Bus Inf Syst Eng 63(3):277–300 (2021)

https://doi.org/10.1007/s12599-020-00649-w

RESEARCH PAPER

Process Mining for Six Sigma

A Guideline and Tool Support

Teun Graafmans • Oktay Turetken • Hans Poppelaars • Dirk Fahland

Received: 24 February 2019 / Accepted: 19 February 2020 / Published online: 27 April 2020
 The Author(s) 2020

Abstract Process mining offers a set of techniques for guideline for practitioners by extending the DMAIC-based
gaining data-based insights into business processes from standard operating procedure. PMSS can help increasing
event logs. The literature acknowledges the potential ben- the efficiency and effectiveness of Six Sigma-based process
efits of using process mining techniques in Six Sigma- improving efforts. This work extends the body of knowl-
based process improvement initiatives. However, a guide- edge in the fields of process mining and Six Sigma, and
line that is explicitly dedicated on how process mining can helps closing the gap between them. Hence, it contributes
be systematically used in Six Sigma initiatives is lacking. to the broad field of quality management.
To address this gap, the Process Mining for Six Sigma
(PMSS) guideline has been developed to support organi- Keywords Process mining
Six Sigma
Define-measure-
zations in systematically using process mining techniques analyze-improve-control
DMAIC
Design science
aligned with the DMAIC (Define-Measure-Analyze-Im- research
Process mining for Six Sigma
PMSS
prove-Control) model of Six Sigma. Following a design
science research methodology, PMSS and its tool support
have been developed iteratively in close collaboration with 1 Introduction
experts in Six Sigma and process mining, and evaluated by
means of focus groups, demonstrations and interviews with Quality management helps companies to improve their
industry experts. The results of the evaluations indicate that business processes, performance, and competitiveness
PMSS is useful as a guideline to support Six Sigma-based (Flynn et al. 1995; Samson and Terziovski 1999). The
process improvement activities. It offers a structured positive benefits of quality management have persuaded
many companies to implement quality management sys-
tems (Peris-Ortiz and Alvarez-Garcia 2014). This has led to
Accepted after two revisions by Jörg Becker. the emergence of a number of process improvement related
quality management methods or frameworks, such as Six
T. Graafmans
O. Turetken (&)
H. Poppelaars Sigma, Total Quality Management (TQM), business pro-
Department of Industrial Engineering and Innovation Sciences,
cess reengineering/management, and relevant standards,
Eindhoven University of Technology, Eindhoven, The
Netherlands such as the ISO 9000 family of quality management system
e-mail: o.turetken@tue.nl standards (ISO 2015; Evans and Lindsay 2016).
T. Graafmans One of the quality management frameworks that has
e-mail: teungraafmans@gmail.com seen a significant increase in usage over the last few dec-
H. Poppelaars ades is Six Sigma. Six Sigma uses a set of quality princi-
e-mail: hans.poppelaars@processgold.com ples and techniques to minimize the number of defects in a
process (George 2002; Pyzdek 2003). It was originally
D. Fahland
developed and used by Motorola in the 1980s and it has
Department of Mathematics and Computer Science, Eindhoven
University of Technology, Eindhoven, The Netherlands now branched out to many industry sectors (Pande and
e-mail: d.fahland@tue.nl Holpp 2002; Tjahjono et al. 2010). The objective of Six

123
278 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

Sigma is to identify and remove errors in business pro- In order to address this research objective, we followed
cesses and, thereby, the causes of variability in processes. the design science research (DSR) methodology (Gregor
In order to do so, Six Sigma practitioners often use a and Hevner 2013) to develop a suitable guideline as a
project methodology known as ‘Define-Measure-Analyze- design artifact, namely Process Mining for Six Sigma
Improve-Control’ (DMAIC) (Pande and Holpp 2002). (PMSS). Following a structured literature review, we
There is an abundance of research on the use and identified as main requirements that PMSS ought to support
effectiveness of Six Sigma in several business domains, practitioners in the effective use of process mining tech-
such as healthcare (Bedgood 2017; Antony et al. 2018), niques in their Six Sigma initiatives. PMSS guides orga-
construction (Siddiqui et al. 2016; Karakhan 2017), supply nizations through different stages of process improvement
chain management (Madhani 2016; Hong 2017), public and offers a direct link between different phases of the
sector (Antony et al. 2017), and education (LeMahieu et al. DMAIC model of Six Sigma and existing process mining
2017; Lu et al. 2017; Nadeau 2017). activities. To help ensure the utility, PMSS was developed
For reducing variations in business processes, Six Sigma in close collaboration with Six Sigma and process mining
often employs statistical methods to quantify changes in experts, both researchers and practitioners.
process performance. However, the data used for such an Following the DSR method, the initial version of PMSS
analysis is usually collected manually through, for was developed by taking an existing process mining
instance, surveys and observations, which makes Six methodology as a basis. The initial version was refined
Sigma a costly and time-consuming endeavor (Park and based on the insights gathered from domain experts in
Kang 2016; van der Aalst 2016). Moreover, such approa- semi-structured interviews. In addition, a prototypical tool
ches often involve subjectivity as they rely heavily on the support, which is integrated into a commercial platform,
knowledge of those involved in the analysis, who often was developed to complement and further increase the
reflect and confirm their own view of the processes and usability of the guideline. The adapted version of PMSS
relevant problems (Johannsen et al. 2011). It is also diffi- and the tool support were refined iteratively using expert
cult to capture all the complexity and variations in process interviews, focus group demonstrations, and usability tests.
executions using such approaches (Fluxicon 2019). The validity of the final version of PMSS and the tool
Information systems and devices record and store large support were evaluated through a series of demonstration
amounts of data (Bücker et al. 2016). Specifically, event sessions with both process mining and Six Sigma experts.
data provides records of the operational steps that have The results of our qualitative evaluation show that PMSS is
been executed in the past. Process mining is an emerging considered an easy to use and useful guideline for orga-
discipline that offers a set of techniques to gain data-based nizations when performing Six Sigma activities.
insights from recorded event data and to conduct further The remainder of this paper is structured as follows:
analysis in order to support process improvements (van der Sect. 2 gives a brief background of the Six Sigma frame-
Aalst 2016). Process mining acts, on the one hand, between work and the concept of process mining. Section 3 dis-
computational intelligence and data mining, and, on the cusses related work on the use of process mining
other hand, between process modeling and analysis, to techniques in Six Sigma initiatives. Section 4 presents the
serve as an enabling technology for process-oriented research design that we followed when developing,
quality management methods and frameworks (van der demonstrating, and evaluating PMSS. In Sect. 5, the final
Aalst et al. 2012a, b; Harmon 2018). version of PMSS, including the tool support, is briefly
The existing literature confirms process mining’s described. The evaluation of PMSS by means of the
potential to enrich the set of techniques that practitioners demonstration sessions and interviews is discussed in
can use in Six Sigma programs (van der Aalst et al. 2016; Sect. 6. Finally, in Sect. 7, we conclude with the discus-
Valle et al. 2017; Garcia et al. 2019). However, current sions of contributions, limitations and future research
research in this field provides very limited guidance in this directions.
direction (van Geffen and Niks 2013; van der Aalst 2016).
In particular, a structured method or a guideline is missing
that helps organizations to discover how and when process 2 Background
mining techniques can be used to support Six Sigma
activities. This leads to the following research objective of 2.1 Six Sigma
this study:
The term Six Sigma refers to a set of tools, techniques, and
To develop a guideline to support organizations in
methods which aim to improve the quality of processes
systematically using process mining techniques
within businesses (George 2002; Pyzdek 2003). It was
aligned with the DMAIC model of the Six Sigma.
developed and introduced by Motorola in the early 1980s

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 279

and afterward adopted by many other companies (Tjahjono discovery, conformance checking, and process
et al. 2010). In order to minimize the number of defects enhancement.
(this refers to the cases that do not produce a desired out- Process discovery takes as input an event log and gen-
come), Six Sigma aims to improve business processes in erates a corresponding process model. Various techniques
such a way that the standard deviation of a process is so can be used for extracting a process model from the raw
small that any value within six standard deviations of the event log (van der Aalst and Dustdar 2012). An example is
mean can be considered as non-defective. the inductive algorithm, which discovers a set of sound
The r in Six Sigma refers to the standard deviation of a block-structured process models from any given event log
normal distribution. The values that lie within a single (Leemans et al. 2013). For a more complete overview of
standard deviation of the mean belong to the one-sigma different process discovery techniques, the reader is
level. A process that runs at one sigma has less than referred to (De Weerdt et al. 2012; Augusto et al. 2019).
690,000 defective cases per million cases – in other words, Process discovery techniques can be used to generate the
at least 31% of the cases are handled properly (Pyzdek structure of the process, to find out about the routing
2003). At two-sigma level, this value goes up to 69.2%. probabilities, to determine the most frequent path in the
The higher the sigma level, the lower the defective cases. process, and to discover the distribution of cases over paths
Motorola eventually settled on the six-sigma level, which (Ailenei et al. 2012).
indicates that 99.99966% of the cases are handled correctly Conformance checking compares a designed process
and thus the process has as few as 3.4 defective cases per model with the corresponding event log. This comparison
million cases. shows where the assumed process model deviates from the
In order to minimize the sigma of a process to such an real-world process as retrieved from the event logs (van der
extent that the process runs at six-sigma, practitioners can Aalst and Dustdar 2012; Carmona et al. 2018). Domain
use a myriad of tools and techniques. These tools and knowledge is often required to locate and explain these
techniques are most often applied through the DMAIC deviations and measure their severity (van der Aalst 2010).
project methodology (Define-Measure-Analyze-Improve- Typical conformance checking use cases include finding
Control) (De Mast and Lokkerbol 2012). The steps of the exceptions from the normal path, determining the degree to
DMAIC model aim to Define the goals of the improvement which the business rules and regulations are followed, and
activity, Measure the existing process to establish a base- measuring the level of compliance to a reference model
line on how the process currently performs, Analyze it to (Ailenei et al. 2012).
identify improvement opportunities, Improve the process Process Enhancement helps to extend or improve an
accordingly, and Control it to ensure that the improvement existing process model with additional information about
was effective (Pyzdek 2003). These steps are looked at in the process present in the event log. Where conformance
further detail in Sect. 5.1. checking is limited to measuring the alignment of the
process and the event log, the goal of enhancement is to
2.2 Process Mining improve or extend the process to gain additional insights
(van der Aalst et al. 2012b). For instance, process mining
Process mining has emerged as a research field that tools can extend a model to show bottlenecks, service
incorporates a set of techniques which enable an organi- levels, throughput times, and frequencies (van der Aalst
zation to gain data-based insights into their processes and and Dustdar 2012; van der Aalst et al. 2012b; van der Aalst
support process improvements (van der Aalst 2016). The 2016).
starting point for Process Mining is an event log. An
information system controls real-world business processes
and records events which are stored in the event logs. An 3 Related Work
event refers to an activity (i.e., a defined step in a process)
which took place at a particular time, and is related to a In this section, we provide a brief overview of the
particular case (i.e., process instance) (van der Aalst et al. methodologies proposed to guide process mining and
2012b). Additional information, such as the resource (e.g., improvement projects, and present the results of our sys-
person or device) executing the activity, or cost incurred in tematic review of relevant works that discuss the applica-
the execution for a single instance of the activity, can also tion of process mining techniques in Six Sigma initiatives.
be stored in the event log (van der Aalst et al. 2012b; van
der Aalst 2016). The process mining techniques can be 3.1 Process Mining Methodologies
used to analyze these event logs and generate fact-based
representations of business processes. Three classes of In order to structurally implement the different types of
analysis can be distinguished in process mining: process process mining in a process improvement project,

