Navigating the Landscape of Organizational Process Mining Setups

PM technology has emerged at the intersection of process science and data science (van der Aalst 2016; vom Brocke et al. 2021b). It leverages digital footprints left by process steps in event logs within an organization’s IT systems (van der Aalst et al. 2012). The technology analyzes event logs both retrospectively (e.g., to identify bottlenecks in the process) and predictively (e.g., to forecast the remaining duration of a process instance during execution) (van der Aalst and Carmona 2022). PM is strongly intertwined with other advanced process technologies. For instance, robotic process automation (RPA) is used to automate tasks and close execution gaps based on insights gained from PM (Hofmann et al. 2020; van der Aalst et al. 2018). Task mining can be understood as PM based on user interaction data with digital systems, including mouse clicks and screen activities (van der Aalst et al. 2020b). As such, we understand PM as an umbrella term for a set of advanced technologies leveraging event data, thus enabling insights into the various operations of companies, ranging from end-to-end processes to individual activities (Hofmann et al. 2020; van der Aalst 2016, 2020b; van der Aalst et al. 2018). The insights enabled by PM are valuable since they help identify process deviations, uncover wasteful resource utilization (e.g., time, costs, or personnel), or identify process inefficiencies (Martin 2021). These findings allow for targeted process improvements with various economic benefits such as cost reductions, time savings, and increased customer satisfaction (Grisold et al. 2021; van Dun et al. 2023).

Research on the managerial and organizational aspects of PM has mainly relied on case studies (Martin et al. 2021). For instance, Andrews et al. (2020) explored PM’s use in analyzing patient transport pathways after traffic accidents, highlighting key findings and lessons learned. Reinkemeyer (2020) compiled 12 successful PM implementations, detailing their impact on specific processes and organizational implications. Case studies have also been analyzed to understand PM applications on an industry level. Rojas et al. (2016) analyzed 74 case studies in healthcare, highlighting the necessity of adapting PM to the specific industry context, suggesting integration into process-aware hospital information systems as one strategy.

Besides case study research, there is a body of literature on PM’s application in various organizational settings. Grisold et al. (2021) conducted a focus group study on practitioners’ requirements and expectations of PM, highlighting the importance of considering organizational and managerial implications, as well as leadership and governance factors. Similarly, Martin et al. (2021) conducted a Delphi study, identifying challenges and opportunities for PM in organizations that require future research, such as a lack of clear guidance and organizational alignment in PM implementation efforts (Martin et al. 2021).

In practice, centers of excellence (CoEs) are commonly used to describe PM setups (Reinkemeyer et al. 2022). They oversee the organization-wide integration of PM, providing services like steering use cases and offering data expertise (Lillig 2020). Despite being a frequently applied organizational setup (Balint et al. 2020; Lechner 2020; Lillig 2020; Reinkemeyer et al. 2022), CoEs are just one possible PM setup and are not clearly defined (Reinkemeyer et al. 2022). Given the lack of a universal definition and understanding of a CoE, we introduce the concept of a PM unit as a broad term to describe the range of all possible PM setups, including a CoE. A PM unit is the organizational entity responsible for conducting PM services, allowing for various organizational characteristics and structures tailored to a firm’s specific needs and context. The PM unit may span different teams or business units, depending on where the individuals in the PM unit are located in the organization. It enables the organization to leverage PM techniques and tools effectively, driving insights, process improvement, and informed decision-making.

To integrate PM into an organization’s structure, we draw on knowledge from BPM governance since PM augments from BPM as a technology that requires reliable event logs for analysis (van der Aalst 2016). BPM governance, as part of BPM, refers to the introduction of roles, structures, and metrics. It is considered a BPM success factor, as the introduction of BPM governance elements is a necessary foundation for the execution of BPM activities (Kerpedzhiev et al. 2021; Rosemann and Brocke 2015; Spanyi 2015). BPM governance enables organizations to properly define an organizational setup for BPM, which helps to clarify responsibilities such as business process owners who are accountable for a whole business process, covering areas such as planning, process performance management, and stakeholder management (Danilova 2019). Governance in BPM can be implemented at the process level (Markus and Jacobson 2015) or the organizational level (Rosemann and Brocke 2015).

