PM compliance is the share of preventive and predictive maintenance work orders due in a period that were completed inside their allowed tolerance window. It's counted in work orders and measured monthly. It answers one question: did your preventive calendar get done on time?
Whether that calendar was worth doing is a separate question, and this metric never answers it.
2 numbers dominate almost everything written about PM compliance: 90%, and the 10% rule. They come from different places and don't agree with each other, yet they're usually printed in the same paragraph as if they were one standard.
Below are the formula, the published versions of "on time" and the one I recommend, and what a good number looks like. Then how your CMMS can move the goalposts, and what to read alongside the number.
My position on the target is more demanding than the industry's, and my position on what the number proves is less generous. Both are deliberate.
How do you calculate PM compliance?
The formula is simple:
PM compliance = (PM and PdM work orders completed on time ÷ PM and PdM work orders due in the period) × 100
SMRP publishes it as metric 5.4.14, with a sample calculation of 416 completed on time out of 476 due, giving 87.4%.
2 words in that formula matter most.
The first is due. Most formulas you'll find divide by the PMs you scheduled, so you can improve the number by scheduling fewer of them. Defer a job before its due date and it drops out of the calculation.
The second is on time. Some formulas don't check a single date. That measures PM completion, a useful number with a different name.
How long is "on time"? The due window
How wide should the window be? There are at least 5 published answers, and none of them is a standard.
| Rule | Window | Source | On a 30-day PM |
|---|---|---|---|
| SMRP 5.4.14, option 1 | Completed by the required date | SMRP Best Practices | No grace |
| SMRP 5.4.14, option 2 | Required date + 1 day | SMRP Best Practices | 1 day late |
| SMRP 5.4.14, option 3 | Required date + 20% of the frequency, capped at 28 days | SMRP Best Practices | 6 days late |
| The 10% rule | 10% of the frequency (most apply it either side) | Stangier & Smith, 2016 | 3 days either side |
| One-step-down rule | Weekly PMs by the day, monthly by the week, quarterly within 2 weeks, annual within a month | Thorman, Plant Services, 2019 | 7 days |
| My modified 10% rule | ±10% of the frequency, capped at 30 days | Reliability Academy | 3 days either side |
Pick one rule and stick to it. SMRP says the same about its 3 options: choose one and apply it consistently.
Why the 10% rule breaks on long-interval PMs
Stangier and Smith don't say which side of the due date the 10% sits. Some people read it as 5% either side. Most of industry applies 10% either side, and that's the version I use.
A yearly PM gets 10% of 365 days, so about 37 days either side of the due date. A 2-yearly PM gets 73 days. A 4-yearly PM gets 146 days. An 8-yearly PM gets 292 days.
Remember that's plus or minus. 292 days each way is a window of 584 days, or 1.6 years in which the job still counts as on time.
In large industrial plants, PMs on those frequencies are normal: major overhauls, statutory inspections, turbine and compressor work. So the 10% rule falls apart on exactly the PMs where the consequences are largest.
The fix: cap the tolerance at 30 days
I've always used a modified version of the 10% rule, with the tolerance capped at 30 days.
That gives you a 2-month window at most on a long-cycle job, which I think is reasonable. And the rule stays simple enough for everyone to remember.
One more thing, and I feel strongly about it. Don't adopt different tolerances for different types of PMs.
I've seen companies run 3 or 4 different formulas depending on the criticality of the job. Nobody remembers which rule applies where, and the arguments start at the monthly review.
Doing a PM early breaks compliance too
Most people treat the window as one-sided: late is a miss, early is a bonus.
Doing a PM too early is just as bad as doing it too late.
Pull a job forward, then do the next one on its original due date, and the next gap runs past your target frequency by however early you went, up to the full tolerance. You've stretched the interval without deciding to.
The reliability side of the argument comes from the nuclear sector. In EPRI report 1009909 (Appendix A), David Worledge describes plants that schedule tasks 20% earlier than the technically optimal interval, so that a grace period still lands inside it. His finding:
"Intrusive PM tasks performed before their technically optimal task intervals are likely to increase the failure rate because maintenance error and material defects are introduced more often."
