Maintenance KPIs and Metrics: Which to Track and How to Use Them

Maintenance KPIs are the small set of numbers that tell you whether your maintenance operation is getting more reliable or quietly falling behind. The set worth tracking is short: MTBF, MTTR, availability, PM compliance, schedule compliance, backlog in weeks, and maintenance cost. This guide explains what each one measures, which lead and which lag, how they get gamed, and how to review them in a weekly cadence that actually changes behaviour.

Why seven numbers beat a dashboard zoo

Most maintenance dashboards fail the same way: they grow. A CMMS makes it cheap to chart anything, so every review adds a metric and nothing ever gets removed. Within a year you have forty tiles, nobody can say who owns which one, and the weekly meeting turns into archaeology instead of decisions. The problem is not that the extra numbers are wrong; it is that attention is the scarce resource, and a dashboard zoo spends it on noise.

A KPI earns its place only if it passes two tests. First, a named person reviews it every week. Second, there is at least one realistic action that person would take if it moved the wrong way. If a number can drift for a month without anyone doing anything differently, it is a curiosity, not a KPI. Keep the curiosities in a report you can pull when needed, and keep the dashboard itself down to a handful of numbers everyone in the room can hold in their head.

The core set: seven KPIs that describe the operation

Seven metrics cover the health of almost any maintenance organisation: two describe reliability, one describes the outcome, three describe discipline, and one describes money. Standardise the definition of each one in writing before you chart anything, because half of all KPI arguments are really definition arguments in disguise.

• MTBF (mean time between failures): total operating time divided by the number of failures. Rising MTBF means assets fail less often. Track it per critical asset, not as a plant-wide blur.
• MTTR (mean time to repair): total repair time divided by the number of repairs. It measures how quickly you restore function, which reflects spares, diagnostics, and access as much as wrench skill.
• Availability: the share of planned production time the equipment was actually able to run. It is the outcome that MTBF and MTTR feed, and the bridge between maintenance and the OEE conversation.
• PM compliance: the percentage of preventive tasks completed within their due window. It is the single best early warning of future breakdowns.
• Schedule compliance: the percentage of planned work-order hours executed in the week they were scheduled. Low values mean the plan is fiction and the crew is living in reactive mode.
• Backlog in weeks: total estimated hours of open, ready-to-schedule work divided by the craft hours available per week. It converts a scary pile of work orders into a plain answer: how far behind are we?
• Maintenance cost: total spend on labour, parts, and contractors, tracked against output or asset replacement value rather than as a raw figure, so growth in production does not masquerade as cost inflation.

Leading versus lagging: know which way each number points

MTBF, MTTR, availability, and cost are lagging indicators: they report what already happened, and you cannot manage last month. PM compliance, schedule compliance, and backlog are leading indicators: they describe the work you are doing right now that determines what the lagging numbers will look like in a quarter. When PM compliance slips, breakdowns follow later, and that lag is exactly what makes the metric feel safe to ignore this week.

In practice, the weekly meeting should spend most of its time on the leading three, because those are the ones you can still change. The lagging four are the scoreboard: check them for trend, and treat a bad trend as a prompt to dig into the leading indicators and the underlying work orders, not as something to fix directly. Chasing a lagging number head-on is how gaming starts.

How KPIs get gamed, and how to design against it

Every metric that carries pressure will be gamed, usually without anyone thinking of it as cheating. MTTR falls when technicians close work orders before the repair is verified, or split one job into three small ones. MTBF rises when minor stoppages simply stop being logged. PM compliance hits a perfect score when checklists are ticked in the office rather than at the machine, a habit old enough to have a name: pencil-whipping. Backlog shrinks when estimates turn optimistic or stale work orders are quietly cancelled instead of completed.

The defence is design, not exhortation. Pair every headline metric with a counter-metric: MTTR with the repeat-failure rate on the same asset, PM compliance with the breakdown count on assets that were supposedly maintained, backlog with the count of cancelled work orders. Audit a small random sample of closed work orders each month. Wherever possible, capture stop and runtime data from the machines themselves rather than from manual logs, because automatically captured data is much harder to polish. And never tie an individual bonus to a single maintenance KPI; that converts gaming from a temptation into a job requirement.

The weekly review: a cadence that changes behaviour

KPIs do nothing on their own; the review is the mechanism. Hold it at the same time every week, keep it to half an hour, and walk the same seven numbers in the same order. Each KPI has one owner who speaks to it. Look at the trend over the last twelve weeks, not the change since yesterday, because week-to-week noise will otherwise generate week-to-week panic. The one hard rule: every number moving the wrong way leaves the room with exactly one action, one name, and one date attached.

Resist the urge to relitigate definitions in the meeting. Standardise them once, write them down, and change them only deliberately, or your trends become meaningless. Once a month, step back and ask whether each KPI still drives action, and prune any that has gone quiet. Above all, prioritise the quality of the inputs, because a beautifully reviewed number built on patchy manual logs is still fiction, and the craft hours utilised chasing a phantom trend are the most expensive hours in the plant.

That input problem is where most plants stall, and it is worth solving properly. The partner we recommend, Fabrico, reads stops directly from the machines and shows the true cause on video, so the numbers on the Monday board reflect what actually happened on the floor rather than what got written down. Fabrico is a partner we recommend; the tools here are free regardless.

FAQ

Related: how to calculate MTBF · how to calculate MTTR · how to calculate availability · planned vs unplanned downtime