The introduction of PM in an organization affects governance structures and challenges existing governance models in BPM (vom Brocke et al. 2021a). PM transforms the work on processes by shifting the focus from traditional tasks such as process modeling to real-time knowledge generation and decision-making for process improvement (vom Brocke et al. 2021a). This transformation also affects the roles and responsibilities in a PM unit. Since existing BPM governance models and approaches focus on tasks of the traditional BPM activities lifecycle (de Boer et al. 2015), they do not address the changes in structures and roles that arise with the introduction of PM. First approaches in the BPM literature have covered single elements of governance in PM, such as data governance (Goel et al. 2021), or case studies that described organizational anchoring, such as the introduction of CoEs (Balint et al. 2020; Lillig 2020), but have not yet taken a holistic perspective on the governance structures in PM. One of the pressing issues in PM is to investigate these governance structures at an organizational level (vom Brocke et al. 2021a). A well-defined governance structure is essential for PM since it properly attributes responsibilities among units and teams, ensuring efficient collaboration between the PM team and, for instance, operations management (Jansen 2020).

Owing to the lack of a comprehensive conceptualization of governance structures in PM, we draw on justificatory knowledge of BPM governance (cf. Section 3 and Appendix B; appendices are available via http://​link.​springer.​com). In the following, we acknowledge seminal contributions from the BPM governance literature that have shaped our taxonomy. While not all concepts were adopted verbatim, each contributed valuable insights that informed our discussions. For instance, Rosemann (2015) describes fifteen distinct services of a BPM CoE, including activities such as change management or governance. Fischer et al. (2020) report on adopting BPM in the context of digital transformation and highlight the impact of (de)centralization on company operations. Nqampoyi et al. (2016) investigate the factors influencing the effectiveness of BPM CoEs. For instance, they provide insights on team structures, reporting lines, and role allocation. Harmon (2016) showcases the heterogeneity of organizational anchoring while demonstrating the widespread support from external parties in BPM. Studies by Santana et al. (2011) and Valença et al. (2013) underscore the importance of internal leadership and executive sponsorship by investigating BPM governance in the context of public organizations. The planning model proposed by Lehnert et al. (2016) provides a systematic approach to project prioritization. For instance, they state that funding for BPM projects can originate from multiple budgets. Lastly, the BPM capability framework by Kerpedzhiev et al. (2021) showcases the relevance of process data governance, such as data ownership structures. Similarly, their updated capabilities on methods and IT illustrate the importance of tool ownership in BPM.

To compile organizational characteristics for PM setups, we developed and evaluated a taxonomy over four iterations, drawing on the taxonomy development methods of Nickerson et al. (2013) and the updated version as per Kundisch et al. (2021). As a system “[…] of groupings that are derived conceptually or empirically” (Nickerson et al. 2013, p. 338), a taxonomy allows one to conceptualize a phenomenon in terms of dimensions and characteristics and serves as a theory for analyzing (Gregor 2006; Kundisch et al. 2021). As part of theory-building, some studies view taxonomy development from the perspective of theories such as configuration theory (El Sawy et al. 2010; Meyer et al. 1993). Configuration theory applies a holistic and systemic perspective that views phenomena in organizations as falling into coherent patterns as part of a collection of possible configurations (Fiss 2007; Meyer et al. 1993). Studies that apply configuration theory in BPM, for instance, measure a configuration’s specific outcome, such as the degree of business process orientation, and develop archetypes based on the developed taxonomy (van Looy et al. 2022). While configuration theory provides a valid perspective on organizational PM setups as the phenomenon studied in this work, we developed our taxonomy without applying the configuration perspective since we focused on an initial exploration of organizational PM setups instead of empirically derived archetypes. Nonetheless, we refer to configuration theory as a foundation for future research in our theoretical implications.

In IS research, Nickerson et al.’s (2013) taxonomy development is a fundamental method widely applied in the literature (Kundisch et al. 2021). Kundisch et al. (2021) integrated and extended Nickerson et al.’s (2013) method by adding structured guidance between steps and additional phases for evaluation. Figure 1 presents the taxonomy development approach and how we applied it.

Following Kundisch et al. (2021), we first defined the taxonomy’s target group and its intended purpose. As stated, we investigate organizational PM setups to lay a theoretical groundwork in the organizational domain for PM and provide practitioners with an overview of and orientation regarding possible PM setups. As for contextual factors from an organizational perspective (vom Brocke et al. 2015), the taxonomy targets large organizations at any level of PM maturity. Its application transcends specific industries. We expect the taxonomy to be applied by individuals responsible for overseeing PM initiatives, such as Heads of PM.

