This chapter is about the philosophical assumptions of the author and the underlying research philosophy of the thesis. Thereupon, an aligned research design and research strategy is chosen that is suited to answer the research questions. The main part forms the description about how the data is collected and analyzed. Finally, measurements to ensure a great research quality as well as a high ethical standard are outlined.
Research is about acquiring new knowledge in a particular field and the nature of that knowledge (Saunders, Lewis & Thornhill, 2009). Research projects are based on different underlying philosophical assumptions that determine the conceptualization of the research project (Saunders, et al. 2009; Easterby-Smith, et al., 2015). Thus, a researcher needs to be aware of his or her own philosophical assumptions before developing an appropriate research design (Saunders et al. 2009; Easterby-Smith et al., 2015). Furthermore, it supports the scholar’s reflective role and the argumentation about how the author generates and interprets knowledge (Saunders et al., 2009).
The basic philosophical directions are ontology and epistemology (Saunders et al. 2009; Easterby-Smith et al., 2015). According to Byrne (2017) ontology comes from the Greek “ontos,” which means being, and “logos,” meaning study” (Byrne, 2017). It is concerned with the nature of reality. Researchers differentiate between the two main concepts of realism (Easterby-Smith et al., 2015) which is similar to objectivism (Saunders et al., 2009), and relativism (Easterby-Smith et al., 2015) that is also named subjectivism (Saunders et al., 2009). In regards of answering my research questions and hence the fact that theory in the related field is still in its infancy, the research relies on the knowledge contribution of the involved research participants. I as a researcher, and the research participants are socially connected human beings and therefore we are part of the reality that is being observed. Thus, I take a relativist ontology. This philosophical position accepts that there are many truths created by people (Easterby-Smith et al., 2015).
In regard of the epistemology, I support the constructionist approach that the social reality is a crucial part of how people acquire knowledge through social interaction like intuition or reflection (Easterby-Smith et al, 2015). Choosing a constructionism research design allows me gathering rich data from the research participants from which ideas are induced (Easterby-Smith et al., 2015, p. 53). A constructionism epistemology is linked to the ontology of relativism (Easterby et al., 2015, p. 54). In the case of my research the context and feelings of the author and the involved research participants or ‘social actors’ are fundamental for a creative and natural design to gather rich data and a reflective (less artificial) interpretation (Saunders et al., 2009). One strength of a relativist constructionist research is the acceptance of multiple data sources (Easterby-Smith et al., 2015). On the other side, data collection and analysis are therefore time-consuming and difficult tasks. Furthermore, constructionism is suited for the development of new theories and the identification of emergent topics (Easterby-Smith et al., 2015). In contrast to an internal realist positivistic philosophy which also facilitates theory generation, the generalization of a relativist constructionist stance is not limited to the given sample. In conclusion, I believe that that a relativist constructionist philosophical stance is well-suited to answer the defined research questions and in addition, provides new insights to the young research area of process mining. Moreover, it also reflects my personal understanding of values which affect the research process, particularly during data collection and analysis (Saunders et al., 2009).
Abductive research approach
My research philosophy is attached to an inductive research approach. A constructionist inductive approach aims towards theory generation through qualitative data collection from small samples (Glaser and Strauss, 1967; Saunders et al., 2009), in contrast to an deductive approach which builds on existing theory (Dubois & Gadde, 2002) and aims towards theory testing by using large samples of quantitative data (Saunders et al.,2009). The fact that theory about process mining is rather emergent than established which would support theory testing favors my decision to adopt a rather inductive approach (Saunders et al., 2009). In addition, deductive approaches are mainly based on a positivistic research philosophy (Saunders, et al., 2009; Easterby-Smith et al., 2015). Both approaches share that they follow a rather linear and artificial research process, which tends to ignore the discovery of new insights that emerge during the research process (Dubois and Gadde, 2002). According to Kovács and Spens (2005) neither a pure inductive nor a deductive approach have brought great research findings.
Dubois and Gadde (2002) present a third approach of abduction. During abduction, theory and reality are continuously confronted, facilitating a non-linear research process (Dubois & Gadde, 2002). By going back and forth between the literature of process mining, BPM and the empirical data reflecting the reality the abductive approach is well-suited to develop theory in process mining further and moreover enable the discovery of new things (Dubois & Gadde, 2002). An abductive approach is also beneficial for the exploratory nature of this study to find out “what is happening; to seek new insights; to ask questions and to assess the phenomena in a new light (Robson, 2002, p. 59).” (Saunders et al., 2009).
Research strategy: illustrative case study
The research strategy must be aligned with the research philosophy and research approach. Saunders et al. (2009) and Easterby-Smith et al. (2015) provide a variety of research strategies. However, not every strategy fits to an abductive research approach. According to Kovács and Spens (2005) action research and case studies use abductive reasoning hence within both strategies data collection and theory development happen in parallel. Eisenhardt & Graebner (2007) agree with Dubois and Gadde (2002) that “case studies provide unique means of developing theory by utilizing in-depth insights of empirical phenomena and their contexts” (p. 555).
