Maintenance KPIs: fourteen measures, defined once and read together
A maintenance KPI is only useful if it is defined once, measured the same way every week and read beside the others. Fourteen measures defined the SMRP way, leading separated from lagging, with the formulas, what a bad number means, how targets get set without inviting editing, and how each one is gamed.
Key takeaways
- A maintenance KPI is only useful if it is defined once, measured the same way every week, and read beside the others. A number on its own is argued about. A set is acted on.
- Leading measures read the process: backlog, planned share, schedule compliance, break-in, PM compliance. Lagging measures read the result: planned maintenance percentage, MTBF, availability, cost. The leading ones move first, which is why they are the ones to manage.
- The SMRP best practice metrics give standard definitions. Adopting them stops the weekly meeting being a debate about what the number means.
- Every KPI has a way to be gamed, and most of them involve editing the schedule or the clock after the fact. The freeze and the fixed definition are what make the numbers honest.
- Fourteen measures is plenty. A site that reports all of them on one page with the trend, every week, knows more than a site with a dashboard of fifty.
Maintenance KPIs are the small set of measures that tell a site whether its maintenance is working: whether work is being identified, planned, scheduled and done, and whether the plant is more reliable and cheaper to run as a result. Every maintenance organisation reports some. Far fewer report measures that are defined once, produced the same way every week and read together, which is the difference between a KPI and a number.
This guide defines fourteen measures the way the SMRP best practice metrics define them, says what each one is for and what a bad number usually means, separates the leading measures from the lagging ones, and covers how to set targets, how the numbers get gamed and how to report them so they stay honest. Where this site has a calculator for a measure, the table links to it.
Leading and lagging, and why the order matters
A maintenance process has a direction. Work is identified, planned, scheduled and executed, and the outcome shows up afterwards as reliability, availability and cost. The measures sit along that line, and the position decides what each one is good for.
The measures near the start are leading: they read the process while there is still time to change the result. Backlog, planned share, schedule compliance, break-in and PM compliance all move within a week of a decision. The measures at the end are lagging: planned maintenance percentage, MTBF, availability and cost confirm, months later, whether the process worked. A site that manages by the lagging measures is steering by the wake.
1Identify
- Backlog, in crew weeks
2Plan
- Planned work share
- Ready backlog share
3Schedule
- Schedule compliance
- Break-in share
4Execute
- PM compliance
- Wrench time
5Outcome
- Planned maintenance percentage
- MTBF
- MTTR
- Availability
- OEE
- Cost of downtime
- Maintenance cost per unit
→ Leading · read the process while it can still change
Lagging · the result
Manage the leading measures weekly. Read the lagging ones monthly to check they are following.
The fourteen measures
The definitions below follow the SMRP best practice metrics where one exists, and the formulas are the ones the calculators on this site use. The range column is a direction drawn from common practice, not a target: a site's own trend is the figure that matters, and the section on targets explains why.
| Measure | Kind | Formula | What it tells you | Common range |
|---|---|---|---|---|
| Backlog, in crew weeks | Leading, identify | Open work hours by trade ÷ weekly capacity by trade | Whether work is being identified and whether the crew can ever catch up | Commonly 2 to 6 weeks, with the ready share reported separately |
| Planned work share | Leading, plan | Hours on work with a job pack ÷ all hours worked | Whether the planner is being allowed to plan | Commonly 80 per cent and above |
| Ready backlog share | Leading, plan | Hours on ready-to-schedule work ÷ all backlog hours | Whether next week's schedule can be filled without compromise | Enough to fill two to three weeks of schedule |
| Schedule compliance | Leading, schedule | Scheduled hours done as scheduled ÷ all scheduled hours, against the frozen schedule | Whether the plan was realistic and whether it was protected | Commonly 85 to 90 per cent |
| Break-in share | Leading, schedule | Hours added after the freeze ÷ all hours worked | Reliability, planning or discipline, and the reasons say which | Commonly 10 per cent or less |
| PM compliance | Leading, execute | Preventive work orders done within their due window ÷ all due | Whether the program is being traded for corrective work | Commonly 90 per cent and above, with critical equipment at 100 |
| Wrench time | Leading, execute | Hours on the job ÷ hours paid, from an occasional sampling study | How much of a scheduled hour reaches the equipment | Commonly 25 to 35 per cent, rarely above 50 |
| Planned maintenance percentage | Lagging, outcome | Hours on planned and preventive work ÷ all maintenance hours | Whether the site is maintaining the plant or chasing it | Commonly 80 per cent and above |
| MTBF | Lagging, outcome | Operating time ÷ number of failures, by equipment and failure mode | Whether reliability is improving where the strategy changed | Against the asset's own history |
| MTTR | Lagging, outcome | Total repair time ÷ number of repairs | How fast failures are restored, and where the delay sits | Against the asset's own history |
| Availability | Lagging, outcome | Run time ÷ planned production time, measured on the constrained circuit | What the plant could run, after maintenance and breakdown time | Set by the business plan |
| OEE | Lagging, outcome | Availability × performance × quality | Which of the three losses is costing output | Commonly 60 to 85 per cent |
| Cost of downtime | Lagging, outcome | Downtime hours × lost throughput × margin, plus standby labour | What a failure costs in dollars, the language of a business case | Per event and per year |
| Maintenance cost per unit | Lagging, outcome | Maintenance cost ÷ units produced, split by labour, parts and contractors | Whether the money is going where the strategy said | Against the site's own trend |
The five to manage weekly
If a site reports nothing else, it should report these five, on one page, every Monday. Together they describe the planning and scheduling cycle well enough to run it, and each one points at a different fix when it is wrong.