For the development of a taxonomy, a meta-characteristic serves as a basis to determine the inclusion of characteristics, each logically derived from the meta-characteristic (Nickerson et al. 2013). In our study, we define characteristics of organizational PM setups as meta-characteristic. Regarding the subjective ending conditions of the taxonomy development process, we followed Nickerson et al.’s (2013) recommendation and opted for concise, robust, comprehensive, extendible, and explanatory as these characteristics form the necessary conditions for a taxonomy to be useful by providing a strong foundation for descriptive evaluations. Concerning the objective ending conditions of the taxonomy development process, we opted for three (Nickerson et al. 2013): every dimension is unique and not repeated; every characteristic is unique in its dimension; no new dimensions or characteristics were added in the last iteration. These were chosen to ensure uniqueness and stability within the taxonomy structure. By specifying that each dimension must be unique and non-repetitive, we wanted to avoid redundancy and maintain a clear conceptual distinction between the different aspects of the taxonomy. Ensuring that each characteristic is unique within its dimension further reinforces this by preventing overlap within the dimensions. Finally, the condition that no new dimensions or characteristics were added in the last iteration indicates the taxonomy’s maturity and completeness, suggesting that it has reached a point of sufficient depth and detail to represent the domain in question comprehensively.

We opted for an E2C (empirical-to-conceptual) iteration as the starting point of taxonomy development. While we initially reviewed the literature on governance in PM, owing to limited conceptual knowledge and very few approaches that specifically address governance structures for PM, we followed Nickerson et al.’s (2013) recommendation to start with an E2C iteration. E2C iterations are characterized by starting with empirical evidence, identifying patterns, and abstracting them into broader concepts, hence creating a conceptual understanding based on real-world objects. An overview of all taxonomy development iterations can be found in Table 1.

For the first set of dimensions and characteristics in iteration 1 (E2C), we drew on knowledge from an international study of PM CoEs conducted among 214 organizations from multiple industries and regions with different approaches and maturity levels in their PM journeys (Reinkemeyer et al. 2022). Thus, we analyzed organizational aspects such as the operating models, organizational setups, roles, responsibilities, value propositions, and budgeting. Although these insights were mainly based on CoEs as one possible organizational PM setup, these findings gave us valuable guidance on which dimensions to consider when introducing a PM setup. With the first version of our taxonomy, we assessed the objective ending conditions of our taxonomy development. Since new dimensions and characteristics were added in this iteration, we continued our taxonomy development.

In iteration 2 (E2C), we continued with another E2C iteration since we sought to gain more empirical insights into PM setups after developing a meaningful set of dimensions that can serve as a basis for interviews. We conducted 15 semi-structured interviews (Myers and Newman 2007). Table 2 provides an overview of our interview partners. We opted for purposive sampling (Etikan 2016) of PM experts in our professional networks, choosing organizations that reflect various industries and PM adoption levels to gain insights from adopters, vendors, and consultants to capture the various available PM setups. Thus, we ensured comprehensive and in-depth observation of PM setups in organizations. We developed a questionnaire (cf. Appendix D) based on version 1 of our taxonomy. At least two authors were present in every interview. On average, each interview lasted around 45 min. All interviews were conducted online via Microsoft Teams. Following Myers and Newman (2007), we started each interview by outlining our research topic and objective. We then followed our questionnaire (Appendix D), asking the interviewees about their current PM setups and how specific components, such as roles, are determined. Since we conducted semi-structured interviews, we allowed for additional questions to be asked if, for instance, an interviewee raised a topic not included in the interview questionnaire. After all 15 interviews, selective coding was used (Wolfswinkel et al. 2013) to aggregate our findings at the levels of characteristics and dimensions. We then discussed our findings and compared them to our initial taxonomy version. After this iteration, we added dimensions, such as the external perspective for organizations (e.g., external support). A third iteration was needed since we had added dimensions and characteristics in iteration 2, violating an objective ending condition. As we repeatedly found similar themes and no new conceptual categories emerged from the interview data after a certain point indicating theoretical saturation (Corbin and Strauss 2008), we pursued a C2E (conceptual-to-empirical) iteration. In C2E iterations, concepts are proposed without examining actual objects but using existing scientific foundations and personal experiences. Afterwards, the proposed concepts are applied to real-world observations to assess their appropriateness.123456