123
280 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

researchers have sought to provide guidance by developing libraries in which the keyword search was conducted. The
process mining methodologies, such as the Process Diag- following keyword string was applied to all fields of the
nostics Method (Bozkaya et al. 2009), the Methodology for publications (i.e., title, keywords, abstracts, and full text
Business Process Analysis in Healthcare (Rebuge and where available): ‘‘Six Sigma’’ and ‘‘process mining’’
Ferreira 2012), the Process Mining Methodology for The initial search resulted in 941 publications in total
Emergency Room Processes (Rojas et al. 2017), the Life- (including duplicates that appear in multiple libraries). In
cycle Model (L*) (van der Aalst 2016), and the Process order to identify the studies that are directly relevant for
Mining Project Methodology (PM2) (van Eck et al. 2015). our research topic, we defined inclusion and exclusion
The first three methodologies are to a large extent for the criteria. For the context, we included papers that are pub-
specific needs of the healthcare domain. The L* covers a lished (1) in English, and (2) in an academic journal,
large set of mining techniques, while focusing mainly on conference proceeding, or a chapter in a scientific book.
the discovery and analysis of structured processes. We excluded publications in the grey literature, i.e., those
PM2 aims to overcome the limitations of the former without bibliographic information (such as publication
methodologies by providing a framework covering the date/type, volume and issue numbers), working papers,
majority of the mining techniques that can serve as guid- white papers, books, or reports. As for the relevance, we
ance for process mining initiatives in diverse business included papers that discuss the use of process mining in
domains. The methodology incorporates the following six sigma initiatives or present a work that shows how
steps: planning, data extraction, data processing, mining & process mining techniques were applied in such initiatives.
analysis, evaluation, and process improvement & support. A thorough review of all publications resulted in 16
The steps of data processing, mining & analysis, and primary studies. Figure 1 presents the distribution of these
evaluation, take place in several iterations. Although PM2 studies by type and year.
provides guidance for the general use of process mining In the list of primary studies, we have distinguished
techniques, it does not explicitly consider the use of pro- between two categories of studies. The first category of
cess mining in a larger context of quality management – in studies endorses the use of process mining techniques in
particular, in the context of process-improvement focused Six Sigma initiatives and discusses the benefits. Ten out of
management strategies, such as Six Sigma or Total Quality 16 studies belong to this category: (Hess 2006; van der
Management (van Cruchten and Weigand 2018). Aalst and Dustdar 2012; van der Aalst et al. 2012b, 2016;
Aguirre et al. 2013; Tomašević and Slović 2013; Sebu and
3.2 Six Sigma and Process Mining Ciocarlie 2014, 2015; Park and Kang 2016; Valle et al.
2017).
In order to gain an accurate understanding of the state-of- The second category of studies are relatively recent
the art research on the works that discuss the application of publications that present discussions of the application of
process mining techniques in Six Sigma initiatives, we process mining techniques in Six Sigma initiatives. In
conducted a structured literature review (SLR) following addition to discussing the benefits, these studies propose
the guidelines proposed by Kitchenham and Charters approaches to how mining techniques can be applied
(2007). We applied a keyword search strategy over a set of within the DMAIC model. Six out of 16 publications are of
academic digital libraries to locate relevant studies pub- this type: (van Geffen and Niks 2013; Dogan and Gurcan
lished until October 2019. Table 1 shows the list of digital 2018; van Kollenburg and Wouters 2018; Boersma et al.
2019; Dahlin et al. 2019; Gupta et al. 2019). In the para-
graphs below, we discuss these works in more detail.
The study by van Geffen et al. (2013) presents a case
Table 1 Digital libraries and search results study that discusses how process mining techniques can be
Digital library No. of relevant publications used to support different phases of the Six Sigma DMAIC
model. It shows that using process mining techniques
ACM Digital Library 4
within the DMAIC model in a particular organizational
AIS Electronic Library 20
setting can decrease a complete rundown duration from 9
IEEE Xplore 22
to 12 weeks to 4-6 weeks. Although the study confirms the
Google Scholar 699 potential benefits of using process mining as a tool in a Six
SCOPUS 96 Sigma program, it does not aim to introduce a structured
Science Direct 24 guideline on how process mining can be methodically used
Springerlink 73 within the phases of the DMAIC model.
Web of Science 3 In the study by Dogan and Gurcan (2018), the authors
Total (with duplicates): 941 discuss the applicability of specific classes of process

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 281

Fig. 1 Distribution of primary

studies by year and type

mining techniques in the DMAIC phases. While the pro- achieved when they are combined. Consequently, the
cess discovery techniques are considered useful in the authors call for future research that should investigate how
Define and Analyze phases, conformance checking is process mining can be integrated into organization-wide
deemed applicable in the Measure and Control phases. For quality improvement initiatives.
the Improve phase, enhancement techniques are proposed. Finally, Gupta et al. (2019) discuss the role of data
van Kollenburg and Wouters (2018) also argue for the analytics techniques – including process mining – in the
usefulness of process mining in continuous and sustainable DMAIC cycle. Similar to the proposition by Dogan and
improvements, and propose applications of process mining Gurcan (2018), the process discovery techniques are con-
in the DMAIC cycle. For the Define phase, the authors sidered useful in the Define phase, while conformance
argue that the goals and scope of the project can be checking is regarded applicable in the Measure phase.
determined using the real data from process executions. In brief, the existing body of research considers the use
Process mining offers a quick and objective analysis of of process mining techniques within Six Sigma initiatives
process performance in the Measure and Analyze phases. It as useful and sound, and the linking of such techniques to
helps to identify opportunities for the Improve phase. existing concepts and methods beneficial are found to
Finally in the Control phase, using process mining as an provide an enriched understanding of processes. Therefore,
established approach in the daily management of opera- a structured guideline that designates how process mining
tions gives the opportunity to sustain the improvements. can be systematically used along the DMAIC cycle would
Boersma et al. (2019) discuss how the discovery, con- facilitate the organizational efforts to realize the claimed
formance and enhancement techniques of process mining benefits.
can be used in the DMAIC cycle in the healthcare domain.
For the initial phases, they show the efficiency of process
mining techniques over qualitative methods, such as 4 Research Design
interviews. In the Analyze phase, conformance checking is
proposed for checking the extent to which the existing The main goal of this research is to develop a new artifact
processes comply with internal and external guidelines. In that supports Six Sigma practitioners to perform their
the Control phase, process mining is considered useful for activities with the use of Process Mining techniques. This
checking adherence to the new (improved) processes and artifact is referred to as Process Mining for Six Sigma
identifying deviations. Enhancement techniques that enrich (PMSS) and consists of three parts: a high-level graphical
the discovered process model with additional perspectives overview, explanatory text in the form of tables, and tool
(such as the time, cost, and resource utilization), are pro- support.
posed for the Control phase to facilitate in-depth The design science research (DSR) methodology is
monitoring. well-suited for studies that aim to develop and evaluate
Dahlin et al. (2019) propose a set of steps depicting how such artifacts (Gregor and Hevner 2013). We followed the
process mining can be applied to improve healthcare pro- DSR process proposed by Peffers et al. (2007) that includes
cesses. Although this study does not refer to the DMAIC the following activities (executed in this nominal
model, it discusses how process mining can be compared to sequence): identifying the problem, defining requirements
the process mapping technique – as an approach commonly of a solution, designing and developing the artifact,
used in Six Sigma initiatives – and what benefits can be demonstrating the artifact in a suitable context, and

123
282 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

DESIGN & DEVELOPMENT DEMONSTRATION & EVALUATION

Development
PROBLEM of the inial PMSS version 1 Interviews with
IDENTIFICATION version of Experts
1 the PMSS 3 4
Literature On
Process Mining Feedback
Methodologies Adapng
and Six Sigma PMSS 5

Design and
PMSS version 2
Development Brainstorm sessions
Soluon of the Tool with Experts
Reqs. PMSS tool reqs.
Support 6 7
DEFINING
REQUIREMENTS
OF THE PMSS + Tool
SOLUTION Support v2 Focus Group session
Need for
addional with Experts
Reqs. 8
2
Feedback

PMSS + Tool
Enhancing
Support v3 Usability Tests with
PMSS and
Experts
Tool Support
9 10
Feedback

Finalizing Demonstraon
PMSS + Tool Evaluaon of ulity
PMSS and Support v4 (final)
Sessions with
using a Survey
Tool Support Experts
11 12 13

