See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3421894

Business Process Analysis and Optimization: Beyond Reengineering

Article  in  IEEE Transactions on Systems Man and Cybernetics Part C (Applications and Reviews) · February 2008
DOI: 10.1109/TSMCC.2007.905812 · Source: IEEE Xplore

CITATIONS READS
253 13,846

3 authors:

Kostas Vergidis Ashutosh Tiwari

University of Macedonia The University of Sheffield
39 PUBLICATIONS   773 CITATIONS    303 PUBLICATIONS   7,479 CITATIONS   

SEE PROFILE SEE PROFILE

Basim Majeed
British Telecom
36 PUBLICATIONS   972 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Design Rationale and Medical Devices View project

Through-life performance: From science to instrumentation View project

All content following this page was uploaded by Ashutosh Tiwari on 14 March 2014.

The user has requested enhancement of the downloaded file.

 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS 1

Business Process Analysis and Optimization:

Beyond Reengineering
Kostas Vergidis, Member, IEEE, Ashutosh Tiwari, Member, IEEE, and Basim Majeed, Member, IEEE

Abstract—There is an abundance of business process modeling proaches and classifies them in three groups based on their
techniques that capture and address different aspects of a business characteristics. Section III identifies the different types of busi-
process. A limited number of these process models allow further ness process analysis techniques and classifies them in a similar
quantitative analysis, and only a few enable structured process im-
provement. This paper reviews and classifies the main techniques way. Section IV justifies the necessity to move from business
for business process modeling with regard to their analysis and op- process improvement to structured optimization and identifies
timization capabilities. Three primary groups are identified, and the scarce optimization approaches in the literature. The last
a selection of representative business process modeling techniques section presents an overview of all the classifications presented
is classified based on these. Similar classification is also presented and highlights the current situation, the research gap, and the
for the analysis and optimization approaches for business processes
that were identified in relevant literature. The main contribution of directions for future development regarding modeling, analysis,
the paper is that it identifies which types of business process mod- and optimization techniques for business processes.
els are suitable for analysis and optimization, and also highlights
the lack of such approaches. This paper offers a state-of-the-art
review in the areas of business process modeling, analysis, and II. BUSINESS PROCESS MODELS: A NOVEL CLASSIFICATION
optimization—underlining that the latter two have not received Havey [28] provides a simple definition of business processes
enough coverage and support in the literature.
as “step-by-step rules specific to the resolution of a business
Index Terms—Business process (BP), BP analysis, BP modelling, problem.” Since the 1990s when the first definitions of business
BP optimization, business processes, reengineering. processes appeared in the literature, many authors attempted
to focus business processes on specific directions. However,
I. INTRODUCTION in almost every reference in this area, two particular business
USINESS processes have received ample attention for process definitions are reverently cited. The first comes from
B more than a decade. Many approaches have been pro-
posed, many promises were made, but the spectacular results
Hammer and Champy [25], who state that “a business process
is a collection of activities that takes one or more kinds of inputs
that the reengineering revolution vowed were never fully real- and creates an output that is of value to the customer,” and
ized, making more and more people hesitant about the whole second is from Davenport [14], who claims that “a business
concept. One of the many reasons is that apart from enthusiastic process is defined as the chain of activities whose final aim is
descriptive proposals, a structured and repeatable methodology the production of a specific output for a particular customer or
that could be generally applied to business process modeling and market.” There are references such as [46], [47], and [67] that
improvement was never established. Compared with the large provide compilations of the various business process definitions.
number of proposed business process modeling techniques and Business process modeling plays a major role in the percep-
qualitative analysis approaches found in literature, business pro- tion and understanding of business processes. In most of the
cess optimization has received little coverage. While the term cases, a business process is as expressive and as communicative
“improvement” implies a qualitative approach of developing an as is the technique that has been used to model it. Therefore, the
existing business process to a better version, “optimization,” as elements and the capabilities of a business process model play
discussed in a later section, is more automated improvement of a significant role in describing and understanding a business
business processes using prespecified quantitative measures of process. There is an abundance of business process modeling
performance (objectives). techniques with approaches that capture different aspects of a
This paper discusses business processes by introducing a business process, each having distinctive advantages and dis-
novel classification scheme for business process models and advantages. Authors such as Kettinger et al. [34], Melao and
presenting the current trends in analysis and optimization ap- Pidd [47], and Aguilar-Saven [2] have provided frameworks for
proaches. The paper is organized as follows. Section II intro- presenting and classifying different business process modeling
duces the most representative business process modeling ap- techniques. Kettinger et al. [34] conducted a thorough study of
business process reengineering methodologies (25), techniques
Manuscript received July 11, 2006; revised October 19, 2006 and January (72), and tools (102) that are adopted by 25 international con-
11, 2007. This paper was recommended by Associate Editor M. Last. sultancy firms. The study reveals that, in every stage of the
K. Vergidis and A. Tiwari are with the Decision Engineering Centre, Cranfield
University, Cranfield MK43 0AL, U.K. (e-mail: [email protected]; reengineering process, there are a variety of approaches fol-
[email protected]). lowed. Kettinger et al. [34] report a widespread use of process
B. Majeed is with the Computational Intelligence Group, Intelligent Systems capture and modeling techniques. They also present a compre-
Research Center (ISRC), BT Research and Venturing, Martlesham IP5 3RE,
U.K. (e-mail: [email protected]). hensive list of the appropriate software tools and the techniques
Digital Object Identifier 10.1109/TSMCC.2007.905812 (e.g., process flowcharting, data flow diagramming) that each
1094-6977/$25.00 © 2008 IEEE
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS

of the tools supports. However, there is not much emphasis on

process modeling itself as it is viewed merely as a technique
among others that constitute the wider picture of business pro-
cess reengineering.
Melao and Pidd [47], on the other hand, focus exclusively in
business processes and their modeling. They adopt four different
perspectives for understanding the nature of business processes
first, and then, identify the most common modeling approaches
for each perspective. The first perspective views business pro-
cesses as deterministic machines, that is, as a fixed sequence
of well-defined activities that convert inputs to outputs in order
to accomplish clear objectives. For this perspective, static pro-
cess modeling is sufficient, with techniques such as integrated
definition methods (IDEF0, IDEF3) and role activity diagrams
(RADs). The second perspective views business processes as
complex dynamic systems, assemblies of interchangeable com- Fig. 1. Classification of business process modeling techniques.
ponents. This second viewpoint focuses on the complex, dy-
namic, and interactive features of business processes. The au-
thors suggest discrete event simulation (discussed later in this modeling techniques, as demonstrated in Fig. 1. The first set
paper) as a suitable way to model the dynamic behavior of (i.e., diagrammatic models) involves business process models
this approach. The third perspective of business processes is that sketch a business process using a visual diagram. The sec-
interacting feedback loops that highlight the information feed- ond set (i.e., mathematical models) corresponds to models in
back structure of business processes. System dynamics model- which all the elements have a mathematical or a formal under-
ers are recommended for this perspective. The last perspective pinning. Finally, the third set (i.e., business process languages)
of business process is social constructs, and emphasizes more contains software-based languages that support business pro-
on the people side. It is the people who made and enact busi- cess modeling and most of the times process execution. The
ness processes, people with different values, expectations, and classification of the most representative modeling techniques is
roles. This soft side of business processes can be modeled with demonstrated using a Venn diagram in Fig. 1. Each of the tech-
soft unstructured illustrative models. However, a real-life busi- niques is further discussed later in this section. Table I presents
ness process involves elements for all the four perspectives, and the classification of Fig. 1 and also cites a selection of references
therefore, it is evident that there is no such modeling technique for each of the key techniques. The remaining of this section dis-
that can embrace all this variety of characteristics that constitute cusses the main features of these process modeling techniques
a business process. based on the set (or sets) that they belong to.
Another notable review regarding business process model-
ing classification comes from Aguilar-Saven [2]. The author
presents the main process modeling techniques and classifies A. Diagrammatic Models—Simple and Communicative
them based on two dimensions: the first dimension is con- The first techniques that were used for business process mod-
cerned with four different purposes of use, and classifies the eling were plain graphical representations (i.e., flowcharts) that
business process models based on whether they are: 1) descrip- were initially developed for software specification [13], [35].
tive for learning; 2) enable decision support for process devel- These simplistic diagrams depicted a business process, but most
opment/design; 3) enable decision support for process execu- of the times without using a standard notation [28]. These tech-
tion, or 4) allow information technology (IT) enactment support. niques are useful for fast and informal process representation,
The second dimension distinguishes between active and passive but they lack the necessary semantics to support more com-
models. As active are considered those models that allow the plex and standardized constructs. This led to the development
user to interact with them (dynamic model), while passive are of standard methodologies such as IDEF and Unified Mod-
those that do not provide this capability. It is important to note eling Language (UML) for process modeling and/or software
that Aguilar-Saven [2] provides an extensive and updated list of development. Business process modeling benefited from these
software tools that are associated with all the process modeling standardized diagrammatic approaches, as well since they are
techniques presented in the paper. simple and easy to use. However, they have also received a series
As seen from the references described before, each of the of criticisms from various authors. The central point of argument
authors provides a different modeling framework according to is that these modeling approaches are based on graphical nota-
his or her focus on specific directions. In the current paper, the tions only [85], thus lacking formal semantics [70]. They also
authors propose a new classification scheme for business pro- lack quantitative information that obstructs any further analysis
cess models. The purpose of this scheme is to classify the most and development of analysis methods and tools [80]; there is no
a range of business process models according to their structural formal underpinning to ensure consistency across models [70].
characteristics and their capabilities for analysis and optimiza- Phalp and Shepperd [52] note that any analysis which attempts
tion. The authors propose three sets to classify business process using these types of models often consists solely of inspection
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