In iteration 3 (C2E), we conducted a literature review to substantiate further and, if necessary, extend our current taxonomy version. For this review, we drew on the PM and BPM governance literature and selected studies from organizational science (Vithayathil and Choudhary 2022). We added layers and guiding questions, enriching certain dimensions with further references. Although no dimensions and characteristics were added in this iteration, and thus no further objective ending condition was violated (Kundisch et al. 2021), we considered the subjective ending conditions comprehensiveness and robustness to be violated, thus requiring a fourth iteration. According to Nickerson et al. (2013), a comprehensive and robust taxonomy can classify all known objects and can clearly differentiate between the objects of interest. We noticed ambiguity in the naming of characteristics, which caused discussions among the authors. Thus, it would likely also hinder users from correctly classifying their PM setup and consequently have a negative effect on the perceived usefulness. Hence, we concluded that the taxonomy was not comprehensive and robust enough, thereby violating two subjective ending conditions and necessitating another iteration of taxonomy development.

After the taxonomy was developed, we evaluated it to ensure the transferability and applicability of our results (Kundisch et al. 2021). Our evaluation activities were threefold: First, we conducted semi-structured interviews (Myers and Newman 2007) with twelve experts for in-depth qualitative feedback. Additionally, we carried out an anonymous online survey with the same experts for supplementary quantitative insights. As for applying our taxonomy, we conducted four additional in-depth interviews with PM adopters. All three activities are considered relevant evaluation methods for taxonomies (Szopinski et al. 2019).

For the evaluation interviews, we contacted the 15 interviewees from iteration 2, ensuring accurate reflection of their insights into the taxonomy. Out of these, eight experts agreed to participate in the evaluation interviews (I2, I3, I6, I7, I10, I12, I13, I15). The remaining seven either did not respond, were unavailable during the interview timeframe, or were no longer engaged with PM. Following Kundisch et al.’s (2021) recommendation, we also introduced four new interview partners (I16-I20) who were not part of the initial taxonomy development. After conducting twelve interviews overall, we concluded the evaluation, as the last three interviews did not yield new insights, indicating saturation (Saunders et al. 2018). In line with the expert selection in iteration 2, our sample selection aimed for diversity, encompassing adopters, consultants, and vendors (cf. Table 2). Each 30 min interview began with an explanation of the research method, emphasizing its descriptive character, purpose, and the concept of a PM unit. We clarified that we envisioned the taxonomy application for each PM unit separately and outlined its overarching structure with layers, dimensions, characteristics, guiding questions, and the meaning of exclusive and non-exclusive dimensions. Dimension by dimension, we then outlined each characteristic with examples, encouraging experts to identify ambiguities, omissions, or elements they deemed important or relevant. Most experts assessed their own setups (adopters) or their customers’ setups (vendors, consultants) to evaluate these aspects and the taxonomy’s applicability. All interviews were recorded and transcribed for analysis.

Following the interviews, all twelve experts also participated in an anonymous survey, rating the taxonomy for understandability, completeness, and usefulness (for definitions of these terms, we refer to Fig. 2). The survey aimed to complement the qualitative insights gathered dimension-by-dimension with an overall quantitative rating. Based on the input from the interviews and the survey, the taxonomy underwent refinement (cf. Appendix A), resulting in our final taxonomy for organizational PM setups.

To further substantiate our findings, we conducted four semi-structured in-depth interviews with adopters (I20-I23) who did not participate in the initial iteration 2 interviews or the evaluation interviews. These interviews provided initial insights into the application of the taxonomy and are presented as three individual exemplary cases (two participants are from the same organization, i.e., one case) in Sect. 5.2 (Kundisch et al. 2021; Limaj and Bernroider 2022).

We conducted twelve 30 min evaluation interviews to evaluate the taxonomy and collect qualitative feedback on a dimension-by-dimension basis. The feedback we received was largely positive. Participants found the taxonomy helpful, as it allowed them to view organizational PM setups from a broad perspective. They appreciated the clarity and comprehensiveness of the taxonomy as well as its practical applicability. Some suggestions for improvement and enhancement included the need for additional clarification of certain characteristics, exploring the reasons behind different organizational setups, and considering the potential for further research based on the taxonomy. A summary of the comments from the interviews is provided in Table 8. Please refer to Appendix A for a comprehensive list of the experts’ suggestions on a dimension-by-dimension basis and their incorporation into the taxonomy.

