley Systems UK and through a Skempton Scholarship from the agement using building information modelling.” Accessed March 19,
Department of Civil and Environmental Engineering, Imperial Col- 2019. https://bsol.bsigroup.com/Bibliographic/BibliographicInfoData
lege, London. During the development of this paper, this author /000000000030333754.
was supported by the Ph.D. enrichment scholarship from the Alan Cai, L., and Y. Zhu. 2015. “The challenges of data quality and data quality
assessment in the big data era.” Data Sci. J. 14 (2): 1–10. https://doi.org
Turing Institute, the United Kingdom’s National Institute for Data
/10.5334/dsj-2015-002.
Science and AI. The second author acknowledges the support of
Cao, D., H. Li, and G. Wang. 2014. “Impacts of isomorphic pressures on
Laing O’Rourke and the Royal Academy of Engineering for co- BIM adoption in construction projects.” J. Constr. Eng. Manage.
sponsoring her Professorship; and Lloyds Register Foundation/ 140 (12): 04014056. https://doi.org/10.1061/(ASCE)CO.1943-7862
ATI Data Centric Engineering Programme. .0000903.
Carrillo, P., J. Harding, and A. Choudhary. 2011. “Knowledge discovery
from post-project reviews.” Constr. Manage. Econ. 29 (7): 713–723.
References https://doi.org/10.1080/01446193.2011.588953.
Chang, C.-Y., W. Pan, and R. Howard. 2017. “Impact of building informa-
Akintola, A., S. Venkatachalam, and D. Root. 2017. “New BIM roles’ le- tion modeling implementation on the acceptance of integrated delivery
gitimacy and changing power dynamics on BIM-enabled projects.” systems: Structural equation modeling analysis.” J. Constr. Eng. Man-
J. Constr. Eng. Manage. 143 (9): 04017066. https://doi.org/10.1061 age. 143 (8): 04017044. https://doi.org/10.1061/(ASCE)CO.1943-7862
/(ASCE)CO.1943-7862.0001366. .0001335.
Batini, C., and M. Scannapieco. 2016. Data and information quality. Chegu Badrinath, A., and S.-H. Hsieh. 2019. “Empirical approach to identify
Cham, Switzerland: Springer. operational critical success factors for BIM projects.” J. Constr. Eng.
Beetz, J., V. L. L. Berlo, D. R. Laat, and V. D. P. Helm. 2010. Manage. 145 (3): 04018140. https://doi.org/10.1061/(ASCE)CO.1943
“BIMSERVER.ORG—An open source IFC model server.” In Proc., -7862.0001607.
27th Int. Conf. CIB W78 2010. Cairo, Egypt: International council Choi, B., H.-S. Lee, M. Park, Y. K. Cho, and H. Kim. 2014. “Framework
of Research and Innovation in Building and Construction. for work-space planning using four-dimensional BIM in construction

© ASCE 04020072-16 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 projects.” J. Constr. Eng. Manage. 140 (9): 04014041. https://doi.org Kassem, M., and B. Succar. 2017. “Macro BIM adoption: Comparative
/10.1061/(ASCE)CO.1943-7862.0000885. market analysis.” Autom. Constr. 81 (Sep): 286–299. https://doi.org/10
Delone, W. H., and E. R. McLean. 2003. “The DeLone and McLean model .1016/j.autcon.2017.04.005.
of information systems success: A ten-year update.” J. Manage. Inf. Kim, K., H. Kim, W. Kim, C. Kim, J. Kim, and J. Yu. 2018. “Integration of
Syst. 19 (4): 9–30. https://doi.org/10.1080/07421222.2003.11045748. IFC objects and facility management work information using semantic
Donato, V., M. Lo Turco, and M. M. Bocconcino. 2017. “BIM-QA/QC in web.” Autom. Constr. 87 (Mar): 173–187. https://doi.org/10.1016/j
the architectural design process.” Archit. Eng. Des. Manage. 14 (3): .autcon.2017.12.019.
239–254. https://doi.org/10.1080/17452007.2017.1370995. Krijnen, T., and J. Beetz. 2017. “An IFC schema extension and binary seri-
Dossick, C. S., and G. Neff. 2010. “Organizational divisions in BIM- alization format to efficiently integrate point cloud data into building
enabled commercial construction.” J. Constr. Eng. Manage. 136 (4): models.” Adv. Eng. Inf. 33 (Aug): 473–490. https://doi.org/10.1016/j
459–467. https://doi.org/10.1061/(ASCE)CO.1943-7862.0000109. .aei.2017.03.008.
Eastman, C., P. Teicholz, R. Sacks, and K. Liston. 2008. BIM handbook: A Kwon, O. S., C. S. Park, and C. R. Lim. 2014. “A defect management sys-
guide to building information modeling for owners, managers, design- tem for reinforced concrete work utilizing BIM, image-matching and
ers, engineers, and contractors. Hoboken, NJ: Wiley. augmented reality.” Autom. Constr. 46 (Oct): 74–81. https://doi.org/10
Eisenhardt, K. M., and M. E. Graebner. 2007. “Theory building from .1016/j.autcon.2014.05.005.
cases opportunities and challenges.” Acad. Manage. J. 50 (1): 25–32. Lee, Y.-C., C. M. Eastman, and W. Solihin. 2018. “Logic for ensuring the
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