VERGIDIS et al.: BUSINESS PROCESS ANALYSIS AND OPTIMIZATION: BEYOND REENGINEERING 3

TABLE I
MAIN MODELING TECHNIQUES, CORRESPONDING SETS, AND SELECTED REFERENCES

of diagrams and the conclusions are heavily dependent upon the der Aalst et al.’s [79] suggestion that business process mod-
skills of the analyst. els “should have a formal foundation” because formal models
Although visual inspection of diagrams tends to be highly do not leave any scope for ambiguity and increase the poten-
subjective, these diagrams are still widely used in business pro- tial for analysis. However, there is a lack of formal methods
cess environments. The unbeatable advantage to visually depict to support the design of processes [30] because business pro-
the flow of a business process in a way that no technical exper- cess elements and constraints are mostly of qualitative nature,
tise is required is very appealing to the business analysts. Even and it is hard to characterize them in a formal way amenable
advanced and more sophisticated modeling techniques are in- to analytical methods [68]. This explains the difficulty of de-
fluenced by this perspective, and they support apart from formal veloping “parametric” models of business processes and the
semantics and a visual representation of the modeled processes. fact that only a few practical examples are found in relevant
A typical example is the Petri nets, discussed later. literature [30].
A Petri net is an example of a business process modeling tech-
nique that combines visual representation using standard nota-
B. Formal/Mathematical Models—Consistent But Complex tion with an underlying mathematical representation. A Petri net
The necessity for formal semantics to business process is a graphical language that is appropriate for modeling systems
modeling led to a second generation of formal models. For- with concurrency [73]. The graph of a Petri net is a directed,
mal models are the ones in which process concepts are de- bipartite graph consisting of two kinds of nodes, called places
fined rigorously and precisely, so that mathematics can be used and transitions. Petri nets have been modified and extended by
to analyze them, extract knowledge from them, and reason various researchers to allow for more powerful modeling ca-
about them. An advantage of formal models is that they can pabilities. Some of their variations include timed Petri nets,
be verified mathematically, and can be checked for consistency stochastic Petri nets, colored Petri nets, and hierarchical Petri
and other properties [38]. These models are in line with van nets [36].
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS

Coming to the approaches that use mathematical models only, ing the most distinctive. van der Aalst et al. [79] remark that
there is not a widely accepted model. This results into different process languages with clear semantics are useful as they can
authors presenting their individual approaches toward math- express business process models and contribute to the analysis
ematical business process modeling. An approach that has a of their structural properties.
mathematical basis is proposed by Hofacker and Vetschera [30]. Havey [28] claims that BPEL is the most popular as it is
They describe a business process using a series of mathematical supported by IBM, Microsoft, and BEA. BPEL is not a nota-
constraints (that define the feasibility boundaries of the busi- tional language but it is also XML-based, and as such, it in-
ness process) and a set of objective functions (that consist of herits XML attributes such as programmability, executability,
the various objectives for business process design). Although and exportablility. BPML is a product of the Business Process
this approach cannot model complex modeling constructs and Modeling Initiative (www.bpmi.org). It is also an XML-based
there is no emphasis on the diagrammatic representation, it is language that encodes the flow of a business process in an exe-
appropriate for further quantitative analysis and improvement cutable form. BPML is accompanied by Business Process Mod-
as it is based on a mathematical model. A similar approach is eling Notation (BPMN), a graphical flowchart language that
presented by Powell et al. [54]. They describe a mathemati- is able to represent a business process in an intuitive visual
cal model that has the main ingredients of a generic business form [28]. Each BPML process has a name, a set of activi-
process. Valiris and Glykas [70] also propose the use of for- ties, and a handler; it also supports subprocesses. Yet Another
mal mathematical notations as a way of introducing business Workflow Language (YAWL) is another—as the name itself
rules and verifying the logical consistency of diagrammatic says—graphical process language created by van der Aalst and
models. ter Hofstede [78]. YAWL is a Petri-net-based language that was
Despite their advantages over simple diagrammatic ap- built with the primary target to support a wide range of business
proaches, criticisms for formal/mathematical business process process patters. It has received criticism for being inadequate in
models have also been reported. Building a formal business terms of expressiveness and system integration capabilities [28].
process model can prove much more complex and demanding JBoss Business Process Management (jBPM) execution lan-
compared to traditional techniques where a process diagram guage named jPDL [37] is also a novel approach to business
is sufficient [30]. These authors also show that the representa- process modeling and execution. This new approach facilitates
tion of real-life processes using mathematical models may be the natural transition from declarative input by the business an-
complex and sometimes not possible as these include complex alyst to the programming logic needed to implement a business
features such as decision points, feedback loops, and parallel or process, thus simplifying business process development and al-
hierarchical flow. Koubarakis and Plexousakis [38] note that the lowing even nonprogrammers to develop business processes
use of complex mathematical notations might discourage the using visual tools. jBPM engine is based on open source soft-
business analyst since “it is a lot of work to create, maintain a ware, providing infrastructure to developers who have access
formal business process and retain its consistency.” However, as to a variety of supplementary software tools with which they
a diagram can lead to ambiguity about the process, the formal can easily design and analyze business processes in a graphical
model ensures that the process is described accurately, and anal- environment.
ysis tools can be used to extract quantitative information about In order to assess and compare the capabilities of these mod-
the process. This is the main advantage of formal business pro- eling techniques, the authors investigated the patterns that each
cess modeling techniques. supports when it comes to business process modeling. A pattern,
according to Riehle and Zuillinghoven [59] is “the abstraction
from a concrete form which keeps recurring in specific non-
C. Business Process Languages—New and Executable arbitrary contexts.” The reason pattern support is essential for a
The third—and most recent—generation of business process process modeling because patterns enable the standardization of
modeling techniques came as an attempt to tackle the complexity solutions to commonly recurring problems within business pro-
of the formal models but retain their consistency and potential cesses and the reuse of these standardized process parts across
for further analysis. As the first generation of business process different process models. Table II presents a selection of pat-
modeling techniques was strongly influenced by the ones used terns from van der Aalst and ter Hofstede [77] that are consid-
in software development; so is this generation. Perhaps it is the ered as the basic constructs for any business process model and
dynamic, complex, and rapidly evolving nature of business pro- identifies which business process modeling techniques support
cess models that makes them similar to software development these patterns. Most of the business process languages are im-
techniques. The third set presented here takes business process plemented taking into account the process patterns of Table II.
modeling a step further as it uses process languages—usually For example, YAWL supports all these patterns since it was
XML-based—to model and execute a business process. This is created primarily for this purpose [77]. BPEL also supports
how business process languages were evolved. These context- most patterns [28], [84] and also BPML [28]. According to
specific executable languages are the latest trend in business Koenig [37], jBPM’s jPDL was also implemented to cover all the
process modeling, a trend that has already produced a number patterns presented here. Therefore, business process languages
of different semantic packages, with Business Process Execu- prove a reliable tool to formally model and visualize a busi-
tion Language for Web Services (BPEL4 WS—also known as ness process in terms of constructing standardized and reusable
BPEL) and Business Process Modeling Language (BPML) be- models.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