Following the interviews, we conducted an anonymous online survey among the interviewees to gather additional quantitative feedback. The survey showed the taxonomy and outlined three evaluation criteria: completeness, understandability, and usefulness (Kundisch et al. 2021), each accompanied by a definition to ensure comparability between individual ratings. Participants were asked to assess the taxonomy against these criteria using a five-point Likert scale (from strongly agree to strongly disagree). The survey results, illustrated in Fig. 2, revealed strong ratings for all criteria. The vast majority rated completeness and understandability as strongly agree (92%), with only one expert rating these as agree (8%). Usefulness was rated with either strongly agree (58%) or agree (42%). No responses fell under neither agree or disagree, disagree, or strongly disagree. The slightly lower ratings for usefulness were attributed to the partially generic naming of characteristics in the dimension key activities, which, however, was a deliberate design decision to ensure the taxonomy’s broad applicability. In response to the feedback, we slightly refined and enhanced the naming where reasonable (cf. Appendix A, Appendix B, Iteration 4). Additionally, one expert (I15) noted that achieving completeness is challenging given the diverse nature of organizational structures and their constant evolution.

In addition to the quantitative assessment, the experts had the opportunity to provide free text comments. They highlighted the taxonomy’s understandability and usefulness positively, with one anonymous expert stating, "The taxonomy is very easy to understand and also very helpful in practice apart from the minor additions (e.g., the addition to demand generation) that I mentioned in the interview." (anonymous). Furthermore, they emphasized its practical use, noting, "One thing this framework will proactively do is push the PM team to understand the client’s org structure, as it is a crucial and important thing to ‘know your client.’” (anonymous). Experts also suggested leveraging the taxonomy as a foundation for future research, stating, “I can very well imagine that it is an excellent basis for further research questions, e.g., how the degree of maturity of a process mining initiative/CoE affects the classification in the taxonomy. Do we see certain patterns over time?” (anonymous).

Following the evaluation of the taxonomy, we applied it in three exemplary cases (Kundisch et al. 2021; Limaj and Bernroider 2022). Thus, we sought to showcase the taxonomy’s applicability to PM setups in diverse organizational contexts as well as to demonstrate the taxonomy’s ability to guide discussions on the rationales for specific PM setups. Drawing on inputs from the interviews during the taxonomy development process (I1, I6, I7, I8, I9, I10, I11, I13, I14), it appeared that organizational PM setups evolve with more processes being analyzed and PM being adopted more broadly in an organization. Thus, we selected three organizations with varying degrees of experience in terms of time since introducing PM, the number of analyzed processes, and the number of active users of the PM tool (cf., Table 9). While not explicitly referring to maturity levels such as those known from process or BPM maturity models (Röglinger et al. 2012), we classified the three organizations according to the abovementioned criteria as beginner, intermediate, or experienced. The beginner company has analyzed two processes since introducing PM a year before, with just over ten regular users. The intermediate company characterized itself as being in the middle of the adoption spectrum, with five analyzed processes and slightly less than 50 active users. (“I would say we’re in the middle. We’re not at the beginning. [But] there are other companies that […] have more maturity than us.”) (I22). The experienced company was an early adopter of PM, with more than five years of experience, five analyzed processes, and more than 100 active PM users. We will provide a detailed account of the rationale behind the choice of its organizational PM setup for each of these companies.

We will now provide a conclusive evaluation of the taxonomy. As substantiated by the evaluation activities presented in Sect. 5.1, we consider the taxonomy useful, comprehensible, and sufficiently complete. Furthermore, the three exemplary cases attest to the taxonomy’s efficacy in characterizing various organizational PM setups and their real-world applicability.

When analyzing the cases, we made three observations. First, there is a high variability in organizational PM setups. Moreover, there is no one-size-fits-all PM setup. Rather, PM adopters must identify the setup that helps them cope with and leverage their individual context (Zelt et al. 2018). Our observations indicate that PM setups are contingent on numerous contextual factors, such as company size, industry, available resources, and budget. Second, the PM setups were not static across all cases but underwent evolutionary changes over time, particularly with increasing PM adoption and the number of processes analyzed. Third, certain characteristics – such as the presence of an executive sponsor and business champions – can yield benefits, irrespective of the respective context and maturity. For instance, I8 stated, "The most important thing about a topic like this […] is that, in addition to pure sponsorship, in other words, in addition to the executive sponsor who provides the budget for it, someone is needed who actively promotes this topic.”

