Metrics
The maintenance metrics that change what you do on Monday
June 25, 2026 · 14 minute read
A maintenance metric earns its place if you can name the decision it changes. If you cannot, it is decoration, and most maintenance dashboards are mostly decoration. This page covers the twelve that survive that test, what each one measures, and where each one misleads.
Start with three, not twelve
If you are starting from nothing, measure these and nothing else for six months. Everything below is worth having eventually. These three change behaviour immediately.
- Preventive work past its date, by building, today
- The proportion of hours spent on planned rather than unplanned work
- Repeat faults on the same asset within ninety days
Availability
The proportion of time equipment was able to do its job. Availability equals uptime divided by total time. A machine available 8,322 hours out of 8,760 in a year is at 95 percent.
Decide up front whether planned downtime counts against you. Both conventions are used and they give very different numbers, so write down which one you mean before anybody sets a target.
MTBF: mean time between failures
Operating hours divided by the number of failures. A pump that ran 4,000 hours and failed 5 times has an MTBF of 800 hours. It measures how reliable the equipment is.
MTTR: mean time to repair
Total repair time divided by the number of repairs. Five repairs taking 20 hours in total gives an MTTR of 4 hours. It measures how good you are at recovering, not how good the equipment is.
These two answer different questions and improve through different actions. MTBF gets better through design, operation and preventive work. MTTR gets better through spares, access, documentation and staffing.
OEE: overall equipment effectiveness
Availability times performance times quality, for one machine or one line. It is the standard measure of how much of the possible output you actually got, and it is the most commonly misquoted number in manufacturing.
PM compliance
The proportion of scheduled preventive jobs completed within their window. The window is the part that matters: without one, a job done four months late still counts as done.
Planned maintenance percentage
Planned hours as a share of all maintenance hours. Count hours rather than jobs, because one breakdown can consume a week and forty small planned jobs can take an afternoon.
Schedule compliance
The proportion of work scheduled for the week that was actually done in that week. This is the honest test of whether your plan means anything, and it is usually the most uncomfortable number in the set.
Wrench time
The share of a technician’s paid time spent actually working on equipment. Typical measured values across the industry are far lower than anybody expects, and the gap is travel, waiting for parts, waiting for access and looking for information.
Maintenance cost as a percentage of asset replacement value
Annual maintenance spend divided by what it would cost to replace the assets. The most useful long-run number in facilities and plant, because it is comparable across sites of different sizes and hard to argue with.
Backlog in crew-weeks
Ready work in estimated hours, divided by the hours your team has in a week. Between two and four is generally healthy. A backlog counted in jobs cannot tell you anything.
Repeat failure rate
The proportion of faults that come back on the same asset within ninety days. This is the metric that finds work being done to the symptom rather than the cause, and it is the one contractors find least comfortable.
Evidence rate
The proportion of completed work carrying a photograph or a reading. Not a classic maintenance metric, and the best early indicator that the record is real rather than being written from memory on Friday.
Four that are mostly decoration
- Total work orders closed. Rewards closing small jobs and says nothing about the building.
- Average time to close. Dragged around by a handful of jobs waiting on parts. Use the median.
- Cost per square foot in isolation. Meaningless without knowing what condition it bought.
- Technician utilisation. Optimising it produces busy people and worse buildings.
Two rules for the whole set
Prefer the median to the average for anything about duration, because maintenance work has a long tail and one eleven-week wait for a part will drag an average somewhere useless.
And never report a ratio without the two numbers underneath it. "PM compliance 94 percent" invites nothing. "94 percent: 312 of 332, and the 20 that slipped are all in Building 4" starts a useful conversation.