VERGIDIS et al.: BUSINESS PROCESS ANALYSIS AND OPTIMIZATION: BEYOND REENGINEERING 5

TABLE II
PROCESS PATTERNS SUPPORTED BY BUSINESS PROCESS LANGUAGES

III. BUSINESS PROCESS ANALYSIS: AN OVERVIEW

Fig. 2. Types of process analysis against the business process modeling sets.
According to Irani et al. [31], businesses should not be ana-
lyzed in terms of the functions in which they can be decomposed
to or in terms of the products they produce, but in terms of the key
business processes that they perform. Due to the complexity of
process design and control encountered in modern businesses, common analysis approach using visual models of business pro-
there is a need for the development of suitable analysis tech- cesses [52]. The observational analysis technique offers a set of
niques [73]. However, business process analysis is a term used options to redesign a process that includes eliminating nonvalue-
with a rather broad meaning including a range of different tactics added activities (e.g., redundant, rework, and supervisory activi-
such as simulation and diagnosis, verification, and performance ties), simplifying activities, combining activities, increasing the
analysis of business processes. van der Aalst et al. [79] under- concurrency of activities and automating activities [39]. How-
line that business process analysis should aim at investigating ever, this analysis approach can be time-consuming and heavily
properties of business processes that are neither obvious nor dependent upon the experience of the modeler whose conclu-
trivial. Boekhoudt et al. [11] justify the necessity for analysis sions are frequently based upon his knowledge of the particular
of business process models in order to clarify the business pro- business domain and his skills [50]. Zakarian [85] recognizes
cess characteristics, identify possible bottlenecks, and compare that diagrammatic process models have qualitative notation, and
any potential process alternatives. Yet most of business process this results in the lack of analysis tools, thus making the ap-
analysis approaches are based on subjective rather than objective plication of quantitative methods very unusual [82] and also
methods [70]. In line with van der Aalst et al. [79], Boekhoudt unattractive. Making business process analysis meaningful and
et al. [11] also report that among the modeling techniques, attractive is not only linked to the construction of ever-more
those that have formal semantics and mathematical basis are detailed maps, which use increasingly sophisticated representa-
the most suitable for analysis. Irani et al. [31] citing Davenport tional techniques, but also the willingness to combine seemingly
[14] state that to understand and analyze a business process irreconcilable strategies for analysis [10].
helps to recognize the sources of problems and ensure that they When analyzing a business process, it is necessary to have
are not repeated in the new process, thus providing a measure mechanisms more sophisticated than simple qualitative anal-
of value for the proposed changes. This approach opposes the ysis of static diagrammatic models. Authors such as Aguilar-
radical attitude toward business process redesign introduced by Saven [2] and Zakarian [85] stress the need for formal techniques
Hammer and Champy [25]. This section presents the differ- for analysis of process models, in order to make process mod-
ent types of business process analysis and presents a variety of eling methodologies more attractive and meaningful. The need
representative approaches found in literature. for quantitative analysis of the business process models is one of
the major reasons for the evolution of process models with for-
mal underpinning (i.e., mathematical models set). These formal
A. Different Analysis Types: From Observational Analysis to approaches to modeling of business processes provide a sound
Performance Evaluation basis for setting performance indicators that measure the attain-
There are different types of analysis related to business pro- ment of strategic goals and objectives by relating these goals
cess. Fig. 2 presents these different analysis types in a Venn and objectives to the core processes [42]. For these to occur,
diagram. It matches the types of process analysis to each of the analyses types that present both dynamic and functional aspects
three business process modeling sets introduced in the previ- of the process are required. According to van der Aalst [73],
ous section. For the first set of business process modeling (i.e., most of the techniques that are used for the analysis of formal
diagrammatic models), only observational analysis is at hand. business process models originate from operations research.
Observational analysis, which primarily entails altering the pro- Fig. 2 demonstrates that the three different types of busi-
cess structure via inspection of the diagrams [3], is the most ness process analyses proposed by van der Aalst [75] (having
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS

workflows—and in particular Petri nets—in mind) belong to the pressions that can be expressed using the process language, and
mathematical models set: thus, be integrated within the process model. However, it is not
1) validation, i.e., testing whether the business process be- sufficient to just develop these techniques. It is important to look
haves as expected in a given context; at methods and tools to make them applicable in the practical
2) verification, i.e., establishing the correctness of a business context [73].
process;
3) performance analysis (or performance evaluation), i.e.,
evaluating the ability to meet requirements with respect to B. Simulation of Business Process Models
throughput times, service levels, and resource utilization Simulation is a popular technique for analyzing business pro-
or other quantitative factors. cesses, and it can involve other types of analyses mentioned
It is obvious that none of the aforementioned analysis types before. According to Volkner and Werners [82], many problems
can be applied to a visual diagram only; formal underpinning of business processes have similarities to problems in project
of the process model is required. Validation checks whether the management or production process planning that have already
system behaves as expected in a particular context, while ver- been solved successfully using simulation. Simulation provides
ification checks whether the business process model is free of a structured environment in which one can understand, analyze,
logical errors [73]. Verification, unlike validation, is context in- and improve business processes [23]. Business process simula-
dependent; it detects, for example, deadlocks in process designs tion is used to assist decision-making by providing a tool that
a logical error independent of the purpose of the process. Per- allows the current behavior of a system to be analyzed and
formance evaluation aims to describe, analyze, and optimize the understood. It is also able to help predict the performance of
dynamic, time-dependent behavior of systems [29], [55]. Valida- the system under a number of scenarios determined by the deci-
tion can be done by interactive simulation: a number of fictitious sion maker [22]. Process simulation facilitates process diagnosis
cases are fed to the system to see whether they are handled well. (i.e., analysis) in the sense that by simulating real-world cases,
However, verification and performance analysis require more what–if analyses can be carried out [73]. The advantage of sim-
advanced analysis techniques [75]. Li et al. [44] present another ulation is that it is a very flexible technique [74] because it can
classification of business process analyses. This classification is be used to obtain an assessment of the current process perfor-
very similar to that proposed by van der Aalst [75]. According mance and/or to formulate hypotheses about possible process
to Li et al. [44], workflow model analysis is conducted mainly at redesign [1].
three levels—the logical, the temporal, and the performance lev- Modern simulation packages allow for both the visualization
els that deal with different aspects of a workflow model. Logical and performance analysis of a given process [74] and are fre-
level focuses on the correctness of the various process events quently used to evaluate the dynamic behavior of alternative
(i.e., verification), and temporal level focuses on the interval designs [3]. Visualization and graphical user interface are im-
dependency relations of a workflow model with imposed tim- portant in making the simulation process more user friendly.
ing constraints (i.e., validation). The logical and temporal levels According to Fathee et al. [18], simulation is most useful for the
ensure only a functionally working workflow but not its opera- analysis of stable business processes and less useful for dynamic
tional efficiency. The performance level focuses on evaluating systems that do not reach equilibrium. The main advantage of
the ability of the workflow to meet requirements with respect to simulation-based analysis is that it can predict process perfor-
some key performance indicators. Although performance anal- mance using a number of quantitative measures such as lead
ysis of business processes is recognized as a significant step time, resource utilization, and cost [22]. As such, it provides
toward quantitative analysis of business processes, it has not a means of evaluating the execution of the business process to
captured the attention of many researchers [63]. determine inefficient behavior [19]. Thus, business process exe-
The concept behind business process languages is to make a cution data can feed simulation tools that exploit mathematical
process executable, and hence, amenable to quantitative analy- models for the purpose of business process optimization and re-
sis. However, for business process languages set, only simula- design [1]. Dynamic process models can enable the analysis of
tion is proposed in literature explicitly. Simulation is a software- alternative process scenarios through simulation by providing
assisted technique for analyzing business process; it is discussed quantitative process metrics such as cost, cycle time, service-
later in a separate section. Although formal languages have been ability, and resource utilization [23]. These metrics form the
exploited in order to define and model business processes, the basis for evaluating alternatives and selecting the most promis-
use of formal languages to handle the performance evaluation ing scenario for implementation [41]. However, these analytical
of workflows has received little coverage [1]. However, some models (mostly mathematical), according to Gunasekaran and
process languages have associated analysis techniques that can Kobu [23], have not received much attention due to their com-
be used for investigating process properties. These techniques plexity despite their ability to play a greater role in measuring
can then be relied upon to provide insight into the behavior performance and in conducting experiments.
and characteristics of a business process model specified in the The advantages of applying simulation are: 1) the possibility
language [79]. According to the authors’ opinion, this level of for the quantitative analysis of business processes with consid-
modeling and execution of business processes (i.e., using a pro- eration to their dynamic characteristics; 2) the possibility for a
cess language) is the most suitable for the application of any systematic generation of alternatives by modifications in iden-
analysis technique. These can be in the form of algorithmic ex- tified weak points; and 3) the high flexibility in modeling as
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

