Choosing the three maintenance KPIs worth reporting upward
A dashboard with twenty numbers on it is a dashboard nobody reads. Pick three that change behaviour.
Maintenance functions under-report or over-report, rarely in between. Both fail for the same reason: the numbers chosen describe activity rather than consequence, so the audience cannot act on them.
Test every candidate metric against one question
If this number moves, what decision changes? A metric that fails that test is operationally interesting and does not belong in an executive pack. Work orders closed per week is a good example. It moves for reasons nobody upstairs can act on.
Three that survive the test
- Unplanned downtime hours on critical assets. The consequence everyone already cares about, expressed in the unit the business feels.
- Cost per critical asset, twelve-month rolling. Turns repair-versus-replace into a decision instead of an argument.
- Percentage of work orders completed within the agreed SLA. A single figure covering approval speed, contractor performance and scheduling realism.
Report the trend and one sentence of cause
A figure with a direction and an explanation is a briefing. A figure alone is a quiz. If downtime rose, say what drove it, and if the driver was an approval queue rather than an engineering failure, that is precisely the sentence worth writing.
Keep the operational set separate
PM compliance, backlog age, mean time to repair, rework rate and technician utilisation all matter, to you. Run them weekly inside the function and leave them out of the board pack. Two audiences, two sets, one source of data.
Publish the definition
Most disputes about maintenance KPIs are disputes about definitions. Write down what counts as downtime and when the clock starts, once, and attach it to the report. The argument then happens once instead of every quarter.
Related reading
Make good practice the default, not the effort
SaharaDesk turns these habits into workflow: thresholds that apply themselves, registers that stay honest, records that survive turnover.