My original idea to conduct a case study with a company that has implemented process mining or is planning to do so has failed due to the reasons outlined in 8.1 Limitations. Against this backdrop, I developed an alternative approach to meet the demand of a young process mining literature of real-life research studying the impacts of process mining. The result is an illustrative case study research strategy. I am convinced that the taken case study research creates an understanding that would have otherwise not been achieved.
Case studies regularly denotes the combination of multiple sources of qualitative or quantitative manner to ensure credibility and accuracy of the research findings (Eisenhardt, 1989). The intermediate position of case study research strategy by Eisenhardt and Graebner (2007) agrees with the abductive reasoning of Dubois and Gadde (2002) whereby the constructionist, inductive theory creation from cases is complemented by a positivist, deductive part to test the emerged data against existing theory. With inductive theory building from cases producing new theory from data and deductive theory testing completing the cycle by using data to test theory. According to Dubois and Gadde (2002) the evolving case is part of the matching process between theory and reality. Applied to my research, I conducted qualitative semi-structured interviews with participants in the field of process mining to collect rich information in order to answer the defined research questions. The illustrative case is a chosen case company that has implemented process mining. In order to establish a high degree of credibility of the chosen case and its context extensive secondary data in terms of company and industry reports were collected. In regard of the need of real-life demonstrations of process mining I applied process mining to an identical case context using exemplary event log data. However, the collected primary interview data builds the empirical backbone of my thesis.
Scholars discuss the limitations of case studies. Yin (1994) and Yin (2003) criticize that case studies are too situation specific and therefore allow only a limited generalization. The criticism of being too situation specific has led scholars arguing for the application of multiple cases (Yin, 1994; Eisenhardt, 1989; Miles and Huberman, 1994). Dubois and Gadde (2002) agrees with Easton (1995) that multiple case studies tend to result in more breadth and less depth, eliminating the actual strength of case study research. Consequently, many case studies describe everything and as a result describe nothing (Easton, 1995; Weick, 1979) because of a rather sloppy research design (Yin, 1994). However, Dubois and Gadde (2002) argue that the most significant choice is not between single and multiple cases it depends rather on the research objective. Furthermore, the authors state that the weakness of being too situation-specific is outdated and argue that it has become an opportunity to generalize findings (Dubois & Gadde, 2002).
Scholars agree that case studies are typically based on a variety of data sources (Eisenhardt & Graebner, 2007; Saunders et al., 2009; Easterby-Smith et al., 2015) which positively support their credibility and accuracy (Yin, 1992; Dubois & Gadde, 2002). According to Yin (2003) the use of multiple data sources often results in large amounts of data which makes authors interpretations biased. To address this problem the following research design aims to define a clear research procedure about the methods and data used for data collection and analysis. Furthermore, the “combining sources of evidence, while shifting between analysis and interpretation, usually denotes triangulation” (Dubois and Gadde, 2002, p. 556), see 3.6 Triangulation.
The inaccessibility and characteristic of event log data has hindered my objective to conduct the research project in cooperation with a case company (see 8.1 Limitations).
Process mining vendors like Celonis SE or Lana Labs GmbH provide publicly accessible process mining case studies on the corresponding websites. The free accessibility of using a case study to simulate the application of a typical process mining projects supports the replicability and the overall credibility of my research. Thus, I applied purposeful sampling to select a case that is based on the decision of the author (Saunders et al., 2009). When reviewing the different case studies, I decided to apply three criteria for my case selection.
The first criterion concerned the choice of the company. To introduce the case and describe the context of phenomena under investigation (6.2 Context of phenomena under investigation) sufficient public accessible information about the company and its industry must be available. The second criterion was the type of industry the company is active in. The conditions of the specific industry should be found in different industries to enable generalization of the research findings. Thirdly, the analyzed process in question should be a typical process that can be find in most businesses further facilitating generalization of the findings.
In conclusion, the chosen case company and industry as well as the case process offer the opportunity to conduct a credible, replicable and trustful case study research. In addition, the case selection does not limit the generalization of the findings. On the other side, one could criticize that the case paper of REWAG published by the process mining vendor Celonis could be untrustworthy or biased. Of course, like any other company, Celonis has an interest in advertising and marketing its products in a positive manner by using reference cases. Even when the capabilities and achieved benefits tend to be emphasized in such reference cases an aspirational company like Celonis has no interest in risking its credibility and reputation by distributing false information.
Primary data collection
The main part of my research strategy is the collection of qualitative data. There are different techniques for gathering qualitative data exists, for example, interviews, observational research or action research (Easterby-Smith et al., 2015). However, I decided to conduct interviews because they “[…] enable researchers to access information in context, and to learn about phenomena which is otherwise difficult or impossible to observe” (Easterby-Smith et al., 2015, p. 134). Eisenhardt (1989) and Eisenhardt and Graebner (2007) state that interviews are an efficient method to collect empirical data for case study research. Furthermore, qualitative non-standardized interviews support me to understand the meanings that participants ascribe to various phenomena which is aligned with the taken epistemology of constructionism and my research strategy and approach.