VERGIDIS et al.: BUSINESS PROCESS ANALYSIS AND OPTIMIZATION: BEYOND REENGINEERING 7

TABLE III
BUSINESS PROCESS ANALYSIS APPROACHES BASED ON MODELING SETS AND ANALYSIS TYPES

well as an adequate consideration of stochastic influences [82]. business process using a combination of fuzzy logic and rule-
However, simulation has some weak points as well. Some au- based reasoning. Using—although not explicitly mentioned—
thors [22], [82] report the large costs involved and the large observational analysis, he extracts IF–THEN fuzzy rules from
amount of time to build a simulation model due to the com- the IDEF3 model and defines a number of linguistic variables.
plexity and knowledge required building such models. van der The linguistic variables are categorized into fuzzy sets that are
Aalst [74] underlines that simulation supports only “what– defuzzified by assigning precise boundaries. The process is ac-
if” analysis and does not suggest any process improvements. curately executed, and its output is quantified and predicted by
Basu and Blanning [8] also claim that while process simulation eliminating the values of each variable. Combinations of differ-
can provide useful insight into process behavior, it does not ent values for each variable can be applied to analyze and test
address questions about the interrelationships among process the process and its outputs. Peters and Peters [51] also present
components. a tool to simulate an IDEF0 model by making dynamic trans-
formations. Other IDEF-based analysis approaches come from
Badica et al. [5] and Shimizu and Sahara [65]. Another group
C. Compilation of Approaches Regarding Business of analysis approaches is related to the quantification of RADs.
Process Analysis Phalp and Shepperd [52] attempt to quantify RADs. The au-
After identifying the main analysis types for business pro- thors extract a metric (coupling ratio) to measure the correlation
cess, the most relevant approaches found in literature are dis- between actions (sole activities of a role) and interactions (in-
cussed. Table III presents the analysis types and approaches for volvement of another role). By reducing coupling, roles can
a selection of business process modeling techniques. For each become more autonomous within the process because they do
process modeling technique, the table cites the modeling set(s) not need to synchronize. Badica et al. [6] attempt to map and
it belongs to (Fig. 1), the types of analyses applicable based on quantify RADs using a similar approach.
these sets (Fig. 2), and a selected number of related approaches When it comes to process models with formal underpinnings,
(references). According to Table III, most analysis approaches two main approaches are identified: those built around Petri nets
reported in the literature are based on models that belong to the and those that use mathematical models of business processes.
diagrammatic models set. Also, no analysis approach is reported van der Aalst has produced a series of papers focusing on differ-
for the business process languages set. ent aspects of Petri nets and workflow analysis (refer to Table III
IDEF models have been a starting point for business process for references), but he tends to focus more on validation, verifi-
analysis for authors such as Kusiak and Zakarian who have cation, and correctness of workflows rather than on performance
published a series of papers (refer to Table III for references) analysis. Other analysis approaches include that of Donatelli
exploring and analyzing various aspects of IDEF models. The et al. [16] that involves process algebra and stochastic Petri nets,
most representative is Zakarian’s [85] where the author is using and Gao et al. [21] that applies fuzzy reasoning to Petri nets.
an IDEF3 model attempting to model and analyze/quantify a In terms of mathematical models, Powell et al. [54] propose a
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS

series of mathematical formulations and ratios to measure, an- activities; but again, the improvement process is not transpar-
alyze, and control business processes. Valiris and Glykas [71] ent. This approach does not guarantee an optimum redesign as
propose a framework that contains a series of metrics for busi- it manually derives alternative process maps starting from the
ness processes. As mentioned previously in this section, as of current process map.
now, there are no reported analysis approaches explicitly for A methodology for business process improvement is only as
business process languages. This gap and its potential are dis- good as the tools and techniques that support it [7]. Unfortu-
cussed later in this paper. nately, the literature restricts itself to descriptions of the “situa-
tion before” and the “situation after,” giving very little informa-
IV. BUSINESS PROCESS OPTIMIZATION tion on the redesign process itself [57]. Valiris and Glykas [71]
BEYOND REENGINEERING criticize this perspective, stating that most of these reengineer-
ing methodologies lack the formal underpinning to ensure the
As the previous section discussed, business process model-
logical consistency of the generation of the improved business
ing does not add much value without further inspection and
process models. This leads to a lack of systematic approach
analysis of the business processes model. Likewise, process
that can guide a process redesigner through a series of (repeat-
analysis has little value, unless it helps in improving or opti-
able) steps for the achievement of process redesign [71]. While
mizing a business process [79]. Process improvement can occur
there are several methodologies for structuring business process
through associated formal techniques [79] that support both the
redesign projects, the task of developing optimal designs of busi-
modeling and the analysis of business processes [80]. A holis-
ness processes is left to the designer’s intuition [30]. Business
tic approach toward business processes should capture a busi-
process optimization is the automated improvement of business
ness process (business process modeling), provide the necessary
processes using prespecified quantitative measures of perfor-
means for bottleneck identification and performance analysis,
mance (objectives), and as discussed in the next section, it is the
and—eventually—generate alternative improved business pro-
appropriate systematic approach to fill in this gap.
cess(es) in terms of specified objectives. But often this last
part (business process optimization) is overlooked—if not com-
pletely neglected—in business process literature. This section B. Two Perspectives for Business Process Optimization
discusses the difference between process improvement and op-
Business process optimization can espouse techniques from
timization, and provides a classification of the current business
relevant disciplines. Gunasekaran and Kobu [23] claim that,
process optimization approaches.
within the business process context, there is a need for a wider
use of decision support systems based on artificial intelligence
A. Improvement Is Not Enough
and expert systems. They also support the need for developing
Business process improvement started as part of business pro- queuing, linear programming, and simulation models to repre-
cess redesign and/or reengineering efforts that promised excep- sent business processes and to select the optimal design. In this
tional results. Gunasekaran and Kobu [23] claim that a business section, we discuss and relate two other disciplines with business
process has to undergo fundamental changes to achieve signifi- processes: scheduling and evolutionary computing. Scheduling
cant performance improvements. According to Soliman [66], the shares a range of common topics with business processes, and
objectives of business process reengineering are to improve the evolutionary computing is an already successful optimization
business processes and reduce costs. However, although most approach in other areas.
of the business process reengineering (or redesign) attempts in 1) Business Processes and Scheduling: Scheduling prob-
literature claim to support business process improvement, there lems are similar to business process optimization problems.
are scarce cases that describe with sufficient details the actual Both disciplines share common topics such as the optimal allo-
improvement steps that need to be undertaken. Jaeger et al. [32] cation of resources to tasks [72]. Having this in mind, a range
is a typical case where business process improvement is limited of already successful optimization approaches from scheduling
to a broad description of steps that need to be undertaken. The can become available to business processes taking into account
steps according to these authors are as follows. what Ernst et al. [17] claim; that optimization capabilities are
1) Specify the system. generally targeted at a specific application area and cannot be
2) Identify the performance bottleneck(s). easily transferred toward another discipline.
3) Choose among the possible modifications to resolve the According to Bellabdaoui and Teghem [9], the development
performance bottlenecks. of optimization models for planning and scheduling is one of the
These—almost obvious—guidelines are not sufficient for a most useful tools for improving productivity in a large number of
structured process improvement as they do not provide the nec- companies. There is a range of review papers regarding schedul-
essary insight and level of detail for the actions that lead to ing optimization approaches. Mathematical programming, espe-
process improvement. Another similar approach is presented by cially mixed integer linear programming (MILP), has become
Aldowaisan and Gaafar [3], and it is based on observational one of the most widely explored methods for process scheduling
analysis. Their technique has a set of options to redesign a pro- problems because of its rigor, flexibility, and extensive model-
cess. This includes eliminating nonvalue-added activities (e.g., ing capability. Floudas and Lin [20] present an overview of the
redundant, rework, and supervisory activities), simplifying ac- developments of MILP-based approaches for scheduling and ob-
tivities, combining activities, and increasing the concurrency of serve increasing application of the formal MILP optimization
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

