Process Mining
Process mining derives process models and performance facts from the event logs that information systems already record, instead of from interviews and workshops. It was shaped largely by Wil van der Aalst and the IEEE Task Force on Process Mining.
Three kinds of process mining
| Type | Input | Output | Question |
|---|---|---|---|
| Discovery | Event log | Process model | What does the process actually look like? |
| Conformance checking | Event log + model | Deviations | Where does reality differ from the rule or model? |
| Enhancement | Event log (+ model) | Improved or enriched model | Where are the bottlenecks, delays, rework? |
Event logs
The minimum is a case identifier, an activity name and a timestamp per event; resource, cost and other attributes add analysis power. Logs usually come from ERP, CRM or ticketing systems and need cleaning. The interchange standard is XES, published as IEEE 1849-2016 and revised as IEEE 1849-2023.
Object-centric process mining
Classic logs assume one case per event (a single order). Real processes mix orders, deliveries, invoices and payments, so one event touches several objects. Object-centric process mining records events against multiple objects. OCEL 2.0, a standard format for object-centric event logs, was specified in 2023 (JSON, XML and SQLite formats) by van der Aalst’s group and co-authors; Celonis announced an object-centric data model in 2023 (celonis).
Tools
celonis (commercial), Disco (Fluxicon), ProM and Apromore (open source), PM4Py (Python library), and process mining inside platforms such as sap-signavio and UiPath (uipath). Details in process-mining-tools.
Where it fits
- Discovery and monitoring phases of the bpm-lifecycle, replacing much manual as-is modelling.
- Selecting automation targets: find high-volume, repetitive paths before building RPA bots (hyperautomation-and-intelligent-automation).
- Compliance: conformance checking against the intended BPMN model.
- AI: LLMs are being applied to process mining tasks (ai-and-bpm).
Caution: logs show what happened in systems, not work done outside them (email, phone, paper). It does not replace talking to the people doing the work.
Related
theory-of-constraints, six-sigma-and-dmaic, BPM hub
Sources
- IEEE Task Force on Process Mining, Process Mining Manifesto, 2011: https://www.vdaalst.rwth-aachen.de/publications/p907.pdf
- XES standard site: https://www.xes-standard.org/
- OCEL 2.0 specification: https://arxiv.org/abs/2403.01975 and https://www.ocel-standard.org/
- van der Aalst, Process Mining: Data Science in Action, Springer, 2016