The Measures Worth Reporting
A short set that shows the process improving, and the longer set that demonstrates project activity while nothing changes.
Reference
Process mining generates a great deal of countable output. Only some of it indicates the process is better.
The measures that matter
Cycle time median and ninetieth percentile, with the boundary stated.
Waiting as a proportion of cycle time. The headline in most processes.
Rework rate, cases containing at least one loop.
Handovers per case.
Conformance rate on the rules that matter, not a fitness score.
Case volume, so the others can be read against it.
Data quality: exclusion rate, incomplete case proportion, unmapped codes.
Findings closed, with the change and the measured effect.
Eight measures, each naming a specific failure when it moves the wrong way.
The measures that mislead
Variant count, which is a function of activity granularity.
Fitness or conformance percentage, which nobody can act on.
Systems connected, processes onboarded, cases loaded — all measure the project.
Average cycle time alone, which hides the tail that customers experience.
Cases analysed, which measures the extract.
Any composite score, which invites target-setting and cannot be acted on.
Reading them together
Waiting falling and cycle time falling: the constraint work is succeeding.
Cycle time falling and volume falling: possibly nothing improved.
Rework falling and conformance rising with no other change: check whether the mapping changed.
Variant count rising after a refinement: a naming change, not a process change.
Data quality declining: read every other figure with less confidence, which is why it belongs on the same page.
Reporting to different audiences
Process owner: the eight measures with normal ranges, plus open findings.
Operation: the specific finding for their area, validated with them first.
Compliance: the rule breach rates and the specific cases, with case numbers.
Executive: cycle time, the measured effect of what changed, and what remains unaddressed. Three things.
One page each. A single dashboard for all four serves none of them.
The savings claim
Realised: the measure moved and the ledger shows it. Claim these.
Avoided: a cost that would have been incurred. Label separately.
Identified: a loss quantified but not yet removed. This is an opportunity, not a saving.
Programmes reporting the third as the first lose credibility with finance permanently the first time someone checks.
The annual view
What the process looked like a year ago and now, same query, same boundary.
What was found, changed and re-measured, including the changes that produced nothing.
What remains unaddressed and why, which is the honest section and the one that justifies the next phase.
Publishing the normal range
The change that stops ordinary movement generating questions.
Compute the ordinary range for each measure from a long baseline.
Mark it on every chart.
State the rule: inside is ordinary, outside is investigated.
Answer questions about movement within it with "within the normal range", confidently.
Investigate the exceptions and report what was found, which is what makes the rule credible rather than a way of avoiding scrutiny.
The one-page report
Six numbers, a sentence on each, nothing else.
Case duration median and ninetieth percentile.
Waiting as a proportion of elapsed time.
Rework rate.
Variant count and the share covered by the top paths.
Conformance deviations, grouped by pattern.
Data quality: exclusion rate and clock agreement.
Plus open findings with owners and dates. Not the number of processes analysed or dashboards built, which measure the project rather than the operation.