VERGIDIS et al.: BUSINESS PROCESS ANALYSIS AND OPTIMIZATION: BEYOND REENGINEERING 9

framework to real scheduling problems in process and related for example, have already been used to find solutions to schedul-
industries. Kallrath [33] gives an overview of the current state- ing problems and their variants [30]. Hart et al. [27] present a
of-the-art of planning and scheduling problems and reaches to review of applied evolutionary computing methods to schedul-
similar conclusions. According to this author, the state-of-the- ing problems, and they report the existence of evolutionary
art technology based on mathematical, especially mixed-integer algorithms that are capable of tackling large and hard real-world
optimization for planning, is quite advanced and appropriate problems and are competitive with traditional techniques. There
for solving real-world planning problems. The reason is that are a number of benefits in using evolutionary-based optimiza-
mixed integer optimization can provide a quantitative basis for tion. One significant advantage lies in the gain of flexibility
decisions, and it has proven itself as a useful technique to re- and adaptability to the task in hand, in combination with robust
duce costs and to support other objectives. Rommelfanger [60] performance and global search characteristics [4]. According
presents another scheduling optimization approach that involves to Moon and Seo [49], the most attractive feature of evolution-
fuzzy mathematical programming. While in the case of classical ary algorithms is the flexibility of handling various kinds of
models, the vague data are replaced by “average data,” fuzzy objective functions with few requirements on fine mathemati-
models offer the opportunity to model subjective judgment of cal properties. Wang et al. [83] note that process optimization
a decision maker as precisely as the decision maker is able to is a difficult task due to the nonlinear, nonconvex, and often
describe it. In contrast to classical systems, in fuzzy systems discontinuous nature of the mathematical models used.
combined with an interactive solution process, the information Regarding business processes, the evolutionary approaches
can be gathered step by step. Another advantage of fuzzy mod- reported are rather limited. Hofacker and Vetschera [30] have
els is the fact that mixed integer programming problems can be attempted to transform and optimize a business process model
solved easily because the boundaries are not crisp. using GAs, but they report nonsatisfactory results. The model
These scheduling problems are inherently combinatorial in is based on a series of mathematical formulations and is highly
nature because of the many discrete decisions involved, such as constrained, thus making it hard for the algorithm to locate
equipment assignment and task allocation over time. Shah [64] solutions. Tiwari et al. [69] and Vergidis et al. [81] extended
examines different techniques for optimizing production sched- their mathematical model and applied multiobjective optimiza-
ules with an emphasis on formal mathematical methods. Pinto tion algorithms, such as the Non-Dominated Sorting Genetic
and Grossmann [53] also present an overview of assignment Algorithm 2 (NSGA2) and the Strength Pareto Evolutionary
and sequencing models used in scheduling with mathematical Algorithm 2 (SPEA2), and report satisfactory results that pro-
programming techniques. A recent review comes from Mendez vide encouraging opportunities for further investigation. These
et al. [48] that present an extensive classification of scheduling and other approaches toward business process optimization are
problem types that demonstrates their great diversity. Address- further discussed in the following section.
ing this diversity, these authors also present a general classifica- In general, evolutionary optimization could benefit business
tion of optimization models as a framework for describing the processes by discovering process designs that are perhaps over-
major optimization approaches that have emerged over the last looked by a human designer. Also, these techniques can evaluate
decade regarding scheduling. a significant number of alternative designs based on the same
From the aforementioned information, one can conclude that process and determine the fittest based on specific objectives.
scheduling optimization is an established research area reporting GAs could also be related with a new concept, the one of au-
successful approaches. These approaches can inspire relevant tomatic process generation. A process design could be either
applications in business process optimization. However, busi- generated or modified in an automatic way based on different
ness processes model other elements not covered by scheduling paths of execution and different objectives each time. It is a
problems, such as decisions, business rules, etc., that are hard to quite new and intriguing area of process optimization where
be expressed mathematically. Ernst et al. [17] report that math- evolutionary techniques can significantly contribute.
ematical programming formulations can only be applied when
constraints and objectives can be expressed mathematically.
Hence, relevant approaches can be applied to simplified ver- C. Current Business Process Optimization Approaches
sions of business processes. As it is later discussed, there are op- Zhou and Chen [88] suggest that business process optimiza-
timization approaches on mathematically formulated business tion should aim at reducing lead time and cost, improving qual-
processes from authors such as Hofacker and Vetschera [30]. ity of product, and enhancing the satisfaction of customer and
These approaches, although consistent, are overly complicated, personnel so that the competitive advantage of an organization
and still deal with simplistic business processes. Taking into ac- can be retained. Reijers [56] suggests that the goals of business
count that scheduling is solely based on mathematical models, process optimization are often the reduction of cost and flow
it is questionable whether business process optimization should time. However, Hofacker and Vetschera [30] underline that the
follow the same path or investigate alternative ways that express concept of “optimality” of process designs is not trivial, and
a business process using a variety of components. the quality of processes is defined by many, often conflicting
2) Business Processes and Evolutionary Computing: Evo- criteria. Both in application and theory, great importance is at-
lutionary techniques use the principles of evolution to guide the tached to the optimization of business processes, mostly without
optimization process, and they have been successfully applied explaining the criteria and the alternatives considered for opti-
to several combinatorial problems. Genetic algorithms (GAs), mization [82]. But Zhou and Chen [87] remark that there is
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

10 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS

structural conflict identification or deadlock removal. Graph-

reduction techniques have also engaged the attention of a series
of authors. These are algorithmic techniques that modify a dia-
grammatic model of a process. Sadiq and Orlowska [62] identify
and try to analyze and resolve two structural conflicts in process
models: deadlock and lack of synchronization. van der Aalst et
al. [76] regard the previous approach as incomplete and propose
a new algorithm. A similar approach is also followed by Lin et
al. [45], who present a complete and minimal set of rules and
a novel algorithm to implement the identification of structural
conflicts in process models. In this case, the correctness and
completeness of the algorithm are proved. Again, graph reduc-
tion techniques are not related with quantifiable performance
measures although they have algorithmic foundation.
The majority of optimization techniques are related to algo-
Fig. 3. Improvement/optimization capabilities of the business process model-
rithmic approaches. Soliman [66] provides a typical description
ing sets. of an optimization problem. According to this author’s approach,
business processes may be considered as a complex network
of activities connected together with decision variables and an
still no systematic optimization methodology for business pro- objective function subject to a number of constraints. Similar
cesses. Fig. 3 classifies the improvement and the optimization approach to the optimization problem is proposed by Hofacker
capabilities of business process models using the same sets like and Vetschera [30], who provide analytical support for optimiz-
Figs. 1 and 2. ing the design of (mainly administrative) business processes.
As mentioned previously, optimization is not an option for Their paper introduces formal models of the business process
diagrammatic process models. This is because optimization re- design problem, which can be used to analytically determine
quires quantitative measures of process performance that cannot optimal designs with respect to various objective functions sub-
be produced in diagrammatic models. However, there are many ject to a number of constraints. It is perceived to be the most
qualitative improvement approaches applied to diagrammatic complete paper in the area of business process optimization
process models such as that by Zakarian [85] and Phalp and because along with the formal business process model, three
Shepperd [52] to name a few. But these techniques are lim- different optimization techniques are examined: mathematical
ited as they develop the existing diagrammatic models based on programming, a branch and bound method, and GAs. Tiwari
trial-and-error approach. Graph reduction technique is another et al. [69] present an extension of the same formal model by ap-
systematic approach for business process optimization applica- plying multiobjective optimization for business process designs
ble to models that have elements from both the diagrammatic and Vergidis et al. [81] demonstrate the optimized alternatives.
and the mathematical models. Current optimization approaches Optimization of a business process under multiple criteria is
are related almost exclusively to the formal modeling techniques quite attractive since business processes often have conflicting
on the mathematical models set. This is because the formality criteria [30].
and quantitative nature of these models allows for systematic Gutjahr et al. [24] present a stochastic branch-and-bound
optimization. Quantitative criteria are considered essential in or- approach for solving hard combinatorial business process re-
der to evaluate the improvements in a business process through lated problems. Jaeger et al. [32] also provide an optimization
modifications to the basic structure [82]. Business process lan- framework based on performance evaluation that makes both re-
guages set could accommodate executable models of process source and process changes to improve a system’s performance.
optimization, but to the authors’ knowledge, there is no litera- Han [26] develops an algorithmic framework to design business
ture reference in this area. processes using decision models. The aim of this methodology
Table IV summarizes the main business process optimiza- is to reduce the total cost of implementing decisions by creating
tion approaches found in literature; mostly related to Petri nets a quantitative model and using four design change patterns: 1)
and mathematical process models. Taking into consideration simple automation for process streamlining; 2) linear sequenc-
the emphasis that has been put on Petri nets for their analy- ing; 3) resequencing involving process parallelization; and (4)
sis capabilities, one would expect that they would also fit for radical process integration that is implemented algorithmically.
optimization purposes. But, according to Lee [40], Petri nets Zhou and Chen [86]–[88] have published three papers re-
are not adequate to solve optimization problems except using garding business process optimization. Zhou and Chen [88]
graph reduction techniques. Although they can capture system introduced the concept of assignment quality and developed
dynamics and physical constraints, they are not suitable for multiobjective evaluation, combining optimization models for
optimization problems with combinatorial characteristics and intra- and interenterprise business processes, and they use the
complex precedence relations. NSGA to solve this problem. Zhou and Chen [86] focus more
Li et al. [43] suggest that another way of analyzing and on time, cost, and resource constraints of a business process
improving a business process is graph reduction technique for model and attempt to optimize it by utilizing a GA to minimize
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