Fig. 2 Research process

evaluating it with respect to a set of criteria. In following guideline and tool support were enhanced accordingly (step
this process, we have performed a number of iterations 9). This was followed by the usability tests with domain
between the design & development, and demonstration. experts, focusing on the usability of the tool support (step
The process that we followed is depicted in Fig. 2. 10). Based on the feedback gathered, PMSS was finalized
As a first step, we conducted a literature review on Six (step 11).
Sigma, on process mining methodologies, and on the use of The interviews and usability tests took place during the
related techniques in Six Sigma initiatives (step 1). Driven construction of PMSS. Therefore, they were ex-ante eval-
by the findings of the first step, the requirements of the uations; that is, they involved the assessment of an un-
solution were identified (step 2). An initial version of instantiated artifact to show its potential (Venable et al.
PMSS was developed based on these requirements (step 3). 2016), (Venable and Pries-heje 2012).
Through interviews, feedback on the initial and interme- The final version of PMSS was demonstrated to both Six
diate versions of PMSS was obtained (step 4), which was Sigma and process mining experts, who were then inter-
later used to iteratively adapt the guideline (step 5). In viewed to provide feedback on the utility of PMSS (step
addition, the tool support was developed for PMSS (step 6) 12). To evaluate the utility (i.e., how useful and easy-to-use
based on the input gathered from the interviews and from the practitioners consider the artifact is), they were asked to
the brainstorm sessions we organized with process mining respond to a questionnaire that involves Technology
and Six Sigma experts (step 7). Acceptance Model (TAM) constructs (step 13). The inter-
The adapted version of the guideline (v2) together with views and survey took place after PMSS was developed;
the tool support was demonstrated to the experts for thus, they were ex-post evaluations (Venable et al. 2016).
gathering feedback to further enhance PMSS (step 8). The

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 283

In the next sub-sections, we provide more detail on the R4. PMSS shall take business domain knowledge into
steps of the research design. account
R5. PMSS shall be domain-independent
4.1 Problem Identification R6. PMSS shall support iterative process analysis
Furthermore, since Six Sigma practitioners will use
The literature acknowledges the applicability and potential
process mining techniques to support their DMAIC activ-
usefulness of process mining techniques in the DMAIC
ities, a complementary tool support for the guideline and
cycle of Six Sigma initiatives. As we have indicated above,
use of these techniques is necessary. Incorporating all
however, the existing literature lacks a method or a
functionality within a single platform would help reducing
guideline that is explicitly dedicated to how these tech-
its complexity (Lepofsky 2015) and thus make it easier for
niques can be systematically used in the DMAIC cycle.
Six Sigma practitioners to use process mining techniques.
This research gap drove the formulation of our research
To fulfill this need, we defined the following requirement:
objective as presented in the first section of this paper. This
study aims to develop a guideline that shows how DMAIC R7. PMSS shall include complementary tool support
activities of the Six Sigma initiatives can be supported with which helps Six Sigma practitioners to use available
process mining techniques, thereby bridging the gap and appropriate process mining techniques in a
between Six Sigma and the field of process mining. single platform

4.2 Define Requirements of the Solution

4.3 Develop the Initial Version of PMSS
To fulfill the above mentioned research gap, we identified
In Sect. 3, we discuss the process mining methodologies
the requirements of the solution that can address our
that are proposed in the academic literature to provide
research objective (Peffers et al. 2007). Since the DMAIC
guidance in the use of mining techniques. However, these
model is the de-facto methodology used in improvement
methodologies lack explicit consideration of a larger
projects by Six Sigma practitioners (Pyzdek 2003), the
organizational objective of quality management and also
guideline should focus on explaining how process mining
lack an evident link between these methodologies and the
techniques can be aligned with the phases of the DMAIC
phases of the quality management frameworks – in our
model. Furthermore, the activities that can be taken within
case, Six Sigma. Therefore, for the initial version of PMSS,
each step in PMSS should be explained in order to provide
we decided to take a process mining methodology with
a systematic guideline for practitioners. To address these
guidance of techniques as a basis and map it to the phases
needs, the following requirements are specified:
of the DMAIC model of Six Sigma to align it with the
R1. PMSS shall be aligned with DMAIC and show phases and objectives of quality management.
where in the model process mining techniques can For this initial version, we took the process mining
be employed methodology of PM2 as a basis, since it was developed as
R2. The tasks incorporated within PMSS shall be an improvement of former methodologies (van der Aalst
thoroughly explained in order to effectively guide 2016), (van Eck et al. 2015). PM2 is domain-independent,
their enactment by practitioners has an explicit consideration of the business level
requirements, and is designed to support iterative analysis
In order to effectively use process mining techniques in
of processes. Hence, we mapped the PM2 activities to the
each DMAIC phase, PMSS should take a process mining
phases of the DMAIC model (as given in Table 2).
methodology as a basis and align it with the DMAIC
As a next step, we performed a series of interviews to
phases. In addition, PMSS should be business-driven to
evaluate (ex-ante) the initial version of PMSS – i.e., the
support (relatively) large and complex projects where the
mapping between DMAIC and PM2 – as described in the
domain knowledge is limited (Suriadi et al. 2013). The
next section.
guideline should also be domain-independent rendering it
to be applicable to any process, and should support itera-
4.4 Interviews with Experts
tive analyses of processes (Suriadi et al. 2013; van Eck
et al. 2015; Dijkman et al. 2019). PMSS should, therefore,
We conducted interviews with 13 field experts, with the
fulfill the following requirements:
objective of investigating how the domain experts would
R3. PMSS shall be based on a process mining decide to use process mining in the phases of the DMAIC
methodology model. A second goal was to evaluate the extent to which
the initial version of PMSS was aligned with their views

123
284 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

Table 2 DMAIC and PM2 activities mapped considered either experts or knowledgeable in Six Sigma
DMAIC Model 2
PM activities
(note that a number of interviewees had considerable
experience and knowledge in both fields). The interviewees
Define Planning had at most 12 and as a minimum 2 years of experience in
Data extraction (preliminary) their main field of expertise.
Data processing (preliminary) The interviews were semi-structured; that is, they fea-
Mining and analysis (exploratory) tured a number of predefined questions as well as room to
Measure Data extraction deviate for open discussions. The questions were driven by
Data processing a case, which involved a realistic process and a scenario
Analyze Mining and Analysis (explanatory) where the interviewee was expected to go through the
Improve Process improvement and support DMAIC phases for improving the process. While going
Mining and Analysis (exploratory) through an example case, the experts were asked to indi-
Control Evaluation cate in detail various process mining techniques that they
would consider employing at each DMAIC phase. The final
part of the discussions with the experts focused on:
and can effectively provide structural guidance in the use
of these techniques. • whether the PMSS can be useful in practice,
Conducting interviews with a group that has a wide • whether its components are correctly structured,
coverage of expertise in related topics enhances the internal • whether it is complete in terms of the components, and
validity of the semi-structured interviews (Gibbert and • whether the level of granularity in the description of
Wicki 2008). Therefore, the interviewees were selected so components is sufficient to guide its use.
that they had expertise in at least one of the two fields, i.e.,
Based on the feedback retrieved from the experts, PMSS
process mining and Six Sigma, to ensure capturing the
was updated and refined as a second version (v2). This
views from both fields. We also distinguished experts with
version incorporated a high-level visual overview of
respect to their roles in related fields (i.e., Six Sigma
adapted PM2 activities mapped to DMAIC phases and
practitioner, process mining tool developer, consultant, and
explanatory text describing what exactly each activity
C-level/project manager, academic, etc.). The interviewee
entails in terms of the process mining techniques and tasks.
profile with respect to these roles and their level of
(We describe the final version of PMSS in Sect. 5.) The
expertise in related fields are presented in Table 3.
feedback from experts gathered at this stage also served as
As depicted in the table, 11 interviewees considered
a point of departure for the design and development of
themselves experts in process mining, while 8 were
PMSS tool support as discussed in the next sub-section.

Table 3 The interviewees and their level of expertise in related fields

Interviewees Level of expertise Years of experience in the (main) field of expertise
Process mining Six Sigma

SSP1 – Six Sigma Practitioner 1 (master black-belt) ? ?? 12

SSP2 – Six Sigma Practitioner 2 (black-belt) ? ?? 2
ACA1 – Academician 1 ??  11
ACA2 – Academician 2 ?? ? 11
PMD1 – Process mining (PM) Tool Developer 1 ??  12
PMD2 – Process mining Tool Developer 2 ??  2
PMC1 – Project Consultant 1 ?? ? 3
PMC2 – Project Consultant 2 ?? ? 2
PMC3 – Project Consultant 3 ?? ?? 6
PMM1 – Project (C-level) Manager 1 ??  12
PMM2 – Project (C-level) Manager 2 ?? ? 10
PMS1 – System Sales Representative 1 ??  5
PMS2 – System Sales Representative 2 ?? ?? 4
‘??’: Expert, ‘?’: Knowledgeable, ‘’: Not familiar

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 285

The interviews were recorded in order to reduce the risk development was based on the use of the process mining
of their subjective interpretation (Blumberg et al. 2008). techniques.
This made it possible to transcribe interviews into text
without leaving certain claims unnoticed, and provided the 4.6 Focus Group session with Experts
opportunity for the interviewees to review and approve the for Demonstration and Feedback
points made (Dul and Hak 2008).
PMSS (v2) and the first prototype of the tool were
4.5 Design and Development of the Tool Support, demonstrated in a focus group setting to a group of experts
and Brainstorm Sessions in order to gather feedback about how PMSS and the
prototype tool can further be improved. The group con-
The feedback from the interviewees allowed us to identify sisted of 24 experts with a range of years of experience and
the classes of process mining techniques and particular fields of expertise in relation to process mining, Six Sigma
graphical indicators that can be useful at each phase of the and process improvement in general. The majority of the
DMAIC model and which formed the requirements to be group members (22) were practitioners in the process
fulfilled by PMSS tool support. To address these require- improvement field (i.e., process mining tool developers and
ments, we iteratively designed and developed the tool sales representatives, process improvement consultants and
support, and conducted a number of brainstorm sessions C-level managers) and 2 were academics with expertise
with 2 experts on process mining and data visualization to both in process mining and Six Sigma. The 2-h session
ensure that the tool support was relevant and addressed the with the group resulted in several points of improvement
requirements set forth. The choice for the brainstorming that were incorporated in a new version (v3) of PMSS and
method was motivated by its recognized potential to gen- its tool support.
erate ideas, find solutions to specific problems, and support
conceptual user interface designs by generating alternatives 4.7 Usability Tests with Domain Experts
(Wilson 2013).
The first expert had 12 years of experience in process To assess the usability of PMSS and the tool support, we
mining and 17 years in data visualization, while the second conducted usability tests with 12 practitioners. A usability
expert had 4 years of experience in data visualization and test aims to gather a better understanding of how real users
3 years in process mining. Both experts were also knowl- interact with a product and to improve the product based on
edgeable about Six Sigma programs and the DMAIC the results (Nielsen 1994). It aims to identify areas where
model. Together with the experts, we identified the features potential users struggle with a product and thus provide
and related user interfaces that should be supported by the feedback for improving its design. Several types of
tool. In particular, we identified the process mining tech- usability tests have been distinguished in the literature, and
niques that should be supported and should be made we chose to conduct informal usability tests, as they are
available to the users at each DMAIC phase. The experts suitable for relatively small-scale prototypical implemen-
were involved during the development to help ensure that tations (Barnum 2011). These informal tests are in the
the requirements originated from the expert interviews in category of formative usability testing and provide the
the previous step were addressed, and that the tool support product developers with a list of findings to analyze and fix,
would be useful for its intended audience. in addition to features that users liked.
There are a number of process mining tool platforms Among the 12 participants of these tests, 5 were Six
available for use in practice and research (van der Aalst Sigma experts (black belts), while the others had expertise
2016). Therefore, instead of developing a new tool to on process mining and improvement in general. The
support specific process mining techniques, we took an average number of years of experience of these experts
existing platform – ProcessGold, that features several were 4 years for Six Sigma experts (maximum of 14, and
process mining techniques – and extended it with addi- minimum of 1 year) and 7 years (maximum of 17, and
tional (functional and visual) layers to support PMSS. The minimum of 2 years) for process mining and improvement
ProcessGold platform (https://processgold.com/en/) is an experts.
enterprise solution for process mining and can be used to Aligned with the usability test protocol (Barnum 2011),
develop web-based process mining applications. That the participants – in individual sessions – went through a
makes it an appropriate platform for the purpose of this task-based scenario that was created for these tests. The
research. In Sects. 5.2 and 5.3, we elaborate the tool sup- scenario incorporated a realistic case of a process
port with a number of user interface elements that present improvement setting with specific steps that implicate the
relevant and applicable graphical indicators whose use of PMSS and its tool support. Adopting a think-aloud
process, we encouraged participants to share their thoughts

