A practitioner’s guide to process mining: Limitations of the directly-follows graph

Process mining techniques use event data to show what people, machines, and organizations are really doing. Process mining provides novel insights that can be used to identify and address performance and compliance problems. In recent years, the adoption of process mining in practice increased rapidly. It is interesting to see how ideas first developed in open-source tools like ProM, get transferred to the dozens of available commercial process mining tools. However, these tools still resort to producing Directly-Follows Graphs (DFGs) based on event data rather than using more sophisticated notations also able to capture concurrency. Moreover, to tackle complexity, DFGs are seamlessly simplified by removing nodes and edges based on frequency thresholds. Process-mining practitioners tend to use such simplified DFGs actively. Despite their simplicity, these DFGs may be misleading and users need to know how these process models are generated before interpreting them. In this paper, we discuss the pitfalls of using simple DFGs generated by commercial tools. Practitioners conducting a process-mining project need to understand the risks associated with the (incorrect) use of DFGs and frequency-based simplification. Therefore, we put these risks in the spotlight.