- Backlog in crew weeks, by trade, with the ready share. Over four to six weeks is under-resourced or a reliability problem; under two means work is not being identified.
- Planned work share. Low means the planner is being pulled into the day, or readiness is not being gated.
- Schedule compliance, against the frozen schedule. Low means unready work was scheduled, capacity was overstated, or break-in displaced the plan. Schedule compliance and break-in are read as a pair.
- Break-in share, with the reasons. The share has three causes with three fixes: reliability, planning or discipline.
- PM compliance, with critical equipment separated. Low means the program is being traded for corrective work, and the backlog will show it next.
The outcome measures, and what they are for
Lagging measures are where maintenance meets the business. Planned maintenance percentage is the one most sites quote and the one most easily misread, because it rises when corrective work falls, not when work orders are reclassified. MTBF and MTTR are only useful at the level of a piece of equipment and a failure mode, where a strategy change can be seen to work; a site-wide MTBF averages bearings with belts and means nothing. The Weibull calculator goes further and says whether the failures are random or age related.
Availability belongs on the constrained circuit, measured from control system data rather than shift logs, and OEE splits the lost output into the three losses so the right team is chasing it. Cost of downtime turns a failure into dollars, which is the language a business case is written in, and maintenance cost per unit shows whether the money is going where the strategy said.
Setting targets without inviting editing
Benchmarks such as 85 per cent schedule compliance or 80 per cent planned work are useful as a direction and dangerous as a target. A site that sets them before the measures are trusted gets the number by editing the schedule after the week, reclassifying work orders, or moving the clock, and the measure stops describing anything.
The sequence that works is to define each measure in writing, report it honestly for a quarter without a target, set a target the next quarter can reach from the trend, and move the target as the trend moves. The definition includes the things that get argued about: whether MTTR includes waiting for parts, whether a break-in approved before the freeze counts, which clock operating hours come from. Written once, they stop being argued about.
- Report the trend, not the week. A quarter of honest trend is worth more than any single number.
- Pair the measures that keep each other honest: schedule compliance with break-in share, PM compliance with backlog, availability with cost.
- Separate critical equipment. A PM compliance of 92 per cent is not reassuring if the eight per cent is the mill.
- Keep the definitions next to the numbers on the report, so a new superintendent reads the same measure the last one did.
A target set before the measure is trusted produces editing, not improvement.
How the numbers get gamed
Every measure has a way to be improved without improving anything, and knowing them is the best defence. Most involve changing the record after the fact rather than the work, and most are prevented by the same two disciplines: a fixed definition and a freeze.
| Measure | How it is gamed | What stops it |
|---|---|---|
| Schedule compliance | Editing the schedule during the week to match what was done | A frozen schedule issued on a fixed day, with changes logged as break-in |
| Break-in share | Approving break-ins before the freeze so they count as scheduled | The freeze date fixed, and additions flagged in the CMMS by date |
| PM compliance | Closing preventive work orders without doing them, or widening the due window | Findings and as-found condition required on closure, and a fixed window |
| Planned maintenance percentage | Reclassifying corrective work as planned once a job pack is attached | Planned means planned before the failure, defined in writing |
| MTBF | Counting fewer events as failures, or dividing by calendar hours | Functional failure defined per equipment class, and operating hours from the historian |
| Backlog | Cancelling old notifications wholesale to shorten the list | Backlog review with a reason recorded for every cancellation |
| Availability | Averaging across every item so the bottleneck disappears | Measured on the constrained circuit, from control system data |
Reporting them so they stay honest
The measures should be queries, not spreadsheets. Each one is a defined query against the CMMS and the historian, with the definition written beside it, produced by the system on a schedule rather than by a person on a Friday afternoon. That is what makes the number the same whoever produces it, and what keeps it being produced when that person is on leave.