123
286 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

while going through the scenario and performing the steps Sigma practitioners in using process mining to support the
using PMSS and its tool support. DMAIC of Six Sigma, and potentially make it more effi-
By going through the scenario, the participants provided cient/effective in order to perform related DMAIC activi-
feedback on the elements of PMSS and elaborated on its ties. To do so, the steps of PMSS have been aligned with
usability. Based on this feedback, PMSS and the tool the phases of the DMAIC model (fulfilling R1). However,
support were finalized (v4). The feedback from each par- as a result, the steps no longer flow in the same order as in
ticipant was recorded and incorporated in the functionali- PM2.
ties to be supported by the tool and in its user interfaces. As depicted in Fig. 3, PMSS consists of eight steps in
We present the results of these efforts in the description of the main flow that are explicitly positioned within the five
the tool support (Sect. 5.2) and its demonstration phases of DMAIC: (Define) Planning, preliminary data
(Sect. 5.3). The next section focuses on the final version of preparation, exploratory mining & analysis, (Measure) data
PMSS. preparation, (Analyze) explanatory mining & analysis,
(Improve) process improvement, and (Control) monitoring
and evaluation. These steps and phases are depicted in
5 PMSS and the Tool Support sequence with a feedback loop connecting the end phase to
the first to represent the cyclic nature of the DMAIC
In this section, we first present an overview of PMSS and model. However, this depicts an ideal flow and PMSS
discuss the activities that need to be taken for each presumes iterations between steps and phases (fulfilling
guideline step. In Sect. 5.2, we briefly introduce the tool R6) to reflect the case in real-life Six Sigma initiatives.
support and we illustrate the application of PMSS on an Inheriting the properties of PM2, PMSS is designed to be
example business case in Sect. 5.3. applicable for the improvement of processes in any busi-
ness domain (fulfilling R5).
5.1 Elements of PMSS The guideline contains a resource layer showing the
roles that act as the leading role for each step. Three roles
A graphical representation of PMSS is shown in Fig. 3. are at the core of PMSS: Data analysts, process analysts,
The bottom part serves as a legend. The white blocks and business user. Briefly put, a data analyst collects,
indicate the steps that are taken in order to conduct a processes, and uses the data related to process executions.
process improvement project. As mentioned before, these The process analyst is an (IT) professional specialized in
steps are based on the process mining methodology of PM2 analyzing business processes and workflows with the
(van Eck et al. 2015) (which fulfills the requirement R3 as objective of finding opportunities for improvement. The
presented in Sect. 4.2). The goal of PMSS is to guide Six business user is an abstract role representing those that are

123 Fig. 3 Process mining for Six Sigma (PMSS)

T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 287

Table 4 The resource responsible, the input, the output, and the activities of each step
Responsible Input Activities Output

Planning Business Information about business Define business goals Business goals
user processes Identify and select business processes Business questions
Potential problems and supporting systems Selected processes and
First insights Identify business questions existing inf. Systems
New business problems Compose a project team Preliminary business case
Create preliminary business case
Data preparation Data Business goals Preliminary data extraction Event logs
(preliminary) analyst Business questions Preliminary data processing Audit trail
Selected processes and Preliminary data verification Data description
existing inf. Systems
Potential problems
Mining and analysis Process Event logs Exploratory process discovery First insights
(exploratory) analyst Audit trail Exploratory conformance checking Additional data (e.g., extra
Data description Exploratory process analysis data attributes)
Business problems
Data preparation Data First insights Data extraction Event logs
analyst Business problems Data processing Audit trail
Additional data (e.g., extra Data verification Data description
data attributes)
Mining and analysis Process Event logs Process discovery Improvement opportunities
(explanatory) analyst Business problems Conformance checking
Audit trail Process analysis
Data description
Process improvement Business Improvement opportunities Assess the impact of improvement Process changes towards
user opportunities/alternatives business goals
Implement improvements/process Process performance
changes indicators
Monitoring Business Process changes towards Diagnose Impact of process changes
user business goals Identify new business problems New business problems
Process performance
indicators
Evaluation Business Impact of process changes Verify and validate Verified and validated
user Process performance Supporting operations improvement
indicators

knowledgeable of organizational processes, such as busi- at: https://goo.gl/nBm3e5). Section 5.3 demonstrates a
ness unit managers, process owners, change managers. number of steps alongside an example business case.
Each PMSS step is assigned to one of these roles as being In the paragraphs below, we elaborate more on the
responsible (leading), although the work at each step PMSS steps that have been defined and enhanced based on
requires a team with members representing all three roles the feedback gathered from field experts along the DMAIC
which are closely collaborating and making use of the phases. (Where necessary, we explicitly refer to the feed-
expertise and viewpoint brought along by each one of back gathered from the experts we interviewed to justify
them. specific design decisions made for PMSS).
Table 4 provides more details regarding each PMSS
step with a brief list of corresponding inputs, outputs,
activities, and the responsible role (fulfilling R2). (Further
details regarding these activities of each step is available

123
288 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

5.1.1 Define Phase regarding KPIs that are focused on the identified business
problem and scope to enable more in-depth explanatory
The Define phase incorporates three steps. In the Planning, mining & analysis for performance measurement.
the business goals and questions are identified, the pro- The (preliminary) data preparation is comprised of three
cesses to be analyzed and improved are selected, and a sub-steps: (preliminary) data extraction, where the process
project team is established. In order to help identify execution data is extracted from selected information sys-
appropriate business goals and accompanying business tems; processing, where the data is prepared so that it can
questions (that are related to one or more aspects of busi- be stored in an event log and can be loaded into a process
ness processes, i.e., quality, time, resource, costs), the mining tool; and verification, where the data is reviewed
preliminary data preparation is performed, so that a brief for correctness and validity to ensure that it correctly
overview of the process can be gathered on the event data depicts the actual process. In the data processing sub-step,
in the exploratory mining & analysis step. However, before the good practices include the definition of an audit trail of
considering the event data, it is important to have clear the changes made to the data and the description of the data
business goals and questions that you expect to be types in the log [PMM1]. At this step, the leading role is
addressed in the following steps. If there is no clear ori- the data analyst [SSP1 and SSP2].
entation for the improvement project, it is likely to fail (as The exploratory mining & analysis step is a special form
maintained by PMC1 and PMM2; please refer to Table 3 of mining & analysis that is carried out with exploratory
for abbreviations). motives to support the identification of process-related
The input for the planning step comprises a broad business problems. It takes the data originating from the
spectrum of information regarding the organization, such previous step as input and incorporates three families of
as its business processes and major issues faced regarding techniques (that we discussed in Sect. 2): process discov-
these processes [PMD1], a selection of processes that are ery, conformance checking and process analysis (covering
known to contain issues and require attention for also enhancement techniques). In addition to process
improvement [PMC3], and first insights gained from mining techniques, the process analysis incorporates sta-
exploratory mining and analysis. As a result of a set of tistical methods and techniques (such as Pareto analysis,
activities in this step, the business goals and questions are histograms, descriptive statistics (Pyzdek 2003)). Such
defined, processes to be improved and supporting systems techniques are often used in Six Sigma initiatives to sup-
are identified, the project team with members that bring port traditional process analysis. The importance of tradi-
different perspectives to the process execution is com- tional process analysis alongside process mining is stressed
posed, and a preliminary description of the business case is by multiple studies in the process mining field in which
defined. The leading role in this step is the business user process analysis makes up for a large share of the actual
acting as the project owner and contributing with the reasoning (Berger 2017; Ryu et al. 2017; Smith and Day
domain knowledge and relevant context information (sat- 2017). Note that the mining & analysis does not enforce
isfying R4) to ensure that the initiative starts in the right any predefined order for the use of process discovery,
direction [PMC1, PMS1]. Although these activities in the conformance checking, and process analysis techniques.
planning step are depicted in an ideal sequence, they are The feedback arrow from the exploratory mining &
iterative as their actual execution unfolds. analysis step to planning indicates that the insights gained
The objective in the preliminary data preparation step is from the exploratory analysis can serve as input for
to provide data for the coming exploratory data analysis, establishing the business goals. The process analyst ensures
and in turn to facilitate the planning step in better identi- that the activities in this step are properly carried out (van
fying business problems and building a business case for Eck et al. 2015).
the initiative. As depicted in Fig. 3, this step is a special The three steps in the Define phase, planning, prelimi-
form of data preparation. At the Define phase, process nary data preparation, and exploratory mining & analysis
mining can provide insights into the current process-related can be conducted iteratively until the business goals for the
problems and complement the traditional techniques improvement project are clearly defined. The define phase
applied in this phase. Therefore, there is a need for a set of also clarifies which additional execution-related informa-
quickly-performed data preparation activities to provide tion should be extracted from the information systems.
sufficient input for the subsequent exploratory mining and
analysis. Hence, we differentiate this set of preliminary 5.1.2 Measure Phase
data preparation activities, which would quickly offer data
that can allow process analysis to be conducted with a Based on the specification of the required additional
wider lens, and the data preparation activities performed in information to be extracted from the information systems,
the Measure phase, which aim to provide enriched data in the data preparation step of the Measure phase, process