VERGIDIS et al.: BUSINESS PROCESS ANALYSIS AND OPTIMIZATION: BEYOND REENGINEERING 11

TABLE IV
OPTIMIZATION APPROACHES FOR FORMAL BUSINESS PROCESS MODELS

the process cost. Lastly, Zhou and Chen [87] develop a system- it is possible to take actions to improve and optimize process
atic design methodology for business process optimization from execution [12].
strategic, tactical, and operational perspectives using structured
and quantitative methods that support the design. This opti-
mization optimally assigns resource capabilities, organizational V. OVERVIEW AND DISCUSSION
responsibilities and authorities, and organizational decision This paper presented and classified a wide selection of refer-
structure. ences regarding business process modeling, analysis, and opti-
Another approach to optimization is the consolidation of the mization. The review was based on a novel classification of the
activities (or tasks) of a business process. Rummel et al. [61] existing business process modeling techniques using a grouping
propose a model that focuses on shortening the cycle time of of three sets according to their modeling capabilities. The anal-
a business process by consolidating activities—assigning mul- ysis and optimization approaches were also related to the same
tiple activities to one actor—thereby eliminating the coordina- classification scheme. These classifications resulted in visually
tion and handoff delay between different activities that occurred highlighting a number of interesting observations, and espe-
when assigned to different actors. As this approach is activity (or cially, the lack of certain approaches. Table V summarizes the
task) focused, it ignores interactivity delay that may contribute main business process models that were discussed in this paper
significantly to overall process cycle time. Dewan et al. [15] along with their associated modeling, analysis, and optimization
claim that there is no systematic methodology to determine the capabilities.
optimal rebundling of information-intensive tasks. They present It is evident from Table V that business process optimization
an approach to optimally consolidate tasks in order to reduce the has not received as much attention in comparison to business
overall process cycle time. The authors present a mathematical process modeling and analysis techniques. Business process
model to optimally redesign complex process networks but a modeling has always attracted attention of researches from a
limitation of the paper is that it refers to business processes with variety of fields. This resulted in a variety of modeling ap-
information flows only. Its main contribution is the effective proaches that are used for business processes. Each of these
business process restructuring and the reduction of the overall diverse modeling approaches has distinctive advantages but still
task time using handoff delay reduction or elimination as a re- what is missing is a holistic approach that will involve elements
sult of a unified methodology applicable to multiple task-based from all the three sets presented in this paper. There is a need
business processes. for defining operational and reusable business process models
Although formal languages have associated analysis tech- within different types of enterprises, in different contexts, and
niques that can be used for investigating properties of pro- at the required level of detail. These models should be able
cesses [79], an optimization approach based on executable pro- to address the complexity of the design and identify problems
cess languages was not observed in the literature. Since most of encountered in modern business processes. Therefore, there is
the optimization approaches—as discussed before—are based an increasing need for formal methods and techniques to sup-
on algorithmic approaches, these could be easily translated to port both the modeling and the analysis of business processes.
executable software programs. Analysis and optimization of However, despite the existence of many formal process model-
business processes can be done best using an approach based on ing notations, the majority of the business process community
explicit and executable process models. Such models would al- still uses simple diagrammatic modeling techniques that have
low evaluating performance in terms of flows, calculating costs no potential for performance analysis and/or optimization.
against objectives, recognizing constraints, and evaluating the Table V demonstrates this gap in the lack of reported perfor-
impact of internal and external events [58]. The idea is that, mance analysis and optimization approaches. For most of the
by being able to assess the process execution quality and costs, business process models, there is no structured and repeatable
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

12 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS

TABLE V
OVERVIEW OF BUSINESS PROCESS MODELS, MODELING SETS, ANALYSIS, AND OPTIMIZATION TYPES

improvement technique reported. In terms of process analysis, This could result in a software-based process language with a
there should be a trend to focus on performance analysis as graphical editor to design the process and the necessary tools
it can be directly used for decision-support and further im- to support process analysis and generation of optimized pro-
provement of the process. Performance evaluation needs to be cess designs. Having these capabilities and being supported by
integrated into the design process from the very beginning so a holistic modeling language, business process optimization can
that the objectives of the process can be rationalized from an take on new tracks. Apart from scheduling-like resource alloca-
early stage. Performance indicators are critical for the control tion problems, business process optimization can move toward
and monitoring of a business process. The knowledge extracted the direction of flow optimization and automatic process modifi-
from performance analysis should be fed back to the process in cation. Flow optimization is the notion of defining the optimum
order to improve it. However, there are very few attempts re- path for a business process during execution and according to
ported in the literature to combine performance evaluation and different objectives at a time. Automatic process modification is
process optimization. Regarding the latter, there are some suc- the real-time construction of a business process design accord-
cessful attempts reported, they are highly complicated, and yet, ing to specific needs. The optimum process design is created
address only simple sequential business processes. The lack of based on the selection and combination of different alternative
optimization approaches can be attributed not only to the static activities. Business process optimization has a potential growth
and complex models but also to the unwillingness of business with direct benefit to the business process community, and there
analysts toward “black box” process improvement. are still a lot remaining to be done.
Business process languages combined with diagrammatic de-
pictions of business processes and associated analysis tech-
niques can be used for investigating the properties of processes. VI. CONCLUSION
These techniques can be used to provide a useful insight into the In this paper, a grouping of three sets provided a classification
behavior and characteristics of a business process model speci- for business process models based on their mathematical, dia-
fied using such a language. As these languages support process grammatic, and language characteristics. The advantage of this
execution, process models can be executed a number of times classification is that it allows a modeling technique to be po-
(i.e., simulated) for their properties to be properly investigated. sitioned based on several features simultaneously. The current
This can prove a distinctive advantage of business process lan- trend in business process modeling is the use of diagrammatic
guages for optimization as the simulation results of multiple models that only visualize the business process. These three
executions can provide a guide for optimum definition of the sets also provided a framework for the classification of analysis
process properties. and optimization approaches. The different analysis types were
With process modeling techniques such as IDEF and Petri identified and classified into the three sets. Then, distinctive
nets still popular, what is missing is a modeling technique that references were discussed to demonstrate the link between the
involves elements from all the three modeling sets, and thus, business process models and their analysis capabilities. Business
supports analysis and optimization. This hybrid modeling tech- process analysis has not yet embraced quantitative performance
nique could: 1) support a visual diagrammatic representation of analysis as a way to assess a business process. Simulation can
the process (thus having all the advantages of visualization); 2) provide the necessary means to analyze a business process and
have a formal mathematical underpinning so that quantitative identify its bottlenecks providing a solid basis for improving
measures can be extracted; and 3) can be expressed using a the process. Lastly, business process optimization (as a concept
software-based process language, and thus, allow optimization of automated improvement) was discussed. What was demon-
extensions. These features are provided individually by exist- strated is the lack of support provided by most business pro-
ing modeling techniques; thus, the proposed novel technique cess modeling techniques for structured process improvement.
could incorporate these features and implement the remaining. The few business process optimization approaches reported in
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

