From Process Mining to Thinking Assistants in Logistics

(a) Situation faced: Vanderlande Industries BV, or Vanderlande for short (VI), is a global market leader for end-to-end material handling solutions (MHS) increasing customer expectations regarding system performance while also increasing system capabilities and complexities. Although process mining fundamentally enables data-driven insight into the causes of performance problems, preexisting process mining technology assumed that processes were single object and executed in isolation. However, MHS are complex multilayered systems with multi-object processes that operate across shared equipment. Thus, despite its potential, VI’s engineers found very limited room for improving MHS performance through classical process mining. (b) Action taken: VI initiated and funded a long-term research collaboration between VI’s process engineers and data scientists and TU Eindhoven’s (TU/e’s) process mining research group involving the authors. After receiving access to system data, users, and use cases, we followed the design science research method to develop and validate novel process mining techniques and tools that provide VI engineers with the information they need to increase system performance. (c) Results achieved: We developed several novel process mining techniques and tools that overcome the limitation of classical process mining by analyzing all cases and their interdependencies within a system together over both time and shared equipment. Validation through case studies confirmed the effectiveness of the tools and several tools are now being industrialized. The overall insights from the project led to the development of the concept of a “thinking assistant” that supports operators in decision-making. (d) Lessons learned: The joint collaboration between industry and research not only enabled the development of a paradigm shift in process mining research, it also facilitated a transition from standard industry tooling for process mining to custom solutions designed for specific use cases. We also identified several large research challenges for process mining directly driven by customer needs.