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 289

data and metrics relevant to the business problem are indicators to help evaluate if the implemented changes
retrieved, a baseline is established, and current process have yielded the expected results. In this phase, we can
performance is determined. This is achieved through the distinguish two steps: monitoring, where the process exe-
sub-steps of data extraction, processing and verification. cution is monitored with respect to the performance indi-
The objective is to extract additional process execution- cators, and evaluation, where the organization determines
related data from the information systems of the organi- if the impact of the changes depicts the expected results
zation, to process it to obtain a clear, filtered, and enriched and whether the process has been successfully improved.
event log, and to verify that the data is correct and valid As a result of monitoring, new findings may also emerge
and can be used as input to the next step, i.e., the ex- leading to new business problems and in turn a re-initiation
planatory mining & analysis. The leading role for this step of the DMAIC cycle. As we mentioned above, although
is the data analyst [SSP1 and SSP2]. Fig. 3 depicts a single feedback loop from the Control to
the Define phase, PMSS assumes iterations that allow
5.1.3 Analyze Phase traversing between any prior phase.
In the monitoring step, the business user (e.g., the
In the Analyze phase, a closer look is taken at the data change manager) is the leading role to identify and prior-
through explanatory mining & analysis. In this phase, the itize the process performance indicators (and metrics) that
team guided by the process analyst performs detailed are of importance to the organization and thus should be
analyses of the data with the aim to detect potential causes monitored [PMS1]. In addition, since it is important to
for the problems identified in the previous phase (Define), have domain knowledge at this point and since the evalu-
and identify improvement opportunities that can be acted ation happens on the business side [PMM2], the business
upon in the next phase (Improve). user (e.g., improvement project leader) is responsible for
The explanatory mining and analysis step comprises evaluating the values of the indicators that are monitored in
three sub-activities: process discovery, conformance order to assess if the process improvements have yielded
checking, and process analysis (as indicated in the legend the expected result [PMM2].
by the block titled mining and analysis in Fig. 3). Although
the techniques used in this phase are the same as those used 5.2 The Tool Support
in the exploratory mining & analysis step, the analyses are
driven by identified business questions aiming to address As indicated in R7 (Sect. 4.2), a complementary tool
the improvement goals. In turn, the level of analysis at this support that allows Six Sigma practitioners to use available
step is deeper and intense. The main input constitutes the process mining techniques in a single platform is important
event logs created in the data preparation step, the audit for the usefulness of PMSS. As described in Sect. 4.5, we
trail, the data description, and business problems. developed the tool support for PMSS as an extension to the
ProcessGold platform. One of the unique features of the
5.1.4 Improve Phase ProcessGold platform is its ability to add tags to cases in
order to group them according to certain relevant proper-
The opportunities identified in the previous step are ties. For example, in the ProcessGold platform the cases
addressed in the Improvement phase. In this phase, process with certain issues can be tagged so that they can later be
mining can be useful in detecting the likely impact of located using filters for the tag. This has made it a suit-
alternative improvement actions and in selecting those that able candidate platform for our tool support.
are likely to bring the highest impact when implemented. For each of the phases of the DMAIC model, one or
As the actual implementation of the process improve- more dashboards have been created based on the draft
ment action takes place on the business side, the business interface sketches developed during the brainstorm ses-
user takes the leading role in this step, and implements and sions with potential users and experts (Sect. 4.5). These
manages the required changes in the processes [PMC1, sketches were refined during the focus group sessions and
PMC2, PMM1, PMM2] taking the improvement opportu- usability tests performed with domain experts (Sects. 4.5
nities found in the previous step as input. The deliverable and 4.6) and during the demonstration and evaluation
of this step is subsequently improvements or so-called sessions that we discuss in the next section. Each dash-
process changes towards business goals. board contains visualizations that support the initial anal-
ysis relevant for each phase. When further analysis is
5.1.5 Control Phase needed, the visualizations link to that part of the applica-
tion that supports more elaborate analyses.
In the Control phase, the process mining techniques can be As an example, a screenshot of the dashboard created
used to monitor the predefined process performance for the Define phase is presented in Fig. 4. Each dashboard

123
290 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

Fig. 4 Screenshot of the dashboard for the define phase

element (numbered 1–3 in the Fig. 4) addresses a certain throughput time, waiting time, or total cost. Hence, the
need to fulfill a particular user objective that we gathered second dashboard element in Fig. 4 (#2) presents the
through the brainstorm sessions. For instance, as main- average completion times of different case types in order to
tained by all experts except one, the most prominent check if cases with certain properties in common have
mining technique that can be used for exploratory mining different average processing times. In relation to that, the
& analysis at this phase is process discovery, which reflects third element (#3) shows the extent to which certain
how the actual process execution took place and what activities or resources contribute to that average. From
bottlenecks can potentially be identified in the process. these dashboard elements, it is possible to navigate to that
Therefore, to check for deviations in the process, the first part of the application that supports more elaborate anal-
dashboard element in Fig. 4 (#1) is placed showing a dis- yses of the information shown in the charts.
covered process graph tuned with respect to the number of A detailed explanation of the tool support and reasoning
execution paths. regarding specific design decisions is available at: https://
As also supported by 4 experts (PMD1, PMD2, PMC1, goo.gl/LPZJpC.
PMC2), particularly when the goal of the project is to
improve the performance of a business process, process
analysis (or enhancement) techniques can be used to ana-
lyze the performance through indicators, such as

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 291

5.3 Demonstration by Means of a Business Case deemed to suffer from long processing times. It sets up a
team involving a number business users, internal data and
In this section, we provide a brief demonstration of the process analysts, and a team leader appointed as the pro-
PMSS and tool support by going through a real-life busi- ject/process owner.
ness case that involves an organization (ABC Inc.) adopt- The planning is facilitated by the preliminary data
ing Six Sigma quality management framework and preparation and exploratory mining & analysis activities in
following the DMAIC model. order to provide a better understanding of the business
In the Define phase, ABC performs the planning, pre- problem in the invoicing process. Relevant data about the
liminary data preparation, and exploratory mining & invoicing is extracted from ABC’s current enterprise
analysis activities. For the planning, ABC considers a information system, prepared and loaded into the PMSS
number of complaints that it has received from its suppliers tool support for the subsequent exploratory mining and
about invoicing, and conducts short discussions with a analysis. At this stage, the data requirements are kept at a
number of employees that are involved in the related pro- minimum to expedite the mining and analysis. (For the
cess with the main objective of defining the scope for the sake of brevity, we do not show the screens of the appli-
initiative and the goals that it aims to attain after cation components that show the loading of the data).
improvement actions are implemented. As result, ABC Figure 4 (given in the section above) shows the screen
considers focusing on its invoicing process, which is for the Define dashboard of this business case. The first

Fig. 5 Process instances with repeated activity selected

123
292 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

screen element shows the process graph for the invoicing Performing similar analyses on other activity repetitions
process. At the bottom of the second element, we can see (and possibly a more elaborate analysis of the information
that the average processing time is 4 days when all 1855 shown in the charts) confirms the initial considerations
process execution instances (cases) are considered. The regarding the potential cause of long processing times.
avg. processing time is highest for medium invoice cases Having gained an initial understanding of how the process
(9 days) and lowest for employee declaration cases (2.3 h). has been executed, its performance and potential problems,
Checking the process graph, the team notices consider- the team closes the Define phase and moves onwards to the
able repetitions of certain activities as a sign of excessive next DMAIC phase.
rework (i.e., loops and self-loops in the process graph) At the Measure phase, the data analyst takes the lead to
which potentially lead to longer processing times. For extract additional process execution-related data from
further analysis, the team selects one of the arrows that ABC’s enterprise system for enriching the event log (e.g.,
indicates a repetition of a certain activity (i.e., the activity with data from an extended period of time and other type of
‘final check of invoice’). Figure 5 shows the updated cases), and verify that the data is correct and valid. Fig-
dashboard screen when the process instances (cases) with ure 6 shows the dashboard for the Measure phase. The data
repetition of this activity are selected by the user. As can be analyst uses the dashboard to check if the data loaded into
seen in the figure, the average processing time for the cases the application is correct (for instance, by comparing the
where this activity is repeated is twice as long (8 days) as numbers of cases, activities, and users with those in ABC’s
the average time when all cases stored in the tool is enterprise system from which the data has been extracted).
considered. Furthermore, the open and closed cases are rechecked to
ensure that the cases started and closed as expected.

Fig. 6 Measure dashboard

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 293

Fig. 7 Analyze dashboard – overview tab

At the top right of Fig. 6, the number of activity repe- analyzes a number of graphs each differentiated for prob-
titions per case is shown. The fact that, on average, each lem cases and other cases. They analyze the average pro-
activity in a case is repeated 1.14 times, provides a further cessing times for different cases (#2), the resources
suggestion regarding the cause of the problem. involved in the process to check if a particular resource is a
When the team has verified the correctness of the data, major contributor to the problem (#3), and the average
they move to the Analyze phase. Figure 7 shows the number of times an activity occurs in the cases (#4). The
dashboard for this phase, which consists of two views: the graph in dashboard element #5 shows the progress of
overview and statistics. In the overview (as shown in ‘defects per million cases’ (DPMO) – a metric that is of
Fig. 7), the cases that are problematic are compared to the particular importance for the Six Sigma initiatives. The
other cases. In the case of ABC’s invoicing process, graph in element #6 shows the workload, i.e., the per-
problem cases refer to those process executions that feature centage of cases that are considered problematic with
activity repetitions, and, in turn, rework. Hence, the over- regard to the repetitions.
view dashboard shown in Fig. 7 is tailored to reflect the Six The team sees that the average processing time (#2) is
Sigma waste problem at hand, i.e., rework (as opposed to, higher for the problem cases than for the other cases for
for instance, inefficiency). majority of the case types. The same situation holds for the
The overview dashboard features a process graph (ele- number of activity repetitions per case (#3). The DPMO
ment #1 in Fig. 7) to determine if the paths of the problem metric, which is expected to be 3.4 for Six Sigma, is
cases with rework (in purple) deviate significantly from the considerably higher – around 175 k.
paths of other cases (in orange). In addition, the team