VERGIDIS et al.: BUSINESS PROCESS ANALYSIS AND OPTIMIZATION: BEYOND REENGINEERING 13

literature were classified and discussed. Although there are some [23] A. Gunasekaran and B. Kobu, “Modelling and analysis of business pro-
successful attempts recorded, they are highly complicated, and cess reengineering,” Int. J. Prod. Res., vol. 40, no. 11, pp. 2521–2546,
2002.
yet, address only simple sequential business processes. This [24] W. J. Gutjahr, C. Strauss, and E. Wagner, “A stochastic branch-and-bound
paper highlights the necessity for developing business process approach to activity crashing in project management,” INFORMS J. Com-
languages with diagrammatic depictions and associated analy- put., vol. 12, no. 2, pp. 125–135, 2000.
[25] M. Hammer and J. Champy, Reengineering the Corporation: A Manifesto
sis techniques. The future trend would be these hybrid modeling for Business Revolution. London, U.K.: Brealey, 1993.
techniques that would support performance analysis and enable [26] H.-S. Han, “Business process change design from decision model per-
process optimization. spective,” Omega, Int. J. Manage. Sci., vol. 9, no. 2, pp. 21–45, 2003.
[27] E. Hart, P. Ross, and D. Corne, “Evolutionary scheduling: A review,”
REFERENCES Genetic Program. Evolvable Mach., vol. 6, pp. 191–220, 2005.
[28] M. Havey, Essential Business Process Modelling. Sebastopol, CA:
[1] A. F. Abate, A. Esposito, N. Grieco, and G. Nota, “Workflow performance O’Reilly, 2005.
evaluation through WPQL,” in Proc. 14th Int. Conf. Softw. Eng. Knowl. [29] H. Hermanns, U. Herzog, and J.-P. Katoen, “Process algebra for perfor-
Eng., 2002, vol. 27, pp. 489–495. mance evaluation,” Theoretical Comput. Sci., vol. 274, pp. 43–87, 2002.
[2] R. S. Aguilar-Saven, “Business process modelling: Review and frame- [30] I. Hofacker and R. Vetschera, “Algorithmical approaches to business pro-
work,” Int. J. Prod. Econ., vol. 90, pp. 129–149, 2004. cess design,” Comput. Oper. Res., vol. 28, pp. 1253–1275, 2001.
[3] T. A. Aldowaisan and L. K. Gaafar, “Business process reengineering: An [31] Z. Irani, V. Hlupic, and G. M. Giaglis, “Business process reengineering:
approach for process mapping,” Omega, Int. J. Manage. Sci., vol. 27, An analysis perspective,” Int. J. Flexible Manuf. Syst., vol. 14, pp. 5–10,
pp. 515–524, 1999. 2002.
[4] T. Back, U. Hammel, and H. P. Schwefel, “Evolutionary computation: [32] T. Jaeger, A. Prakash, and M. Ishikawa, “A framework for automatic
Comments on the history and current state,” IEEE Trans. Evol. Comput., improvement of workflows to meet performance goals,” in Proc. 6th Int.
vol. 1, no. 1, pp. 3–17, Apr. 1997. Conf. Tools Artif. Intell., Nov. 1994, pp. 640–646.
[5] C. Badica, A. Badica, and V. Litoiu, “A new formal IDEF-based modelling [33] J. Kallrath, “Planning and scheduling in the process industry,” OR Spectr.,
of business processes,” in Proc. 1st Balkan Conf. Inf., Y. Manolopoulos vol. 24, pp. 219–250, 2002.
and P. Spirakis, Eds., Thessaloniki, Greece, 2003, pp. 535–549. [34] W. J. Kettinger, J. T. C. Teng, and S. Guha, “Business process change: A
[6] C. Badica, A. Badica, and V. Litoiu, “Role activity diagrams as finite study of methodologies, techniques and tools,” MIS Q., vol. 21, no. 1,
state processes,” in 2nd Int. Symp. Parallel Distrib. Comput., Oct. 2003, pp. 55–80, 1997.
pp. 15–22. [35] D. E. Knuth, “Computer-drawn flowcharts,” ACM Commun., vol. 6, no. 9,
[7] J. Bal, “Process analysis tools for process improvement,” TQM Mag., pp. 555–563, Sep. 1963.
vol. 10, no. 5, pp. 342–354, 1998. [36] A. Knutilla, S. T. Polyak, C. Schlenoff, A. Tate, S. Ray, S. C. Cheah, and
[8] A. Basu and R. W. Blanning, “A formal approach to workflow analysis,” R. C. Anderson, Process Specification Language: An Analysis of Existing
Inf. Syst. Res., vol. 11, no. 1, pp. 17–36, 2000. Representations. Gaithersburg, MD: Nat. Inst. Standards Technol., 1997,
[9] A. Bellabdaoui and J. Teghem, “A mixed-integer linear programming NISTIR 6133.
model for the continuous casting planning,” Int. J. Prod. Econ., vol. 104, [37] J. Koenig, JBoss jBPM White Paper, version 2004. 2004, 2006.
no. 2, pp. 260–270, 2006. [38] M. Koubarakis and D. Plexousakis, “A formal framework for business
[10] S. Biazzo, “Approaches to business process analysis: A review,” Bus. process modelling and design,” Inf. Syst., vol. 27, pp. 299–319, 2002.
Process Manage. J., vol. 6, no. 2, pp. 99–112, 2000. [39] A. Kusiak, N. T. Larson, and J. Wang, “Reengineering of design and
[11] P. Boekhoudt, H. Jonkers, and M. Rougoor, “Graph-based analysis of busi- manufacturing processes,” Comput. Ind. Eng., vol. 26, no. 3, pp. 521–
ness process models,” in Mathematics and Computers in Modern Science, 536, 1994.
Proc. of the WSES/MIUE/HNA International Conference, N. Mastorakis, [40] Y. H. Lee, K. G. Min, C. Han, K. S. Chang, and T. H. Choi, “Pro-
Ed., Montego Bay, Jamaica, Dec. 2000, pp. 227–235. cess improvement methodology based on multivariate statistical analysis
[12] M. Castellanos, F. Casati, D. Umeshwar, and S. Ming-Chien, “A compre- methods,” Control Eng. Pract., vol. 12, pp. 945–961, 2004.
hensive and automated approach to intelligent business processes execu- [41] A. Levas, S. Boyd, P. Jain, and W. A. Tulskie, “Panel discussion on the
tion analysis,” Distrib. Parallel Databases, vol. 16, pp. 1–35, 2004. role of modelling and simulation in business process reengineering,” in
[13] N. Chapin, Flowcharts. Princeton, NJ: Auerbach, 1971. Proc. IEEE Winter Simul. Conf., Dec. 1995, pp. 1341–1346.
[14] T. H. Davenport, Process Innovation: Reengineering Work Through In- [42] C. Lewis, “A source of competitive advantage,” Manage. Accounting,
formation Technology. Boston, MA: Harvard Business School Press, vol. 71, no. 1, pp. 44–46, 1993.
1993. [43] H. Li, Y. Yang, and T. Y. Chen, “Resource constraints analysis of workflow
[15] R. Dewan, A. Seidmann, and Z. Walter, “Workflow optimization through specifications,” Syst. Softw., vol. 73, pp. 271–285, 2004.
task redesign in business information processes,” in Proc. Hawaii Int. [44] J. Li, Y. Fan, and M. Zhou, “Performance modelling and analysis of
Conf. Syst. Sci., Jan. 1998, vol. 1, pp. 240–252. workflow,” IEEE Trans. Syst., Man, Cybern. A, Syst. Humans, vol. 34,
[16] S. Donatelli, M. Ribaudo, and J. Hillston, “A comparison of performance no. 2, pp. 229–242, Mar. 2004.
evaluation process algebra and generalised stochastic Petri nets,” in Proc. [45] H. Lin, Z. Zhap, H. Li, and Z. Chen, “A novel graph reduction algorithm
6th Int. Workshop Petri Nets Perform. Models (PNPM 1995), p. 158. to identify structural conflicts,” in Proc. 35th Hawaii Int. Conf. Syst. Sci.,
[17] A. T. Ernst, H. Jiang, M. Krishnamoorthy, and D. Sier, “Staff scheduling 2002, vol. 9, p. 289.
and rostering: A review of applications, methods and models,” Eur. J. [46] A. Lindsay, D. Downs, and K. Lunn, “Business processes—Attempts to
Oper. Res., vol. 153, pp. 3–27, 2004. find a definition,” Inf. Softw. Technol., vol. 45, pp. 1015–1019, 2003.
[18] M. M. Fathee, R. Redd, D. Gorgas, and B. Modarres, “The effects of [47] N. Melao and M. Pidd, “A conceptual framework for understanding busi-
complexity on business process reengineering: Values and limitations of ness process modelling,” Inf. Syst., vol. 10, pp. 105–129, 2000.
modelling and simulation techniques,” in Proc. 1998 Winter Simul. Conf., [48] C. A. Mendez, J. Cerda, I. E. Grossmann, I. Harjunkoski, and M. Fahl,
Dec., vol. 2, pp. 1339–1345. “State-of-the-art review of optimisation methods for short-term scheduling
[19] A. Ferscha, “Optimistic distributed execution of business process models,” of batch processes,” Comput. Chem. Eng., vol. 30, no. 6/7, pp. 913–946,
in Proc. 31st Annu. Hawaii Int. Conf. Syst. Sci., Jan. 1998, vol. 7, pp. 723– 2006.
732. [49] C. Moon and Y. Seo, “Evolutionary algorithm for advanced process plan-
[20] C. A. Floudas and X. Lin, “Mixed integer linear programming in process ning and scheduling in a multi-plant,” Comput. Ind. Eng., vol. 48, no. 2,
scheduling: Modelling, algorithms and applications,” Ann. Oper. Res., pp. 311–325, 2005.
vol. 139, pp. 131–162, 2005. [50] M. A. Ould, Business Processes: Modelling and Analysis for Re-
[21] M. Gao, M. Zhou, X. Huang, and Z. Wu, “Fuzzy reasoning Petri nets,” Engineering and Improvement. Chichester, U.K.: Wiley, 1995.
IEEE Trans. Syst., Man, Cybern. A, Syst. Humans, vol. 33, no. 3, pp. 314– [51] L. Peters and J. Peters, “Using IDEF0 for dynamic process analysis,”
324, May 2003. in Proc. 1997 IEEE Int. Conf. Robot. Autom., Albuquerque, NM, 1997,
[22] A. Greasley, “Using business-process simulation within a business- pp. 3203–3208.
process reengineering approach,” Bus. Process Manage. J., vol. 9, no. 3, [52] K. Phalp and M. Shepperd, “Quantitative analysis of static models of
pp. 408–420, 2003. processes,” Syst. Softw., vol. 52, pp. 105–112, 2000.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

