What is process mining?
Process mining is a technique that rebuilds the real map of how work flows through a business by reading the timestamps its software already records, rather than by asking people how they think it works.
Also called process discovery, automated process discovery.
Process mining in plain words
Every system stamps a time on what it does: an order raised at 09:14, an invoice approved at 16:02 on Thursday, a payment released eleven days later. Process mining collects those stamps into an event log and draws the path each case actually took. The result is the process as it happened, including the detours, the loops and the steps that were skipped.
Process mining: a worked example
One RA bill, followed end to end through the event log.
- Work executed
- 3 Mar · the first stamp on the case
- Joint measurement
- 5 Mar · 2 days later
- RA bill submitted
- 8 Mar · 3 days later
- Sent back for correction
- 19 Mar · 11 days in client review
- RA bill resubmitted
- 22 Mar · the rework pass
- Certified
- 2 Apr · 11 more days in review
- Throughput time
- 30 days, of which 22 were client review
The flowchart says this process has six steps. The event log says this case took eight, and that the two it added cost 11 days.
Why does process mining matter?
The map on the wall shows six steps. The mined map usually shows sixty paths, and the cash is sitting in the ones nobody drew. You cannot fix a delay you have never seen.
What can and cannot be mined?
| What you have | What process mining can tell you |
|---|---|
| A case id, an activity and a timestamp | The full map: every path, every queue, every loop |
| Those three plus an amount | The same map with cash on each path — process intelligence |
| Those three plus a user or team | Who holds each queue, and where work crosses a boundary |
| Only monthly totals | Nothing. Aggregates have no journey in them |
The minimum is three columns. Everything past that adds a dimension to the same map rather than a new one.
Where does process mining mislead?
Process mining only sees what a system recorded. Work that happens over WhatsApp, on site, or in an inbox leaves no stamp, so it appears in the map as one long unexplained wait rather than as the three approvals it really was. That gap is a finding, not a failure — but it has to be read as one.
What do people get wrong about process mining?
- Mistaking the happy path for the process
- The map's thickest line is the most common path, not the correct one. On most projects the majority of cases take a path nobody designed, and treating the thick line as 'the process' hides exactly that.
- Mining from the wrong timestamp
- Systems often stamp when a record was saved, not when the thing happened. If approvals are keyed in on Friday for a week's work, every wait before Friday disappears and the map understates the delay.
- Stopping at the picture
- A discovered map is a diagnosis, not a decision. Without a cost on each wait, every bottleneck looks equally urgent and the team fixes the visible one instead of the expensive one.
How does Crestline measure process mining?
Crestline reads the event trail directly from your ERP, read-only, and rebuilds the map per project and per process. Nothing is exported, cleaned by hand, or uploaded — the case, activity and timestamp fields are mapped once during setup and the map refreshes on every sync.
How Crestline mines your ERP