Fan, W. 2015. “Data quality: From theory to practice.” ACM SIGMOD Rec. data exchange integrity of building information models.” Autom.
44 (3): 7–18. https://doi.org/10.1145/2854006.2854008. Constr. 85 (Jan): 249–262. https://doi.org/10.1016/j.autcon.2017
Farias, T. M. D., A. Roxin, and C. Nicolle. 2018. “A rule-based methodology .08.010.
to extract building model views.” Autom. Constr. 92 (Aug): 214–229. Loshin, D. 2010. The practitioner’s guide to data quality improvement.
https://doi.org/10.1016/j.autcon.2018.03.035. Burlington, MA: Morgan Kaufmann.
Giretti, A., A. Carbonari, G. Novembri, and F. Robuffo. 2012. “Estimation Mirarchi, C., and A. Pavan. 2019. “Building information models are dirty.”
of job-site work progress through on-site monitoring.” In Proc., 9th In Proc., 2019 European Conf. of Computing in Construction (2019
Int. Symp. of Automation and Robotics in Construction, ISARC EC3 ). Crete, Greece: European Council for Computing in Construction.
2012. Eindhoven, Netherlands: International Association for Automa- Naumann, F., and C. Rolker. 2000. “Assessment methods for information
tion and Robotics in Construction. quality criteria.” In Proc., Fifth Conf. on Information Quality (IQ 2000).
Goedert, J. D., and P. Meadati. 2008. “Integrating construction process Cambridge, MA: MIT Sloan School of Management.
documentation into building information modeling.” J. Constr. Eng. Oti, A. H., J. H. M. Tah, and F. H. Abanda. 2018. “Integration of lessons
Manage. 134 (7): 509–516. https://doi.org/10.1061/(ASCE)0733-9364 learned knowledge in building information modeling.” J. Constr. Eng.
(2008)134:7(509). Manage. 144 (9): 04018081. https://doi.org/10.1061/(ASCE)CO.1943
Gu, N., and K. London. 2010. “Understanding and facilitating BIM adop- -7862.0001537.
tion in the AEC industry.” Autom. Constr. 19 (8): 988–999. https://doi Pauwels, P., R. De Meyer, J. Van Campenhout, R. D. Meyer, and J. V.
.org/10.1016/j.autcon.2010.09.002. Campenhout. 2010. “Interoperability for the design and construction
Han, K. K., and M. Golparvar-Fard. 2017. “Potential of big visual data and industry through semantic web technology.” Lect. Notes Comput.
building information modeling for construction performance analytics: Sci. 6725 (1): 143–158.
An exploratory study.” Autom. Constr. 73 (Jan): 184–198. https://doi Pauwels, P., S. Törmä, J. Beetz, M. Weise, and T. Liebich. 2015. “Linked
.org/10.1016/j.autcon.2016.11.004. data in architecture and construction.” Autom. Constr. 57 (Sep): 175–177.
Hartmann, T. 2008. “A grassroots model of decision support system impli- https://doi.org/10.1016/j.autcon.2015.06.007.
cations by construction project teams.” Ph.D. dissertation, Dept. of Civil Pazlar, T., and Ž. Turk. 2008. “Interoperability in practice: Geometric
and Environmental Engineering, Stanford Univ. data exchange using the IFC standard.” J. Inf. Technol. Constr. 13 (1):
Harty, C., and J. Whyte. 2010. “Emerging hybrid practices in construction 362–380.
design work: Role of mixed media.” J. Constr. Eng. Manage. 136 (4): Pedro, A., D. Y. Lee, R. Hussain, and C. S. Park. 2017. “Linked data
