Operational Support

Online (runtime) process mining, as distinct from offline retrospective analysis (sources/2011-vanderaalst-process-mining-book, Ch. 10). Operational support provides the conceptual foundation for the entire PPM literature.

Three runtime activities (§10.3–10.5)

Given a running (partial) trace and a reference model:

Detect — notify the moment a running case violates the model (online conformance). Example: alert when a loan application skips the required credit check.
Predict — estimate future values for a running case: remaining time, next activity, case outcome, cost. The direct ancestor of the PPM sub-discipline — concepts/next-activity-prediction, concepts/remaining-time-prediction, concepts/outcome-prediction are all predict operations in Van der Aalst's taxonomy.
Recommend — suggest the action most likely to achieve a goal (minimise time / cost / risk) for a running case. Decision-support layer on top of predict.

Refined process mining framework (§10.1)

Operational support lives in a 3×2 grid — three activities × two regimes:

	Offline (retrospective)	Online (at runtime)
Cartography (build process maps: discovery, enhancement)	Discover a model, annotate with performance	Update maps as new events stream in
Auditing (compare log and model: conformance)	Retrospective deviation reports	Online conformance checking (detect)
Navigation (use models for action)	Simulation, scenario analysis	Predict and recommend on live cases

This 3×2 grid is the Process Mining Spectrum — Van der Aalst's refined reframing of the three original mining types (discovery / conformance / enhancement) in terms of the use of the model rather than the direction of the mining.

Processes are not in steady state (§10.6)

Operational support must handle:

Daily / weekly / seasonal patterns — arrival rates, resource availability fluctuate.
Contextual factors — exogenous variables (weather, macroeconomy, campaigns) shift behaviour.
Concept drift — the process itself changes over time (new product line, policy, law). Drift can be sudden, gradual, recurring, or incremental. Handling drift is a named sub-problem in the PPM literature (e.g., sources/2019-verenich-survey-ppm discusses this).

Relation to PPM

PPM = operational support narrowed to the Predict activity. The PPM synthesis treats this book's Ch. 10 as the seed of the field: everything in raw/Predictive process monitoring/ elaborates Van der Aalst's predict mode with richer encoders (LSTMs, transformers, GNNs), richer targets (outcomes, remaining time, next activity), and richer inputs (inter-case features, text, resource data).

methods/process-mining-basics · concepts/conformance-checking · concepts/predictive-process-monitoring · concepts/next-activity-prediction · concepts/remaining-time-prediction · concepts/outcome-prediction · concepts/process-mining-spectrum · concepts/mape-k-loop (the agentic-BPM cousin of detect/predict/recommend)