123
294 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

Furtheron in the analyze phase, the team performs 6.1 Demonstration Sessions and Interviews
additional investigations and analyses supported by a for Validation
number of dashboard elements and related graphs with the
aim to locate the causes that contribute most to the prob- The demonstration sessions (step 12 in Fig. 2) and follow-
lem, and hence to decide on the improvement alternatives up interviews were conducted with 11 experts. Table 5
to act upon in the improve phase. shows the profile of the experts with respect to their roles
In order to keep this section concise, we present the and level of expertise in related fields. Five experts were
further demonstration steps on the following page: http:// certified (black-belt or master black-belt) on Six Sigma,
tiny.cc/pqesdz. and (at least) knowledgeable about process mining. The
remaining six were process mining experts, four of them
also with considerable expertise in Six Sigma. Note that
6 Evaluation of PMSS seven experts in this list were also participants of the first
interview sessions (which we describe in Sect. 4.4 and list
In Sect. 4, we described how the initial and intermediate in Table 3). Similar to the case for the initial set of inter-
versions of PMSS were evaluated ex-ante during its views, we aimed at choosing experts that have different
development through expert interviews (Sect. 4.4), brain- roles in Six Sigma initiatives and/or process mining pro-
storm sessions (Sect. 4.5), a focus group session jects in order to enhance the internal validity of the infer-
(Sect. 4.6), and usability tests (Sect. 4.7). The ex-post ences we obtain from the follow-up interviews (Gibbert
evaluation of the final version of PMSS guideline and tool and Wicki 2008).
support focused on its validity and utility. In the first part of the demonstration sessions, we pre-
In order to collect evidence for the validity of the artifact sented and discussed the final version of PMSS and tool
(i.e., the extent to which it is applicable and can be used for support. After a brief presentation and discussion of the
its intended purpose (Gregor and Hevner 2013)), we held PMSS visual guideline (Fig. 3) and how it should be
demonstration sessions with domain experts and inter- interpreted, the details regarding each step and phase (i.e.,
viewed them to obtain their opinion on the instantiated Table 4 and additional details that are provided as external
artifact. Next, we asked the experts of the demonstration sources for this paper at https://goo.gl/nBm3e5) were dis-
sessions to respond to a questionnaire to evaluate PMSS’ cussed. Next, the tool support was demonstrated by going
utility, i.e., how useful and easy-to-use they considered the through a realistic case scenario, and participants were
guideline and tool support to be for Six Sigma initiatives. interviewed for their opinion on the method and tool sup-
In the sub-sections below, we elaborate on the demon- port regarding its validity and completeness. During the
stration sessions and the conduct of the survey, and discuss demonstration sessions and follow-up interviews, experts
our findings. also suggested minor changes for the guideline and tool-
support, which were incorporated in the final version.

Table 5 The experts in the demonstration sessions and their level of expertise in related fields
Interviewees with the participants of the demo Level of expertise Years of experience in the (main) field of expertise
Process mining Six Sigma

SSP1 – Six Sigma Practitioner 1 (master black-belt) ? ?? 12

SSP2 – Six Sigma Practitioner 3 (black-belt) ? ?? 2
SSP3 – Six Sigma Practitioner 4 (black-belt) ?? ?? 17
SSP4 – Six Sigma Practitioner 5 (black-belt) ?? ?? 3
SSP5 – Six Sigma Practitioner 6 (black-belt) ? ?? 5
PMC2 – Project Consultant 2 ?? ? 2
PMC3 – Project Consultant 3 ?? ?? 6
PMM2 – Project (C-level) Manager 2 ?? ? 10
PMM2 – Project (C-level) Manager 3 ??  10
PMS1 – System Sales Representative 1 ??  5
PMS2 – System sales representative 2 ?? ?? 4
‘??’: Expert, ‘?’: Knowledgeable, ‘’: Not familiar

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 295

The experts displayed an agreement that a standard that the way the process mining techniques can be used in
operating procedure as presented in the guideline is useful combination with the traditional Six Sigma tools should be
to help Six Sigma practitioners to perform their activities in made more explicit. Many Six Sigma experts commented
an improvement project more effectively and efficiently on this point:
with the help of process mining. Below, we include a
From a master black belt perspective, I still need to
number of quotes from interviews.
integrate your guideline in the traditional (Lean) Six
Six Sigma practitioners analyze in a very structured Sigma guideline because I have to train and coach my
way and also want to know how process mining can black belts in the entire process. For me, they cannot
contribute to that. So, when we are able to present be separate. [SSP1]
that, in a structured and predefined way, then I think
In addition, the experts suggested incorporating addi-
that those people will benefit from that. They will
tional (statistical) traditional Six Sigma techniques into the
also be convinced of process mining in general.
tool support, such as the distribution plots and control
Therefore, I think it’s quite useful to have it. [PMC3]
charts, which would help making it more useful and more
I think it’s definitely useful. It helps people to better
likely to be embraced by Six Sigma practitioners. Sup-
understand what they have to do everywhere and it
porting different perspectives with respect to common
guides them through it. I think that guidance is very
process problems was also brought up as another point of
very important. [SSP5]
improvement regarding the tool support. Instead of having
Another point of agreement among the majority of generic dashboards applicable to any kind of process
experts was the guideline’s ease of use. They indicated that problems, dashboards that are tailored to a specific process
the guideline might be hard to grasp at first sight – par- problem (e.g., rework) could be generated:
ticularly for those new to these fields -, but that the level of
The guideline is self-explanatory and the tool support
granularity and complexity engrained in the guideline is
is very easy to use. With the addition that it would be
needed to ensure usefulness and still remain relatively
even easier to use if you added some templates on
easy-to-use.
traditional lean analysis: waiting time, lead time,
Everything can always become much easier. But, rework, etc. [SSP1]
easiness will also lead to the loss of some function-
ality as well. I’m not sure if I’m, at this point, willing 6.2 Survey
to give that up, [PMM2]
I think the guideline is easy to use, it’s a bit over- The experts who joined the demonstration sessions and
whelming at the beginning maybe. But, you definitely were interviewed, were asked to participate in a short
need all the things that are in there. There is nothing survey to express their opinion on the utility of PMSS (step
that I would leave out when I look at it. [PMS1] 13 in Fig. 3). From the 11 experts that were interviewed, 9
participated, which resulted in a participation rate of 82%.
Furthermore, the experts indicated that they would rec-
The questionnaire was composed by taking as a basis the
ommend PMSS to Six Sigma practitioners. Not only do
Technology Acceptance Model – TAM (Davis 1989;
they think that this structured way of conducting process
Venkatesh and Davis 2000). The TAM and its derivatives,
mining is useful for Six Sigma practitioners, but that it can
e.g., (Venkatesh et al. 2003), have been used as a theo-
be useful for organizations that have not adopted Six Sigma
retical basis for several empirical studies in the information
principles.
systems field, including the acceptance of IS methods and
(This) is a structured way of doing Six Sigma using models (e.g., Turner et al. 2010; Stojanov et al. 2015;
process mining. These are two fields that can really Schriek et al. 2016; Turetken and Grefen 2017; Dikici et al.
benefit from each other. [PMS1] 2018; Turetken et al. 2019). The original TAM has three
I’d definitely recommend the guideline to Six Sigma primary constructs: perceived ease of use, perceived use-
practitioners. Also to non-Six Sigma guys. …. guys fulness, and intention to use (Davis 1989). Perceived use-
who are not that familiar with Six Sigma can make fulness refers to users’ perception of the utility of the
use of it. That’s quite nice I think. [PMC3] design artefact in providing gains to its user (Venkatesh
et al. 2003). Perceived ease of use refers to ‘‘the degree to
The experts also indicated a number of points for
which a person believes that using a particular design
improvement that we accepted as tasks for future work.
artefact will be free from physical or mental effort’’.
The main point of criticism of the guideline was the lack of
Finally, intention to use can be defined as the extent to
clarity regarding the relation between process mining
which a person intends to use a particular design artefact.
techniques and the traditional Six Sigma tools. It is argued

123
296 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

Intention to use is the most proximal antecedent to the (Q1–Q4), perceived ease of use (Q5–Q8), and intention to
artefact use and believed to be determined by perceived use (Q9–Q10).
usefulness and ease of use (Davis 1989; Venkatesh and The positive tendency regarding the ‘perceived useful-
Davis 2000). ness’ statements indicate that the experts considered the
The constructs of TAM are operationalized using mul- guideline to be useful. Particularly the indicators (items)
tiple indicators, and their reliability and validity have been regarding the perceived usefulness link closely to the
rigorously evaluated (Davis 1989). In line with (Venkatesh objective of our research study, and the majority of experts
and Davis 2000), we used 4 indicators for perceived use- indicated a positive view on these indicators and agreed
fulness and ease of use, and 2 for intention to use. To that PMSS can support organizations that adopt Six Sigma
accommodate this research, the wording of the indicators principles to perform their activities more effectively and
(items) were modified in accordance with the approach efficiently with the help of process mining techniques.
followed in (Moody 2003). In addition, 3 statements were The positive outlook also held for the ‘perceived ease of
negated in order to prevent monotonous responses. The use’. The majority of the experts believed that the guideline
participants were asked to express their level of agreement was easy to use and that they would be able to become
with each statement on a 5-point Likert scale, ranging from skillful at doing so. The responses to the ‘intention to use’
1 (strongly disagree) to 5 (strongly agree). indicators also showed a positive outcome. However, only
The questions and the experts’ responses are presented few experts seemed to ‘strongly agree’ with the statements.
in Fig. 8. The results indicate a positive view of the experts Furthermore, the experts mostly had a neutral opinion on
towards the measured constructs, i.e., perceived usefulness preferring the guideline over another approach of using
process mining techniques to support Six Sigma related

Fig. 8 Results of the survey about perceived usefulness, perceived ease of use, and intention to use of PMSS

