Monitoring programmes generate a great deal of countable activity. Only some of it indicates the operation is better.
The measures that matter
Waiting time by stage, as a proportion of elapsed time. The headline.
Cycle time percentiles, median and ninetieth.
Travel per order line.
Exception rate by origin stage, not by detection stage.
Equipment availability against demand, by hour.
Near-miss reporting rate, which should rise, and detected events, which should fall.
Data quality: completeness, exclusion rate, clock agreement.
Findings closed, with the change made and the re-measurement.
Eight measures. Each names a specific failure when it moves the wrong way.
The measures that mislead
Units per hour per person, which attributes system performance to individuals.
Sensors deployed, dashboards built, data points collected.
Aggregate utilisation, which conceals both peak shortage and trough surplus.
Total throughput without the demand curve beside it, so a fall in demand reads as a fall in performance.
Injury rate alone, which is too rare to move usefully and lags by months.
Any single composite score, which cannot be acted on and invites target-setting.
Reading them together
Waiting falling and cycle time falling: the constraint work is succeeding.
Throughput flat with waiting falling: work in progress reduced, which is a real improvement even without more output.
Exception rate falling and near-miss reporting also falling: treat with suspicion. Reporting may have degraded rather than the process improved.
Productivity rising alongside injury, absence or turnover: not an improvement. A cost has moved from the operation to the people.
Data quality declining: every other figure should be read with less confidence, which is why it belongs on the same page.
Reporting to different audiences
Floor: current queue, shift target, what is blocked. Aggregate, never individual.
Supervisors: the constraint now, exceptions needing a decision, equipment availability, staffing against the remaining demand curve.
Facility management: the eight measures, weekly, with normal ranges and open findings.
Above the facility: cycle time, cost per order line, safety measures, and what changed.
One page each. A single dashboard for all four serves none of them.
The savings claim
Realised: the cost actually fell, visible in the ledger. Claim these.
Avoided: a cost that would have been incurred. Claim separately and label.
Identified: a loss quantified but not yet removed. This is an opportunity, not a saving.
Programmes that report the third as the first lose credibility with finance permanently the first time someone checks.
The annual view
What the operation looked like a year ago and now, same measures, same method.
What was found, changed and re-measured, including the changes that did not work.
What remains unaddressed and why, which is the honest section and the one that justifies the next investment.
The one-page weekly report
Eight numbers, a sentence on each, nothing else.
Waiting time by stage as a proportion of elapsed time.
Cycle time median and ninetieth percentile.
Travel per order line.
Exception rate by origin.
Equipment availability against demand at the peak hour.
Near-miss reporting rate and detected events.
Data quality: completeness and exclusion rate.
Open findings with owners and dates.
A sentence per number naming what moved and why. Trends matter more than levels, and a number without a comment is ignored within three months.
Reporting a figure you cannot fully support
Most operational figures are estimates, and how they are presented determines whether the programme stays credible.
State the sources the figure is built from.
State the exclusions and the volume they removed.
Give a range where a range is honest.
Put the caveat in the report, not in the covering email, because the report is what circulates.
Update the figure when data quality improves, and explain that a change reflects better measurement rather than a changed operation.