Earlier product guidance · version not specified
Which fields each analysis needs
Available fields determine which analyses can run. Declare event order with a step number or sequence column; timestamps do not substitute for order.
- Work object, step order, activity, start time
- Paths and lead time
- Add an end time
- Waits between steps, processing time, work in progress
- Add a person and a team
- Handovers and role-separation checks; hours also need work-time evidence
- Add a due date
- On-time share, past due and forecast risk
Five checks before you run
Check identity, order and time in the source records before calculating process measures.
Keep work objects distinct.
A shipment, order and invoice can be related without being the same work item. Declare object links and their grain.
Declare event order.
A CSV's physical row order is not necessarily a business sequence. Use declared order evidence.
Preserve missing timestamps.
If a step has no end time, it does not have a zero-duration interval. Leave the missing value empty.
Identify each event.
Repeated exports, reopened work and corrected records can change the path. Establish event identity.
Separate waiting from working time.
A gap between steps does not establish employee work time. Use declared calendars and duration evidence to determine what can be measured.
Reference and example
External reference
Microsoft's process mining data preparation guide explains why operational history, not just current state, is needed.
Inspect the fields in a worked sample.
Open the published process map, rule checks and source-linked result.
Inspect the worked sampleAn event log is different from a current-state export
An export containing one row per open ticket may describe only the latest state. It cannot establish earlier transitions that were never exported. A history with one row per recorded activity supports different questions from an object list with creation and closure fields.
Check what one row represents before assigning it a role. Use the role, domain, grain and authority declarations to distinguish event history, work intervals and object records. The event-log example shows several events belonging to one claim.