One page, weekly, with the trend, read in one meeting by the people who can act on it. The leading five at the top, the outcome measures below, critical equipment separated, and the reasons for every miss on schedule compliance and break-in listed underneath. A site that does this consistently for a quarter has a maintenance organisation it can describe in numbers, which is rarer than it should be. Automated operations reporting describes what that looks like when it is built properly.
Common mistakes
- Fifty measures on a dashboard, none of them defined, none of them trended.
- A site-wide MTBF that averages failure modes that behave nothing alike.
- Schedule compliance quoted without break-in share, so a small plan scores well.
- Targets set from a benchmark before the measure has been reported honestly for a quarter.
- Definitions held in a planner's head and argued about every week.
- Availability averaged across the plant, hiding the circuit that limits the tonnes.
- Measures assembled by hand in a spreadsheet that stops when its owner is away.
Putting the measures in place
Define
Each measure written down, with the arguments settled: clocks, windows, what counts as a failure.
Build the queries
From the CMMS and the historian, so the number is the same whoever runs it.
Report without targets
One page, weekly, with the trend, for a quarter.
Set reachable targets
From the trend, and move them as it moves.
Pair and separate
Compliance with break-in, PM with backlog, and critical equipment on its own.
Act in the meeting
Reasons for every miss, and a decision for each reason.
Define
Each measure written down, with the arguments settled: clocks, windows, what counts as a failure.
Build the queries
From the CMMS and the historian, so the number is the same whoever runs it.
Report without targets
One page, weekly, with the trend, for a quarter.
Set reachable targets
From the trend, and move them as it moves.
Pair and separate
Compliance with break-in, PM with backlog, and critical equipment on its own.
Act in the meeting
Reasons for every miss, and a decision for each reason.
Common questions
What are the most important maintenance KPIs?
Five leading measures of the process (backlog in crew weeks, planned work share, schedule compliance, break-in share and PM compliance) and a few lagging measures of the result (planned maintenance percentage, MTBF and MTTR on critical equipment, availability of the constrained circuit, and maintenance cost). Read together, weekly, with the trend, they describe a maintenance organisation well enough to run it.
What is the difference between a leading and a lagging maintenance KPI?
A leading measure reads the process while there is still time to change the outcome: whether work is planned, scheduled and done as scheduled. A lagging measure reads the outcome after the fact: failures, availability, cost. Leading measures move first and are the ones a maintenance manager can act on this week. Lagging measures confirm, months later, whether it worked.
What is planned maintenance percentage?
The share of all maintenance hours spent on planned and preventive work rather than reactive repair. Eighty per cent and above is the commonly quoted range for a well run site. It is a lagging measure, because a site reaches it by holding schedule compliance and PM compliance up for months, not by reclassifying work orders.
What is a good schedule compliance?
Commonly quoted targets sit around 85 to 90 per cent, measured against a schedule frozen before the week began. A site introducing a frozen schedule usually starts well below that, and the trend over a quarter matters more than any target. Compliance only means something if break-in share is read beside it.
What is wrench time?
The share of a tradesperson's paid hours spent with tools on the equipment, as opposed to waiting, travelling, finding parts and permits, or in meetings. It is measured by an occasional sampling study rather than weekly. Typical figures are 25 to 35 per cent, and the gap is mostly planning and logistics, which is why the planned work share is the measure to fix first.
How do you set maintenance KPI targets?
From the site's own trend rather than from a benchmark. Industry figures such as 85 per cent schedule compliance are a direction, not a target to hit next month. Define the measure, report it honestly for a quarter, then set a target the next quarter can reach, and move it as the trend moves. A target set from a benchmark before the measure is trusted produces editing, not improvement.
Where should maintenance KPIs be reported?
On one page, weekly, with the trend, from the CMMS and the historian rather than from a spreadsheet beside them. The measures should be queries anyone can run, with the definitions written next to them, so the number is the same whoever produces it. A dashboard that needs a person to assemble it each week will drift as soon as that person is on leave.
Key terms
Plain-language definitions from our glossary for the concepts this article leans on.
Standards and further reading
- Best practices, metrics and guidelines for maintenance and reliability (SMRP)
- ISO 22400-2:2014 Key performance indicators for manufacturing operations management (ISO)
- ISO 14224:2016 Collection and exchange of reliability and maintenance data for equipment (ISO)
- ISO 55001:2024 Asset management system requirements (ISO)
- Operations and maintenance best practices guide (US DOE Federal Energy Management Program)
- The Asset Management Landscape, third edition (GFMAM)
Related case studies and tools
- Automated operations reporting (case study)
- Work management and scheduling uplift (case study)
- Schedule Compliance Calculator (tool)
- MTBF, MTTR and Availability Calculator (tool)
- OEE Calculator (tool)
- Weibull Reliability Calculator (tool)
Related reading
Working through something like this?
See how we approach these initiatives, or tell us what you are dealing with.