EPRI's recommendation is to do intrusive tasks as close as possible to the end of the grace period.
EPRI also cautions against reading this as a proven direct link to failures. So treat early work as a risk to manage.
Do your PMs inside the tolerance you set.
The 2 CMMS fields that make this calculable
Alongside the due date in your CMMS, get the system to calculate 2 more dates from that due date and your tolerance rule.
Earliest allowable start date (EASD). The earliest date the job can start and still count as compliant. Due date minus 10% of the frequency, capped at 30 days.
Latest allowable finish date (LAFD). The latest date by which it must be complete to count as compliant. Due date plus 10% of the frequency, capped at 30 days.
Those 2 fields make PM compliance easy to calculate, and they end the argument about what counts.
Your monthly denominator becomes every job with an LAFD falling in that month. That's almost the same population as "PMs due in the period", with the tolerance already built in.
Your schedule should still aim for the due date.
The tolerance is there to give you operational flexibility. Sites that schedule everything to land just before the LAFD put huge pressure on the crew and set themselves up to fail.
Check your regulatory position before you set the rule. In most jurisdictions, a 2-yearly or 4-yearly PM done up to 30 days late wouldn't normally be a major issue, which is part of why I like the 30-day cap.
Where the 10% rule came from
The best-known source for the 10% rule is Doug Stangier and Ricky Smith's Preventive Maintenance Made Simple (Reliabilityweb, 2016):
"A time-based PM must be accomplished in ten percent of the time frequency or it is out of compliance."
It's a practitioner rule of thumb.
Terry Wireman's 1998 PM compliance indicator carries no tolerance window at all, with a stated goal of 100% completion.
There's no international standard setting this tolerance. You can choose what's best for your company, and you should choose deliberately.
SMRP allows twice as much lateness
SMRP is the body most often credited with the 90% target. SMRP's most generous on-time option is the required date plus 20% of the frequency, capped at 28 days.
On a monthly PM, SMRP allows 6 days late. The 10% rule allows 3 days late and 3 days early. Same width, different position.
PM compliance vs schedule compliance
Schedule compliance tells you how much of the frozen week got done. PM compliance tells you whether your preventive calendar was done on time. They use different denominators and run on different cadences.
| Schedule compliance | PM compliance | |
|---|---|---|
| Question it answers | How much of the frozen week got done? | Was the preventive calendar done inside its tolerance? |
| Denominator | The jobs in the frozen weekly schedule | PM and PdM work orders due in the period |
| Unit | Work orders or planned hours | Work orders |
| Cadence | Weekly | Monthly |
| Timing test | Done inside the week. The day doesn't matter | Done inside the tolerance window around the due date |
| Is early a problem? | No | Yes. Early is as bad as late |
| What breaks it | Emergency work, under-loading, poor estimates | Resource shortages, poor-quality PMs, floating due dates |
| Is 100% good? | No. Sustained 100% means the week was never full | Close to it is the target, provided the PM list is worth doing |
Measure PM compliance monthly, using the latest allowable finish date. If you set it up weekly, it starts to clash with schedule compliance, which is the weekly metric you need for maintenance productivity. Review progress weekly in your schedule review meeting, and report the metric monthly.
What is a good PM compliance percentage?
SMRP 5.4.14 says this about 90%:
"Above 90%: Investigate non-compliance reasons, implement improvements, monitor results and look for improvement trend."
SMRP treats 90% as a point to keep investigating and improving from.
I'd expect PM compliance over 95%, and ideally over 98%, consistently, month in month out. High PM compliance needs to be non-negotiable in your organisation.
If you don't do your PMs on time, you can't assess whether they work.
When a task fails to prevent a failure, you need to answer one question: was the task wrong, or was it never done? If your PMs aren't done on time, you can't tell those 2 apart, and every question about PM effectiveness becomes unanswerable.
