Schlüchtermann, JörgRöglinger, MaximilianWynn, MoeDun, Christopher vanChristopher vanDun2023-02-222023-02-222022urn:nbn:de:bvb:703-epub-6418-4https://publica.fraunhofer.de/handle/publica/43630410.15495/EPub_UBT_00006418Business processes are at the core of every organisation’s effort to deliver services and products to customers and, thus, achieve the organisation’s goals. The discipline that deals with the design, analysis, execution, and improvement of such business processes is called business process management (BPM). Over the years, the BPM research discipline has created a large number of methods and tools to support practitioners in managing and improving their business processes. In recent years, the increasing abundance of process data available in organisational information systems and simultaneous progress in computational performance have paved the way for a new class of so-called data-driven BPM methods and tools, the most prominent of them being process mining. This cumulative doctoral thesis concentrates on two challenges related to data-driven BPM methods and tools that impede faster and more widespread adoption. First, while data-driven methods and tools have found quick adoption in BPM lifecycle phases such as process discovery and process monitoring, the lifecycle phase of process improvement has so far been neglected. However, process improvement is considered to be the most value-adding BPM lifecycle phase since it is the necessary step to address existing issues in as-is processes or to adapt these processes to constantly changing environments and customer needs and expectations. Process improvement is often expensive, time-consuming, and labour-intensive, which is why there is a particular need to support process stakeholders in redesigning their processes. Second, there is a need for high-quality process data in all phases of the BPM lifecycle. In practice, process data, e.g., in the form of event logs for process mining, is often far from the desired quality and process analysts spend the majority of their time on identifying, assessing, and remedying data quality issues. Thus, in the BPM community, the interest in exploring the roots of data quality problems and the related assurance of high-quality process data is rising. Hence, it is essential to have a means for detecting and quantifying process data quality. Against this backdrop, this cumulative doctoral thesis comprises five research articles that present advances in process data quality management on the one hand and data-driven process improvement on the other hand. Taking on a design-oriented research paradigm and applying different qualitative and quantitative research methods, this thesis proposes several IT-enabled artifacts that support stakeholders in managing process data quality and improving business processes. The insights contained in this thesis are relevant for academia and practice as they provide both scientific perspectives and practical guidance. Concerning process data quality management, research article #1 presents an approach for (semi-) automated and quality-informed event log extraction from process-agnostic relational databases. It applies metrics for data quality dimensions that are relevant to process mining in order to quantify the data quality of the source data in selected database tables and simultaneously allows users to extract event logs in XES format from the database tables. Research article #2 presents an approach for detecting and quantifying timestamp data quality issues in events logs already present in XES format. The approach applies metrics for identifying timestamp imperfection patterns and allows users to interactively filter, repair, and annotate the event log. Furthermore, this thesis provides several concrete approaches to data-driven business process improvement. First, it focuses on process improvement in itself and aims to create artifacts for supporting process improvement initiatives. Therefore, research article #3 provides a model based on generative adversarial networks to create new process designs. Specifically, it uses event logs and annotated information on process variants and process deviance to generate a new process model which provides suggestions for process improvement to the user. Second, this thesis targets data-driven decision support in business processes. In particular, research article #4 uses multi-criteria decision analysis to extend traditional vehicle routing problems in last-mile delivery with a customer-centric perspective. The customer-centric vehicle routing uses process and customer data and the concept of customer lifetime values to predict customer satisfaction and, thus, optimise delivery routes. Finally, research article #5 presents a modelling approach for IT availability risks in smart factory networks based on Petri nets. The modelling approach uses modular components of information systems and production machines to model, simulate, and analyse production processes. The thesis concludes by pointing to limitations of the presented research articles as well as directions for future research. Overall, this thesis contributes to several important research streams in BPM while applying a broad range of qualitative and quantitative research methods such as simulation, normative analytical modelling, multi-criteria decision analysis, and interview studies within an overarching design science research paradigm. It builds upon and extends existing research on process data quality management and business process improvement.endoctoral thesis