In regards to the interview structure Easterby-Smith et al. (2015) and Saunders et al. (2009) distinguishes between structured, semi-structured and unstructured. I decided to conduct semi-structured interviews which are based on a list of questions covering the research topic (Saunders et al., 2009; Easterby-Smith et al.; 2015). Semi-structured interviews are well-suited to the abductive approach by providing the opportunity let interviewees explain to develop and in parallel build on their responses to test theory (Saunders et al., 2009). In contrast to standardized structured interviews which aim to collect quantifiable data (Saunders et al., 2009), semi-structured interviews for the purpose of my research allow flexibility for a more natural but guided conversation that is also open for the discovery of new significant ideas and themes (Saunders et al, 2009; Easterby-Smith et al.; 2015).
In order to achieve a certain level of guidance a topic guide is a helpful instrument of conducting semi-structured interviews. According to Saunders et al. (2009) data quality issues related to reliability, form of bias, and validity and generalizability, can come up when using semi-structured interviews. Reliability is concerned with whether alternative researchers would reveal similar information (Easterby-Smith et al. 2015). Related to issues of reliability, the risk of bias is present during data collection and interpretation. The topic guide should support avoiding that I impose my own beliefs and frame of reference through the questions I will ask (Saunders et al., 2009)
Selection of interview participants
When I decided to write my master thesis about the topic “Process Mining” I purposefully extended my professional network using the Social Media platforms of LinkedIn and Xing with contacts from vendors of process mining applications as well as end-users and other experts in the area of process mining or BPM. With the help of a selling thesis proposal (see Appendix 1) and a respectful manner of contact I received positive feedback from 35 contacts that would willing to support my research project. I have applied purposive sampling (Easterby-Smith et al., 2015) to select interview participants that cover the topic of process mining from different angles and different levels of expertise. However, only eight interview respondents confirmed the invitation for conducting an interview. Despite that I was able to interview a heterogenous and knowledgeable set of interviewees which could be divided into two groups. The first group “A” consists of four interviewees who are working for process mining vendors and are involved in customer projects. The second group “B” of the other four respondents are two current and knowledgeable end-users of process mining with profound expertise in BPM as well as two respondents with little knowledge about process mining but huge experience in BPM research and the application of business process improvement in a sales context.
Development of topic guide
After selecting the interview respondents, a topic guide needs to be developed. A topic guide serves as the interview preparation and is an important instrument to collect relevant data to answer the research questions. Furthermore, it supports the interviewer to choose an appropriate attitude and language (Easterby-Smith et al., 2015) The topic guide is divided into three parts (Easterby-Smith et al., 2015). The opening questions aim to understand the context and experience of the interview respondents. In addition, the first part should support the establishment of a trustful atmosphere and relationship (Easterby-Smith et al., 2015) by clarifying open questions and the agreement of the informed consent. The informed consent is a crucial instrument to ensure compliance with ethical standards in research, outlined in 3.5.1 Research ethics and quality and at this stage to further enhance the important interview element of trust.
The second part of interview questions is about the relevant key topics. According to Easterby-Smith et al. (2015) questions should be clear and easy to understand. In addition, questions should rather be open-ended in a semi-structured interview (Saunders et al., 2009). Those questions support open-ended answers whereby the interview respondents are given the “space” to make use of real-life examples and experiences (Easterby-Smith et al., 2015). Although open questions avoid becoming biased, probes can sometimes be a useful technique “[…] to improve, or sharpen up, the interviewee’s response (Easterby-Smith et al., 2015, p. 143). Furthermore laddering-up and laddering-down techniques are helpful to gather in-depth details of topics that are of great relevance (Easterby-Smith et al., 2015).
The third part of my topic guide contains closing questions (Easterby-Smith et al., 2015) which are of holistic nature to facilitate correct understanding and interpretation of the interviewee’s answers. Finally, I communicate my appreciation of the respondent’s participation. The German and English versions of the topic guide are outlined in Appendices 2 and 3.
Table of Contents
1.2 Research problem
1.3 Research purpose
1.4 Scope and delimitations
1.6 Definition section
2 Theoretical background
2.1 Process of extracting the relevant literature
2.2 Business processes
2.3 Integrating perspectives on process mining
2.4 Traditional BPM Life Cycle
2.5 Critical factor: business process modeling
2.6 Process mining capabilities
2.7 Impacts of process mining
2.8 Challenges and limitations of process mining
2.9 Motivation for empirical research
3 Research Method
3.3 Methods of data analysis
3.4 Research ethics and quality
4 Empirical results
4.1 Impacts of process mining
4.2 Success factors of process mining
5 Empirical data analysis and discussion
5.1 Impacts of process mining
5.2 Success factors of process mining
6 Illustrative case
6.1 Case company
6.2 Context of the phenomena under investigation
6.3 Process mining in action
6.4 Reflection of identified impacts of process mining
7.3 Managerial implications
8 Limitations and future research
8.2 Future research
GET THE COMPLETE PROJECT
Data-driven business process improvement An illustrative case study about the impacts and success factors of business process mining