Process or case? — the litmus test
Who owns the order of work?
- The organisation (“always X before Y”) → process (BPMN).
- The practitioner, within guardrails (“depends on the case”) → case (CMMN).
Apply per stage, not per system — most flows are processes containing at most one discretionary pocket (model that pocket as a case, or the case calls processes: lesson 03).
Signals
| Toward process | Toward case |
|---|---|
| regulator/policy mandates sequence | expert judgment drives sequence |
| same path for 95% of instances | “every one is different” is actually true |
| SLA per step | SLA on the outcome only |
| steps mostly automated | steps mostly human investigation |
| deviations are exceptions to route | deviations are the normal mode |
Traps
- Hairball BPMN — loops + inclusive gateways emulating discretion: you wanted a case (or one case-shaped stage).
- Fake CMMN — a case whose sentries chain every item into a fixed sequence: you rebuilt BPMN with worse tooling; go back.
- “Dynamic” as an excuse — teams claim discretion to avoid agreeing on a process; the litmus test forces the real conversation about who should own the order.
- Discretion without guardrails — a case with no required items, no sentries, no milestones is just a shared to-do list; the value is the guardrails.