Your target is to make PM compliance a non-metric. You want to reach the point where discussing it holds little interest, because you always do your PMs on time. Then the review meeting can move on to the questions that matter.
What a high score doesn't prove
Typically 40% to 60% of the tasks in a PM program add little or no value. John Moubray put that range in RCM II, and it's the figure I teach.
A plant at 98% PM compliance, on a list where half the tasks address no credible failure mode, has done a lot of useless work on time.
Steve Turner, who built PMO2000, put it bluntly back in 2010:
"Achieving 100% compliance with a program that is 50% useful and 50% wasteful can not be good asset management!"
The light fitting
A light fitting falls and nearly hits someone. It was never installed properly, so it's a construction defect. Because it's a safety issue, a manager insists on a quarterly PM to check every light fitting on site.
Now you've got dozens, if not hundreds, of PM hours a year that are largely a waste of time. Every PM needs a credible failure mode, and this one doesn't have one.
That task will run at 100% compliance forever. On your dashboard it looks exactly like the vibration route on your critical pumps.
More PM can mean more failures
The mechanism goes back to the Second World War. C. H. Waddington ran operational research for RAF Coastal Command, studying the aircraft hunting U-boats in the Atlantic. He wrote:
"inspection tends to increase breakdowns, and this can only be because it is doing positive harm by disturbing a relatively satisfactory state of affairs."
James Ignizio named it the Waddington effect in 2010: a spike in unscheduled events right after a scheduled maintenance event, then a gradual decline. Once Waddington's recommendations went in, by Ignizio's account the effective size of the fleet rose by more than 60%.
Plot your own unscheduled work against time since the last PM. If you see a similar spike, it could be down to ineffective PMs. Or the right PMs aren't being done correctly, which points straight at your work instructions.
No study links this metric to reliability
There's no published study showing that higher PM compliance gives you higher reliability, availability or MTBF.
If someone quotes you a figure linking the 2, ask them where the study is.
2 plants, the same 96%
The figures below are illustrative, built to show the mechanism.
Both plants had 200 PM and PdM work orders due in June. Both reported 96% PM compliance.
| Plant A | Plant B | |
|---|---|---|
| PMs due in June | 200 | 200 |
| Never generated, because the previous call was still open | 0 | 22 |
| Skipped in the scheduling transaction | 0 | 6 |
| Denominator reported | 200 | 172 |
| Completed on time | 192 | 165 |
| Reported PM compliance | 96% | 96% |
| Tolerance rule | ±10% capped at 30 days, as EASD and LAFD in the CMMS | "The 10% rule", one-sided, applied by whoever writes the report |
| Due dates | Fixed | Floating |
| Task list | Reviewed in the last 24 months | Last reviewed at commissioning in 2009, plus 60 tasks added after incidents |
| Condition-based share of PM hours | 61% | 18% |
| Follow-on work raised from PMs | 1 corrective per 7 PMs | 1 corrective per 41 PMs |
| Emergency work | 6% of hours | 24% of hours |
Measured against the 200 originally due, Plant B's real compliance is 165 ÷ 200, or 82.5%. Apply the 40% to 60% figure to its unreviewed list and the useful work sits somewhere between 66 and 99 tasks. Plant A is doing 2 to 3 times the useful work on the same headline number. And 1 follow-on corrective per 41 PMs tells you Plant B is inspecting and finding nothing, which usually means over-maintaining.
How the number gets gamed, including by your CMMS
People game it, and so do some CMMS settings.
| Who | The move | What it does to the number |
|---|---|---|
| People | Delete or cancel overdue PMs before month end | Removes the misses from the denominator |
| People | Close the work order without doing the work | Tick and flick. Perfect score, no maintenance |
| People | Convert a missed PM into a corrective work order | The PM disappears, the failure risk doesn't |
| People | Widen the tolerance when the number dips | Last month and this month are no longer comparable |
| CMMS | Floating due dates re-base the next due date after a late completion | A late PM is never late twice |
| CMMS | The next work order isn't generated while the last one is open | The worst assets leave the denominator |
| CMMS | Reports that divide by completed work orders | Finish 90 of 100 due, leave 10 untouched, and you score 100% |
Fixed or floating due dates
Fixed means a 30-day PM falls due every 30 days from the start date, whether the last one was done early, late or on time. Floating means the next due date is calculated from the last completion date.