Since the taxonomy was designed and evaluated by analyzing diverse PM setups of established organizations operating in various contexts, we claim that it can be applied in similar contexts. To assess the taxonomy’s transferability, we draw on the generalization framework proposed by Lee and Baskerville (2003), which differentiates four distinct strategies for generalization: data-to-description (EE), theory-to-description (TE), description-to-theory (ET), and concepts-to-theory (TT). Drawing on empirical insights in iterations 1 and 2 of the taxonomy development process, we developed a taxonomy that generalizes data to description (type EE generalizability), facilitating the characterization of organizational PM setups.

To answer our research question on organizational PM setups’ characteristics, we proposed a multilayer taxonomy, laying the foundation for further research into PM’s organizational aspects.

From a theoretical perspective, our taxonomy has contributed to the descriptive knowledge of PM and BPM governance, delving into an underexplored research area (Martin et al. 2021). Our main contribution is a theoretically well-founded taxonomy developed with the active involvement of PM experts. Our taxonomy is a theory for analyzing (type I) (Gregor 2006), offering descriptive insights into various organizational PM setups and presenting relevant dimensions. Further, by elaborating on selected cases, we have demonstrated that PM setups exhibit considerable variability, driven by the multifaceted organizational characteristics and contextual factors that influence their design and implementation (Reinkemeyer et al. 2022). This underlines that conventional BPM governance approaches cannot fully capture PM setups’ intricacies. From a practical perspective, we have provided a holistic overview of the relevant dimensions and characteristics of a PM setup and possible manifestations. Diving into selected practical examples of a diverse group of organizations, we shed light on the rationales involved in designing and developing an organizational PM setup.

We have made two primary theoretical contributions: We have advanced the research at the intersection of PM, BPM governance, and organizational design, providing a solid foundation for further theorizing on organizational aspects of PM.

First, we suggest that research into PM benefits from a holistic perspective that exceeds existing studies’ focus on specific elements or setups. As we have observed by the variety of possible characteristics, our findings suggest investigating the relationships between certain characteristics to see whether there are set combinations of characteristics that occur in specific industries or organizations of specific sizes. Future endeavors could extend the body of knowledge on integrating PM into an organization’s structures or the empirical derivation of archetypes through a configuration perspective. Future endeavors could extend the body of knowledge on integrating PM into an organization’s structures or the empirical derivation of archetypes (Wanner et al. 2023). Similar to other configuration theory studies (Van Looy et al. 2022), these PM archetypes can be used to measure the relationships between specific configurations and performance.

Second, our results can serve as a basis for further theorizing on organizational and managerial aspects of PM and can be used by other scholars to develop higher-level theories (Gregor 2006) in this regard. The three exemplary cases demonstrate how explanatory knowledge can be created based on our taxonomy, i.e., an organization’s rationale for certain PM configurations. Given several such indications in the initial interviews, the evaluation interview, and the exemplary cases, the links between common PM setups and organizational characteristics (e.g., size, industry, or maturity level) should be investigated. Researchers could indicate the extent to which an organization’s contextual factors influence its choices of certain organizational characteristics. This could provide further guidance on which setup may suit a certain organization type at a particular point in its PM adoption journey or depending on its maturity level. Lastly, as we have broken down the relevant dimensions and characteristics of organizational PM setups, we have laid a foundation for further theorizing on the specific effects of single characteristics, such as choosing an organizational anchoring at the executive rather than the business level. Our taxonomy supports practitioners in approaching the establishment and integration of PM in an organization. It fosters structured discussion and justifies the relevant dimensions and characteristics of organizational PM setups. Besides describing a current PM setup, it also provides orientation in determining future characteristics of an organization’s PM setup.

While our focus is on PM, we acknowledge that the taxonomy’s dimensions and characteristics may have applicability beyond the PM domain. We claim validity for PM within the scope of our research, but we do not assert exclusivity. It should be investigated whether some findings also apply to similar disciplines within data analytics and technology-driven process improvement.

Our work has limitations, which affect our findings’ transferability and applicability. First, our sample selection for the interviews during the taxonomy development did not fully cover various organization sizes and industries. Researchers could study a larger sample to enrich our insights with size- or industry-specific conclusions. Second, applying the taxonomy could be challenging, especially in organizations with complex PM setups or structures that are unknown to a user. This complexity is further increased when multiple (partially) independent distinct PM setups operate in different parts of the organization. In such cases, the taxonomy would need to be applied repeatedly to capture the diverse setups and variations. Further, organizations implementing PM may not have all the characteristics of their PM setup (adequately) defined. Thus, we acknowledge that applying our taxonomy may be complex, potentially limiting practitioners’ adoption of our approach.