468–476. https://doi.org/10.1061/(ASCE)CO.1943-7862.0000146. system for sharing construction safety information.” In Proc., 34th
Hendler, J., and T. A. Pardo. 2012. “A primer on machine readability for Int. Symp. on Automation and Robotics in Construction (ISARC
online documents and data.” Accessed October 10, 2019. https://www 2017). Taipei, Taiwan: The International Association for Automation
.data.gov/developers/blog/primer-machine-readability-online-documents and Robotics in Construction.
-and-data. Preidel, C., A. Borrmann, C. Oberender, and M. Tretheway. 2016. “Seam-
Hu, Z.-Z., X.-Y. Zhang, H.-W. Wang, and M. Kassem. 2016. “Improving less integration of common data environment access into BIM author-
interoperability between architectural and structural design models: ing applications: The BIM integration framework." In Proc., 11th
An industry foundation classes-based approach with web-based tools.” European Conf. on Product and Process Modelling (ECPPM 2016),
Autom. Constr. 66 (Jun): 29–42. https://doi.org/10.1016/j.autcon.2016 119. Limassol, Cyprus: European Association of Product and Process
.02.001. Modeling.
Hwang, B.-G., X. Zhao, and K. W. Yang. 2019. “Effect of BIM on rework Quintero, D., W. Genovese, K. Kim, M. Li, F. Martins, A. Nainwal, D.
in construction projects in Singapore: Status Quo, magnitude, impact, Smolej, M. Tabinowski, and A. Tiwary. 2015. IBM software defined
and strategies.” J. Constr. Eng. Manage. 145 (2): 04018125. https://doi environment. IBM Redbooks.
.org/10.1061/(ASCE)CO.1943-7862.0001600. Redmond, A., A. Hore, M. Alshawi, and R. West. 2012. “Exploring
ISO. 2019. “ISO/DIS 21597 Information container for data drop— how information exchanges can be enhanced through cloud BIM.”
Exchange specification.” Accessed March 19, 2019. https://www.iso Autom. Constr. 24 (Jul): 175–183. https://doi.org/10.1016/j.autcon
.org/standard/74389.html. .2012.02.003.
Janssen, M., H. van der Voort, and A. Wahyudi. 2017. “Factors influencing Sacks, R., C. Eastman, G. Lee, and P. Teicholz. 2018. BIM handbook:
big data decision-making quality.” J. Bus. Res. 70 (1): 338–345. https:// A guide to building information modeling for owners, designers, engi-
doi.org/10.1016/j.jbusres.2016.08.007. neers, contractors, and facility managers. Hoboken, NJ: Wiley.
Jaradat, S., J. Whyte, and R. Luck. 2013. “Professionalism in digitally Sacks, R., L. Ma, R. Yosef, A. Borrmann, S. Daum, and U. Kattel. 2017.
mediated project work.” Build. Res. Inf. 41 (1): 51–59. https://doi “Semantic enrichment for building information modeling: Procedure
.org/10.1080/09613218.2013.743398. for compiling inference rules and operators for complex geometry.”
Jayawardene, V., S. Sadiq, and M. Indulska. 2015. An analysis of data qual- J. Comput. Civ. Eng. 31 (6): 04017062. https://doi.org/10.1061/(ASCE)
ity dimensions. ITEE Technical Report. St Lucia, Australia: Univ. of CP.1943-5487.0000705.
Queensland. Sebastian, R. 2011. “Changing roles of the clients, architects and contrac-
Jordani, D. A. 2010. “BIM and FM: The portal to lifecycle facility man- tors through BIM.” Eng. Constr. Archit. Manage. 18 (2): 176–187.
agement.” J. Build. Inf. Model. 13: 16. https://doi.org/10.1108/09699981111111148.