My strong recommendation is to use fixed due dates as much as possible.
In the real world, PMs often get done a couple of days early or late. With floating dates on everything, your due dates end up all over the shop within a year or 2.
Fixed dates create stability. They build a culture of standard work that just needs to get done on time, and that's what you need from a PM program.
There's a legitimate exception. Where usage is uneven, run hours or cycles can matter far more than calendar days. Major compressors and turbines are the obvious case. Accept that the due date moves, and go in with your eyes open.
Under floating due dates, a late PM moves its own next due date. A late PM is never late twice, and a backlog never builds up. Your score goes up while your interval stretches.
Most CMMSs let the next due date float from the last completion. In SAP it's called the shift factor. Take a task planned for 1 August and confirmed on 16 August. At a 0% shift factor the next date stays 1 September. At 100% it moves to 16 September. Maximo offers the same choice between fixed and floating schedules.
A second setting removes PMs from the count altogether. Many systems don't generate the next work order while the previous one is still open. So one missed PM stops the next work order being generated, and it never reaches the denominator. Your worst-performing assets drop out of the measurement.
In my experience, you'll build the PM compliance metric yourself, whether in your CMMS or in something like Power BI. Whoever built it made choices, and those choices are baked into the number. So the first question to ask about anybody's PM compliance figure, including your own, is how it was calculated. It's also why comparing PM compliance across plants tells you little.
What to do when the number is low
Ask why, and expect more than one root cause. There are almost always several factors at play. Get down to the systemic issues, and don't stop at blaming individuals for not doing PMs.
4 causes cover most of what I see:
- Productivity is low and you don't have the resources to do all the PMs on time. That's a resourcing problem or a planning and scheduling problem, and talking about compliance won't fix it.
- Some PMs are poor quality. The team doesn't do them because they know they add no value, can't see the point, or don't understand them.
- The team is in reactive mode, too busy with breakdowns to protect the preventive calendar.
- Nobody looks at the metric regularly, or people don't understand why it matters.
Start with cause 2. A PM that keeps getting skipped is telling you something about the task. Check the task before you blame the people.
2 more things.
Don't hide behind a split by criticality. It's tempting to say overall compliance is low but critical PMs are fine. But do you want a culture where late PMs become normal?
And don't report just the monthly number. Publish a trend of at least the last 12 months.
What to read alongside it
PM compliance on its own tells you about execution. Read it with 3 other measures, in this order.
PM/CM ratio. The PM hours you spend in a month divided by the CM hours. Use hours, because work order counts are very misleading. Break it down by equipment type. General area lighting should show a very low PM/CM ratio, emergency lighting a high one. If the numbers don't match your strategy, find out why. And don't compare your ratio with other companies unless you're sure they capture the data the same way. Usually they don't.
PM-related bad actor reports. Annualise PM cost and hours before you rank anything. A monthly task gets done 12 times a year and a 2-yearly task once every 2 years, so raw totals will mislead you. A system at the top of the list may be fine, if it's high effort and gives you the reliability you need.
Follow-on work. How many of your PMs lead to new work. It tells you whether your PMs find anything. Lots of inspections and no findings usually means over-maintaining. Treat it as a pointer for further investigation.
Start with PM compliance. Add the PM/CM ratio and your bad actor reports. Only add follow-on work once you have good discipline around planning, scheduling and work management.
Make PM compliance a non-metric
Get the denominator right. Pick a tolerance rule and put it in the CMMS as computed dates. Then hold the bar at 95% and push for 98%.
Below that bar you can't tell a bad PM from a PM that never happened.
So get it boring, and then ask the better question: are these PMs worth doing at all?
Our article on whether your preventive maintenance is even worth doing is where to start, and how to improve your preventive maintenance program covers what comes next.