14 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS

[53] J. M. Pinto and I. E. Grossmann, “Assignment and sequencing models for [81] K. Vergidis, A. Tiwari, and B. Majeed, “Business process improvement us-
the scheduling of process systems,” Ann. Oper. Res., vol. 81, pp. 433–466, ing multi-objective optimisation,” BT Technol. J., vol. 24, no. 2, pp. 229–
1998. 235, 2006.
[54] S. G. Powell, M. Schwaninger, and C. Trimble, “Measurement and control [82] P. Volkner and B. Werners, “A decision support system for business pro-
of business processes,” Syst. Dyn. Rev., vol. 17, no. 1, pp. 63–91, 2001. cess planning,” Eur. J. Oper. Res., vol. 125, pp. 633–647, 2000.
[55] A. B. Raposo, L. P. Magalhaes, and I. L. M. Ricarte, “Petri nets based [83] K. Wang, A. Salhi, and E. S. Fraga, “Process design optimisation using
coordination mechanisms for multi-flow environments,” Int. J. Comput. embedded hybrid visualisation and data analysis techniques within a ge-
Syst. Sci. Eng., vol. 15, no. 5, pp. 315–326, 2000. netic algorithm optimisation framework,” Chem. Eng. Process., vol. 43,
[56] H. A. Reijers, “Product-based design of business processes applied within pp. 663–675, 2004.
the financial services,” J. Res. Pract. Inf. Technol., vol. 34, no. 2, pp. 110– [84] P. Wohed, W. M. P. van der Aalst, M. Dumas, and A. H. M. ter Hofst-
122, 2002. ede, “Pattern-based analysis of BPEL4 WS,” Queensland Univ. Technol.,
[57] H. A. Reijers and S. Liman-Mansar, “Best practices in business process Brisbane, Australia, QUT Tech. Rep. FIT-TR-2002-04, 2002.
redesign: An overview and qualitative evaluation of successful redesign [85] A. Zakarian, “Analysis of process models: A fuzzy logic approach,” Int.
heuristics,” Omega, Int. J. Manage. Sci., vol. 33, pp. 283–306, 2005. J. Adv. Manuf. Technol., vol. 17, pp. 444–452, 2001.
[58] C. Reyneri, “Operational building blocks for business process modelling,” [86] Y. Zhou and Y. Chen, “Business process assignment optimisation,” in
Comput. Ind., vol. 40, pp. 115–123, 1999. Proc. IEEE Int. Conf. Syst., Man, Cybern., 2002, vol. 3, pp. 540–545.
[59] D. Riehle and H. Zullinghoven, “Understanding and using patterns in [87] Y. Zhou and Y. Chen, “The methodology for business process optimised
software development,” Theory Pract. Object Syst., vol. 2, no. 1, pp. 3– design,” in Proc. Ind. Electron. Conf. (IECON), 2003, vol. 2, pp. 1819–
13, 1996. 1824.
[60] H. J. Rommelfanger, “The advantages of fuzzy optimization models [88] Y. Zhou and Y. Chen, “Project-oriented business process performance
in practical use,” Fuzzy Optim. Decis. Making, vol. 3, pp. 295–309, optimization,” in Proc. IEEE Int. Conf. Syst., Man Cybern., 2003, vol. 5,
2004. pp. 4079–4084.
[61] J. L. Rummel, Z. Walter, R. Dewan, and A. Seidmann, “Activity consolida-
tion to improve responsiveness,” Eur. J. Oper. Res., vol. 161, pp. 683–703,
2005.
[62] W. Sadiq and M. Orlowska, “Analyzing process models using graph re-
duction techniques,” Inf. Syst., vol. 25, no. 2, pp. 117–134, 2000.
[63] K. Salimifard and M. Wright, “Petri net-based modelling of workflow
systems: An overview,” Eur. J. Oper. Res., vol. 134, pp. 664–676, Kostas Vergidis (M’xx) received the B.Sc. de-
2001. gree in applied informatics from the University of
[64] N. Shah, “Single- and multisite planning and scheduling: Current status Macedonia, Thessaloniki, Greece, in 2003, and the
and future challenges,” in Proc. 3rd Int. Conf. Found. Comput.-Aided M.Sc. degree in information technology for prod-
Process Oper., J. F. Pekny and G. E. Blay, Eds., 1998, pp. 75–90. uct realization from Cranfield University, Cranfield,
[65] Y. Shimizu and Y. Sahara, “A supporting system for evaluation and review U.K., in 2005.
of business process through activity-based approach,” Comput. Chem. He served in the Greek Army for a year. He is
Eng., vol. 24, pp. 997–1003, 2000. currently a Doctoral Researcher at Cranfield Univer-
[66] F. Soliman, “Optimum level of process mapping and least cost business sity in the area of business process optimization. His
process re-engineering,” Int. J. Oper. Prod. Manage., vol. 18, no. 9/10, current research interests include business process
pp. 810–816, 1998. definition and modeling, process optimization, and
[67] M. Tinnila, “Strategic perspective to business process redesign,” Bus. multiobjective optimization for large number of objectives.
Process Re-Eng. Manage. J., vol. 1, no. 1, pp. 44–59, 1995.
[68] A. Tiwari, Evolutionary Computing Techniques for Handling Variables
Interaction in Engineering Design Optimisation. Cranfield, U.K.: SIMS,
Cranfield Univ., 2001.
[69] A. Tiwari, K. Vergidis, and B. Majeed, “Evolutionary multi-objective Ashutosh Tiwari (M’04) received the qraduate de-
optimisation of business processes,” in Proc. IEEE Congr. Evol. Comput.,
gree in mechanical engineering from IIT, Kan-
Jul. 2006, pp. 3091–3097.
pur, India, the Master’s and Ph.D. degrees from
[70] G. Valiris and M. Glykas, “Critical review of existing BPR methodologies:
Cranfield University, Cranfield, U.K., in 1999 and
The need for a holistic approach,” Bus. Process Manage. J., vol. 5, no. 1, 2001, respectively. He is currently leading the re-
pp. 65–86, 1999.
search in applied soft computing in the Decision
[71] G. Valiris and M. Glykas, “Business analysis metrics for business process
Engineering Centre, Cranfield University, Cranfield,
redesign,” Bus. Process Manage. J., vol. 10, no. 4, pp. 445–480, 2004. U.K., and has developed new research areas in
[72] W. M. P. van der Aalst, “Petri net based scheduling,” OR Spectr., vol. 18,
the application of soft computing to product and
pp. 219–229, 1996.
process design. He is a member of the Editorial Board
[73] W. M. P. van der Aalst, “The application of Petri-nets to workflow man-
of the Applied Soft Computing Journal and an Asso-
agement,” J. Circuits, Syst. Comput., vol. 8, no. 1, pp. 21–66, 1998. ciate Editor of the International Journal of Applied Fuzzy Sets Theory. He is
[74] W. M. P. van der Aalst, “Re-engineering knock-out processes,” Decision
the author or coauthor of more than 58 research papers in refereed journals,
Support Syst., vol. 30, pp. 451–468, 2001.
conferences, and books, and an edited book on applied soft computing.
[75] W. M. P. van der Aalst, “Business process management: A personal view,”
Bus. Process Manage. J., vol. 10, no. 2, pp. 135–139, 2004.
[76] W. M. P. van der Aalst, A. Hirnschall, and H. M. W. Verbeek, “An alter-
native way to analyse workflow graphs,” in Lecture Notes in Computer
Science, Springer-Verlag, 2002, vol. 2348, p. 535.
[77] W. M. P. van der Aalst and A. H. M. ter Hofstede, “Workflow patterns: On Basim Majeed (M’01) received the Master’s and
the expressive power of (Petri-net-based) workflow languages,” in Proc. Ph.D. degrees in intelligent control systems from the
4th Workshop Practical Use Coloured Petri Nets CPN Tools, Aarhus, University of Manchester, Manchester, U.K., in 1987
Denmark, 2002, pp. 1–20. and 1992, respectively.
[78] W. M. P. van der Aalst and A. H. M. ter Hofstede, “YAWL: Yet an- He is currently a Principal Research Professional
other workflow language (revised version),” Queensland Univ. Technol., at the Intelligent Systems Research Centre, BT Re-
Brisbane, Australia, QUT Tech. Rep., FIT-TR-2003-04, 2003. search and Venturing, Martlesham, U.K. He is en-
[79] W. M. P. van der Aalst, A. H. M. ter Hofstede, and M. Weske, “Business gaged in the area of real-time business intelligence
process management: A survey,” in Lecture Notes Computer Sciences, and business process management.
Springer-Verlag, 2003, vol. 2678, pp. 1–12. Dr. Majeed is a Member of the Institution of En-
[80] W. M. P. van der Aalst and K. M. van Hee, “Business process redesign: A gineering and Technology and is also a Chartered
Petri-net-based approach,” Comput. Ind., vol. 29, pp. 15–26, 1996. Engineer.

View publication stats