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 297

activities. Only a single expert among the expert group the help of process mining techniques. The tool support
exhibited a relatively negative attitude to the guideline and that has been developed using a commercially available
indicated this in the responses. However, overall, the enterprise solution and its qualitative evaluation with
responses of the experts indicate a generally positive atti- domain experts demonstrates its potential to be deployed as
tude towards using PMSS. a support for Six Sigma practitioners. Recent initiatives in
practice prove the importance of using process mining to
provide Six Sigma practitioner with new perspectives and
7 Conclusions tools to find root causes quickly (Fluxicon 2019; Gartner
2019).
Six Sigma focuses on improving business processes by
statistically quantifying process performance changes, but 7.1 Limitations and Future Work
data used for such analyses is typically collected manually,
which makes Six Sigma a costly and time-consuming Our work has a number of limitations which also drive our
endeavor (Pyzdek 2003; van der Aalst 2016). In the past future work. PMSS has been ex-post validated through
few years, process mining has proven to be a useful tech- interviews with domain experts, to whom the artifact
nique to conduct process analysis in improvement projects (PMSS and tool support) was explained and demonstrated
in a potentially time-efficient way (Garcia et al. 2019). with an example business case. After the demonstrations,
Hence, process mining can serve as an important support the experts were invited to indicate their opinion on the
technology for process improvement frameworks, such as validity and utility of the artifact. Although this method of
Six Sigma (van der Aalst and Dustdar 2012; van der Aalst evaluation is valid and reliable, it would be considerably
et al. 2012b; van der Aalst 2016; Harmon 2018). more conclusive if PMSS and the tool were used by the
Despite the agreed potential, there is no research on how experts in real-life settings to perform Six Sigma activities.
process mining can be systematically incorporated into the This would provide a stronger confirmation of the validity
DMAIC model of Six Sigma (van Geffen and Niks 2013; and utility of the artifact. Our future work will incorporate
Sebu and Ciocarlie 2014; van der Aalst et al. 2016; Valle these evaluation techniques that involve applications by
et al. 2017). No structured guideline exists that specifies practitioners in real-life settings to improve the artifact and
how and when process mining can support Six Sigma strengthen the conclusions regarding its validity and use-
related activities and lead to a potentially more efficient fulness. Similarly, it will include performing controlled
and effective performance of these improvement activities. experiments in real-life business settings with groups of
This research aimed to fill this research gap by developing practitioners to test the tangible influence and value of
PMSS and its tool support. using the PMSS over not using it at all.
This work has implications for both research and prac- The Six Sigma experts that we interviewed indicated
tice. It not only extends the body of knowledge in the fields that the relationship between the process mining tools and
of process mining and Six Sigma, but also helps to close techniques and the traditional techniques used in Six Sigma
the gap between them, thereby contributing to the broad initiatives (e.g., distribution plots, control charts) can be
field of quality management. It is an initial and explicit step further clarified and potentially integrated. There is a need
to bridge these fields, which will contribute to future to explicitly describe how the tools and techniques of
research regarding the design and development of process process mining can complement those that are commonly
mining tools and techniques that are tuned to the needs of employed in DMAIC phases, and to incorporate this
Six Sigma practitioners. This work contributes to the pro- structure into the standard operating procedure of the
cess mining research with an attempt to structure its DMAIC model. This research work has taken process
application, and therefore paves the way for exploiting the mining as a point of departure and incorporated it into the
capabilities and potential of process mining algorithms, DMAIC phases. Future work should also proceed from the
tools and methods that have been widely researched over Six Sigma’s perspective, and aim at exploring how and
the last decade (Thiede et al. 2018). when traditional tools and techniques of Six Sigma can be
This works contributes also to the practice of DMAIC- complemented and supported by process mining tools and
based process improvement. PMSS offers Six Sigma techniques. Similarly, some of these Six Sigma techniques
practitioners a guideline that extends their existing standard can be combined in such a way that together they make
operating procedure and will potentially increase the effi- process mining insights more comprehensive for Six Sigma
ciency and effectiveness of process improving efforts in practitioners. In addition, PMSS is currently intended as
their organizations. PMSS can also serve as a generic complementary to the DMAIC toolkit, while future work
process mining methodology for organizations willing to can also consider incorporating traditional tools and tech-
follow a structured approach for improving processes with niques within PMSS to support all DMAIC activities.

123
298 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

The tool support that we developed is prototypical and Ailenei I, Rozinat A, Eckert A, van der Aalst WMP (2012) Definition
suffers from the limitations that are described above (e.g., and validation of process mining use cases. In: Daniel F,
Barkaoui K, Dustdar S (eds) Business process management
limited support for traditional Six Sigma techniques). workshops. BPM 2011, vol 99. Lecture notes in business
Hence, it requires further work to streamline and improve information processing. Springer, Heidelberg. https://doi.org/10.
the functionalities offered. Several improvement points 1007/978-3-642-28108-2_7
were raised in expert interviews that provide valuable input Antony J, Rodgers B, Cudney EA (2017) Lean six sigma for public
sector organizations, is it a myth or reality? Int J Qual Reliab
for future work. For instance, enhancing the dashboards Manag 34:1402–1411. https://doi.org/10.1108/MRR-09-2015-
with respect to certain types of business problems (e.g., 0216
increasing throughput, increasing resource utilization) was Antony J, Palsuk P, Gupta S et al (2018) Six sigma in healthcare: a
considered to be a useful feature for the tool support. The systematic review of the literature. Int J Qual Reliab Manag
39(12):1663–1694. https://doi.org/10.1108/MRR-09-2015-0216
current dashboards purposely have generic designs and are Augusto A, Conforti R, Dumas M et al (2019) Automated discovery
not designed to support custom elements for a particular of process models from event logs: review and benchmark. IEEE
business problem type. Allowing users to select a particular Trans Knowl Data Eng 31(4):686–705. https://doi.org/10.1109/
problem to focus on for improvement and populating the TKDE.2018.2841877
Barnum CM (2011) Usability testing essentials: ready set…test!
dashboards with elements that are relevant to that particular Morgan Kaufmann, Burlington
problem is considered as a useful feature for the new Bedgood C (2017) The art of lean six sigma in healthcare. Ind Manag
version of the tool. 59(6):22–25
Finally, we used a commercially available enterprise Berger F (2017) Mining event log data to improve a loan application
process. In: International business process intelligence challenge
process mining solution to host the tool support and made (BPIC’17), pp 1–29
used of the functionality available for process mining Blumberg B, Cooper DR, Schindler PS (2008) Business research
within a single platform. However, the developed guideline methods, 2nd edn. McGraw-Hill Education, Berkshire
and related concepts are general and can act as a basis for Boersma HJ, Leung TI, Vanwersch R et al (2019) Optimizing care
processes with operational excellence & process mining. Fun-
the implementation of a tool support in other platforms. damentals of clinical data science. Springer International Pub-
Hence, for future work, the constructs on which the dash- lishing, Cham, pp 181–192
boards are created can be further generalized so that it Bozkaya M, Gabriels J, Werf JM Van Der (2009) Process diagnostics:
becomes possible to use publicly available process mining a method based on process mining. In: International Conference
on Information, Process, and Knowledge Management. IEEE,
platforms, such as ProM (van der Aalst 2016). pp 22–27
Bücker I, Hermann M, Pentek T, Otto B (2016) Towards a
Acknowledgements The authors especially thank all experts for their methodology for industrie 4.0 transformation. In: Abramowicz
participation in this research and for the valuable feedback they W, Alt R, Franczyk B (eds) Business information systems (BIS
provided during interviews and focus group sessions. The authors also 2016), vol 255. Lecture notes in business information process-
thank ‘‘ProcessGold’’ for its support of this research and, in particular, ing. Springer, Cham, pp 209–221
of the development of the tool support for PMSS. Carmona J, van Dongen B, Solti A, Weidlich M (2018) Conformance
checking: relating processes and models. Springer International
Open Access This article is licensed under a Creative Commons Publishing, Cham
Attribution 4.0 International License, which permits use, sharing, Dahlin S, Eriksson H, Raharjo H (2019) Process mining for quality
adaptation, distribution and reproduction in any medium or format, as improvement: propositions for practice and research. Qual
long as you give appropriate credit to the original author(s) and the Manag Health Care 28(1):8–14. https://doi.org/10.1097/QMH.
source, provide a link to the Creative Commons licence, and indicate 0000000000000197
if changes were made. The images or other third party material in this Davis F (1989) Perceived usefulness, perceived ease of use, and user
article are included in the article’s Creative Commons licence, unless acceptance of technology. MIS Q 13(3):319–340
indicated otherwise in a credit line to the material. If material is not De Mast J, Lokkerbol J (2012) An analysis of the six sigma DMAIC
included in the article’s Creative Commons licence and your intended method from the perspective of problem solving. Int J Prod Econ
use is not permitted by statutory regulation or exceeds the permitted 139:604–614. https://doi.org/10.1016/j.ijpe.2012.05.035
use, you will need to obtain permission directly from the copyright De Weerdt J, De Backer M, Vanthienen J, Baesens B (2012) A multi-
holder. To view a copy of this licence, visit http://creativecommons. dimensional quality assessment of state-of-the-art process
org/licenses/by/4.0/. discovery algorithms using real-life event logs. Inf Syst
37(7):654–676. https://doi.org/10.1016/j.is.2012.02.004
Dijkman R, Turetken O, van IJzendoorn GR, de Vries M (2019)
References Business processes exceptions in relation to operational perfor-
mance. Bus Process Manag J 25(5):908–922. https://doi.org/10.
1108/BPMJ-07-2017-0184
Aguirre S, Parra C, Alvarado J (2013) Combination of process mining Dikici A, Turetken O, Demirors O (2018) Factors influencing the
and simulation techniques for business process redesign: a understandability of process models: a systematic literature
methodological approach. In: Cudre-Mauroux P, Ceravolo P, review. Inf Softw Technol 93:112–129. https://doi.org/10.1016/j.
Gašević D (eds) International symposium on data-driven process infsof.2017.09.001
discovery and analysis (SIMPDA 2012), vol 162. Lecture notes Dogan O, Gurcan OF (2018) Data perspective of lean six sigma in
in business information processing. Springer, Heidelberg, industry 4.0 era: a guide to improve quality. Proc Int Conf Ind
pp 24–43 Eng Oper Manag 2018:943–953