© ASCE 04020072-17 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072

 Singh, V., N. Gu, and X. Wang. 2011. “A theoretical framework of a Won, J., G. Lee, C. Dossick, and J. Messner. 2013. “Where to focus for
BIM-based multi-disciplinary collaboration platform.” Autom. Constr. successful adoption of building information modeling within organiza-
20 (2): 134–144. https://doi.org/10.1016/j.autcon.2010.09.011. tion.” J. Constr. Eng. Manage. 139 (11): 04013014. https://doi.org/10
Solihin, W., C. Eastman, and Y.-C. Lee. 2015. “Toward robust and quantifi- .1061/(ASCE)CO.1943-7862.0000731.
able automated IFC quality validation.” Adv. Eng. Inf. 29 (3): 739–756. Younas, M. 2019. “Research challenges of big data.” Serv. Oriented
https://doi.org/10.1016/j.aei.2015.07.006. Comput. Appl. 13 (2): 105–107. https://doi.org/10.1007/s11761-019
Taylor, J. E. 2007. “Antecedents of successful three-dimensional computer- -00265-x.
aided design implementation in design and construction networks.” Zadeh, P. A., G. Wang, H. B. Cavka, S. Staub-French, and R. Pottinger.
J. Constr. Eng. Manage. 133 (12): 993–1002. https://doi.org/10.1061 2017. “Information quality assessment for facility management.”
/(ASCE)0733-9364(2007)133:12(993). Adv. Eng. Inf. 33 (Aug): 181–205. https://doi.org/10.1016/j.aei.2017
.06.003.
Wang, L. 2017. “Heterogeneous data and big data analytics.” Autom. Con-
Zhang, C., and J. Beetz. 2015. “Model checking on the semantic web: IFC
trol Inf. Sci. 3 (1): 8–15. https://doi.org/10.12691/acis-3-1-3.
validation using modularized and distributed constraints.” In Proc.,
Wang, R. Y., and D. M. Strong. 1996. “Beyond accuracy: What data quality
32nd CIB W78 Conf. 2015, 819–827. Eindhoven, Netherlands:
means to data consumers.” J. Manage. Inf. Syst. 12 (4): 5–33. https://doi International Council of Research and Innovation in Building and
.org/10.1080/07421222.1996.11518099. Construction.
Whyte, J. 2011. “Managing digital coordination of design: Emerging
Downloaded from ascelibrary.org by Rice University on 04/27/20. Copyright ASCE. For personal use only; all rights reserved.

Zhang, J., Q. Liu, Z. Hu, J. Lin, and F. Yu. 2017a. “A multi-server

hybrid practices in an institutionalized project setting.” Eng. Project information-sharing environment for cross-party collaboration on a pri-
Organization J. 1 (3): 159–168. https://doi.org/10.1080/21573727 vate cloud.” Autom. Constr. 81 (Sep): 180–195. https://doi.org/10.1016
.2011.597743. /j.autcon.2017.06.021.
Whyte, J., A. Stasis, and C. Lindkvist. 2016. “Managing change in the de- Zhang, S., F. Pan, C. Wang, Y. Sun, and H. Wang. 2017b. “BIM-based
livery of complex projects: Configuration management, asset informa- collaboration platform for the management of EPC projects in hydro-
tion and ‘big data’.” Int. J. Project Manage. 34 (2): 339–351. https://doi power engineering.” J. Constr. Eng. Manage. 143 (12): 04017087.
.org/10.1016/j.ijproman.2015.02.006. https://doi.org/10.1061/(ASCE)CO.1943-7862.0001403.

© ASCE 04020072-18 J. Constr. Eng. Manage.

J. Constr. Eng. Manage., 2020, 146(7): 04020072