123
T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021) 299

dos Garcia CS, Meincheim A, Faria Junior ER et al (2019) Process conference on information systems (ECIS 2003). Association for
mining techniques and applications – a systematic mapping Information Systems
study. Expert Syst Appl 133:260–295. https://doi.org/10.1016/J. Nadeau S (2017) Lean, six sigma and lean six sigma in higher
ESWA.2019.05.003 education: a review of experiences around the world. Am J Ind
Dul J, Hak T (2008) Case study methodology in business research. Bus Manag 7(5):591–603. https://doi.org/10.4236/ajibm.2017.
Routledge, New York 75044
Evans JR, Lindsay WM (2016) Managing for quality and perfor- Nielsen J (1994) Usability engineering. AP Professional/Morgan
mance excellence, 10th edn. South-Western, Cengage Learning, Kaufmann, Burlington
Mason Pande PS, Holpp L (2002) What is six sigma?. McGraw-Hill, New
Fluxicon (2019) Combining lean six sigma and process mining – part York
I. Eindhoven Park S, Kang YS (2016) A study of process mining-based business
Flynn BB, Schroeder RG, Sakakibara S (1995) The impact of quality process innovation. Procedia Comput Sci 91:734–743. https://
management practices on performance and competitive advan- doi.org/10.1016/j.procs.2016.07.066
tage. Decis Sci 26(5):659–691. https://doi.org/10.1111/j.1540- Peffers K, Tuunanen T, Rothenberger MA, Chatterjee S (2007) A
5915.1995.tb01445.x design science research methodology for information systems
Gartner (2019) Market guide for process mining. Gartner, Stamford research. J Manag Inf Syst 24(3):45–77. https://doi.org/10.2753/
George ML (2002) Lean six sigma: combining six sigma quality with MIS0742-1222240302
lean speed. McGraw-Hill, New York Peris-Ortiz M, Alvarez-Garcia J (2014) Action-based quality man-
Gibbert M, Wicki B (2008) What passes as a rigorous case study? agement: strategy and tools for continuous improvement.
Strateg Manag J 29(13):1465–1474. https://doi.org/10.1002/smj Springer, Heidelberg
Gregor S, Hevner AR (2013) Positioning and presenting design Pyzdek T (2003) The six sigma handbooks. McGraw-Hill, New York
science research for maximum impact. MIS Q 37(2):337–355. Rebuge A, Ferreira DR (2012) Business process analysis in healthcare
https://doi.org/10.2753/MIS0742-1222240302 environments: a methodology based on process mining. Inf Syst
Gupta S, Modgil S, Gunasekaran A (2019) Big data in lean six sigma: 37(2):99–116. https://doi.org/10.1016/j.is.2011.01.003
a review and further research directions. Int J Prod Res. https:// Rojas E, Sepúlveda M, Munoz-Gama J et al (2017) Question-driven
doi.org/10.1080/00207543.2019.1598599 methodology for analyzing emergency room processes using
Harmon P (2018) Process mining grows up. BPTrends. https://www. process mining. Appl Sci. https://doi.org/10.3390/app7030302
bptrends.com/harmon-on-bpm-process-mining-grows-up/ Ryu D, Lim J, Jeong D, Song M (2017) Analysis of loan process
Hess H (2006) From corporate strategy to process performance — using various process mining techniques: the BPIC 2017. In:
what comes after business intelligence? In: Corporate perfor- Seventh international business process intelligence challenge
mance management. Springer, Heidelberg, pp 7–29. https://doi. (BPIC’17), pp 1–27
org/10.1007/3-540-30787-7_2 Samson D, Terziovski M (1999) The relationship between total
Hong C (2017) Implementing lean six sigma to achieve inventory quality management practices and operational performance.
control in supply chain management. In: AIP conference J Oper Manag 17(4):393–409. https://doi.org/10.1016/S0272-
proceedings 6963(98)00046-1
ISO (2015) ISO 9000 family – quality management. Int Organ Stand. Schriek M, Turetken O, Kaymak U (2016) A maturity model for care
https://www.iso.org/iso-9001-quality-management.html. Acces- pathways. In: European conference on information systems
sed 19 Sept 2018 (ECIS 2016). Association for Information Systems
Johannsen F, Leist S, Zellner G (2011) Six sigma as a business Sebu ML, Ciocarlie H (2014) Applied process mining in software
process management method in services: analysis of the key development. In: 9th IEEE international symposium on applied
application problems. Inf Syst E-bus Manag 9:307–332. https:// computational intelligence and informatics (SACI – 2014),
doi.org/10.1007/s10257-010-0128-2 pp 55–60. IEEE
Karakhan A (2017) Six sigma & construction safety: using the Sebu ML, Ciocarlie H (2015) Business activity monitoring solution to
DMAIC cycle to improve incident investigations. Prof Saf detect deviations in business process execution. In: 2015 IEEE
62(6):38–40 10th jubilee international symposium on applied computational
Kitchenham B, Charters S (2007) Guidelines for performing system- intelligence and informatics, pp 437–442. IEEE
atic literature reviews in software engineering. EBSE Technical Siddiqui SQ, Ullah F, Thaheem MJ, Gabriel HF (2016) Six sigma in
Report, Version 2.3. https://doi.org/10.1145/1134285.1134500 construction: a review of critical success factors. Int J Lean Six
Leemans SJJ, Fahland D, Van Der Aalst WMP (2013) Discovering Sigma 7(2):171–186. https://doi.org/10.1108/IJLSS-11-2015-
block-structured process models from event logs – a constructive 0045
approach. Lect Notes Comput Sci 7927:311–329. https://doi.org/ Smith S, Day C (2017) Insights into the consumer loan application
10.1007/978-3-642-38697-8_17 process of a dutch bank. In: International business process
LeMahieu PG, Nordstrum LE, Cudney EA (2017) Six sigma in intelligence challenge (BPIC’17), pp 1–36
education. Qual Assur Educ 25(1):91–108. https://doi.org/10. Stojanov I, Turetken O, Trienekens JJM (2015) A maturity model for
1108/QAE-11-2012-0046 scaling agile development. In: Proceedings - 41st Euromicro
Lepofsky A (2015) Why your organisation should buy a collaboration Conference on Software Engineering and Advanced Applica-
platform instead of best-of-breed solutions. Constellation tions, SEAA 2015
Research Inc., Monte Vista Suriadi S, Wynn MT, Ouyang C, et al (2013) Understanding process
Lu J, Laux C, Antony J (2017) Lean six sigma leadership in higher behaviours in a large insurance company in Australia: a case
education institutions. Int J Product Perform Manag study. In: Advanced information systems engineering (CAISE
66(5):638–650. https://doi.org/10.1108/MRR-09-2015-0216 2013), pp 449–464
Madhani PM (2016) Application of six sigma in supply chain Thiede M, Fuerstenau D, Barquet APB (2018) How is process mining
management: evaluation and measurement approach. IUP J technology used by organizations? A systematic literature review
Supply Chain Manag 13(3):33–53 of empirical studies. Bus Process Manag J 24(4):900–922.
Moody DL (2003) The method evaluation model: a theoretical model https://doi.org/10.1108/BPMJ-06-2017-0148
for validating information systems design methods. In: European

123
300 T. Graafmans et al.: Process Mining for Six Sigma, Bus Inf Syst Eng 63(3):277–300 (2021)

Tjahjono B, Ball P, Vitanov VI et al (2010) Six sigma: a literature performance analysis. Wiley Interdiscip Rev Data Min Knowl
review. Int J Lean Six Sigma 1(3):216–233. https://doi.org/10. Discov 2:182–192. https://doi.org/10.1002/widm.1045
1108/20401461011075017 van der Aalst WMP, Adriansyah A, De Medeiros AKA, et al (2012b)
Tomašević I, Slović D (2013) Facilitating wasteful activities discov- Process mining manifesto. In: BPM 2011 workshops, LNBIP 99.
ery in pure service environment through usage of process Springer, Heidelberg, pp 169–194
mining. Int J Ind Eng Manag 4(4):199–206 van der Aalst WMP, La Rosa M, Santoro FM (2016) Business process
Turetken O, Grefen P (2017) Designing service-dominant business management: don’t forget to improve the process! Bus Inf Syst
models. In: European conference on information systems (ECIS Eng 58(1):1–6. https://doi.org/10.1007/s12599-015-0409-x
2017). Association for Information Systems, Guimarães, van Eck ML, Lu X, Leemans SJJ, van der Aalst WMP (2015) PM2: a
pp 2218–2233 process mining project methodology. In: Advanced information
Turetken O, Dikici A, Vanderfeesten I et al (2019) The influence of systems engineering (CAiSE 2015). LNCS 9097, pp 297–313
using collapsed sub-processes and groups on the understand- van Geffen F, Niks R (2013) Accelerate DMAIC using process
ability of business process models. Bus Inf Syst Eng mining. In: Proceedings of 3rd BPI challenge BPIC’13, CEUR
62(2):121–141. https://doi.org/10.1007/s12599-019-00577-4 workshop proceedings (CEUR-WS.org)
Turner M, Kitchenham B, Brereton P et al (2010) Does the van Kollenburg T, Wouters S (2018) The future of continuous
technology acceptance model predict actual use? A systematic improvement. In: European lean educators conference 2018,
literature review. Inf Softw Technol 52(5):463–479. https://doi. Braga
org/10.1016/j.infsof.2009.11.005 Venable JR, Pries-heje J (2012) Design science research in informa-
Valle AM, Santos EAP, Loures ER (2017) Applying process mining tion systems. Adv Theory Pract. https://doi.org/10.1007/978-3-
techniques in software process appraisals. Inf Softw Technol 642-29863-9
87:19–31. https://doi.org/10.1016/J.INFSOF.2017.01.004 Venable J, Pries-Heje J, Baskerville R (2016) FEDS: a framework for
van Cruchten RMER, Weigand HH (2018) Process mining in evaluation in design science research. Eur J Inf Syst
logistics: the need for rule-based data abstraction. In: 12th 25(1):77–89. https://doi.org/10.1057/ejis.2014.36
International conference on research challenges in information Venkatesh V, Davis FD (2000) A theoretical extension of the
science (RCIS 2018), pp 1–9. IEEE technology acceptance model: four longitudinal field studies.
van der Aalst WMP (2010) Process discovery: capturing the invisible. Manag Sci 46(2):186–204. https://doi.org/10.1287/mnsc.46.2.
IEEE Comput Intell Mag 5(1):1–28 186.11926
van der Aalst WMP (2016) Process mining: data science in action, Venkatesh V, Morris MG, Davis GB, Davis FD (2003) User
2nd edn. Springer, Heidelberg acceptance of information technology: toward a unified view.
van der Aalst WMP, Dustdar S (2012) Process mining put into MIS Q 27(3):425–478. https://doi.org/10.2307/30036540
context. IEEE Internet Comput 16:82–86. https://doi.org/10. Wilson C (2013) Brainstorming and beyond, 1st edn. Morgan
1109/MIC.2012.12 Kaufmann, Burlington
van der Aalst W, Adriansyah A, van Dongen B (2012a) Replaying
history on process models for conformance checking and

123