MTBF and MTTR: maintenance KPIs with formulas

What MTBF and MTTR measure

MTBF (mean time between failures) is the average operating time between one failure of a repairable asset and the next; MTTR is the average time it takes to get the asset working again after a failure. Together they describe reliability and maintainability, and they give availability. They matter because they turn a pile of work orders into evidence for staffing, spares and replacement decisions.

Both are averages over a period and a population, so the first job is to decide which period, which assets and which failures count. A KPI that changes definition every quarter is worse than none, because it invites decisions based on artefacts of the formula rather than changes in the equipment.

Three mistakes distort these figures more than any others. Counting preventive visits as failures makes MTBF look worse than it is. Averaging over very few failures makes one bad night swing the result: an asset with two failures a year can show an MTBF that halves or doubles on a single event, so report the failure count next to every MTBF. And mixing different asset types in one average hides the problem asset; calculate per asset or per model first, then roll up.

Where maintenance KPIs are standardised

In Europe, the reference is EN 15341:2019+A1:2022, Maintenance Key Performance Indicators. It lists KPIs for the maintenance function and gives guidance on choosing a suitable set to appraise and improve the effectiveness, efficiency and sustainability of maintaining existing physical assets, from industrial plant and infrastructure to facilities, civil buildings and transport systems. Its vocabulary sits on EN 13306:2017, Maintenance terminology.

Both texts are sold by national standards bodies, so this guide paraphrases rather than quotes them. One point of terminology is worth knowing: in IEC and EN dependability vocabulary, MTTR is mean time to restoration, which can be longer than the hands-on repair time many systems measure. Industry usage varies, so every report should say which one it shows.

MTBF: formula and worked example

MTBF = total operating time in the period / number of failures in the period. Count only failures that stopped or degraded the asset and needed corrective work; preventive visits and cosmetic jobs are not failures.

Worked example: chiller CH-02 runs continuously, so the year has 8,760 hours. It failed four times, and the four outages lasted 6, 10, 4 and 12 hours, 32 hours in total. Operating time is 8,760 - 32 = 8,728 hours, so MTBF = 8,728 / 4 = 2,182 hours, about 91 days.

For a fleet of identical devices, pool the time. Forty infusion pumps over 365 days give 40 x 365 = 14,600 device-days; if the fleet logged 10 failures, fleet MTBF is 14,600 / 10 = 1,460 device-days. Comparing that figure across two pump models of similar age is a fair reliability test; comparing two single pumps rarely is.

For items that are replaced rather than repaired, such as lamps or sealed batteries, the usual measure is mean time to failure (MTTF): the average life until the single failure that ends the item's use.

MTTR: repair time or restoration time

MTTR = total time to restore (or to repair) / number of failures. The choice of start and end points changes the answer more than any other decision in maintenance reporting.

Take the chiller's 6-hour outage: the alarm sounded at 02:00, the technician arrived at 03:30, diagnosis and repair ran from 03:30 to 07:00, and the restart and test ran from 07:00 to 08:00. Restoration time is 6 hours, from failure to working again. Active repair time is 3.5 hours. If the four outages had repair times of 3.5, 7, 2.5 and 9 hours, repair-based MTTR is 22 / 4 = 5.5 hours, while restoration-based MTTR is 32 / 4 = 8 hours.

Both are useful. Restoration time is what the users of the asset feel and what availability depends on; repair time shows how efficient the technicians are once they start. The gap between them, here 2.5 hours on average, is waiting: for a technician, for a permit, for parts or for a test, and it is often the cheapest part to reduce.

Availability: putting MTBF and MTTR together

Availability = MTBF / (MTBF + MTTR), using restoration time. For the chiller, availability = 2,182 / (2,182 + 8) = 2,182 / 2,190 = 99.63 percent. The same answer comes from the hours directly: 8,728 operating hours out of 8,760 = 99.63 percent, which is a useful check that the inputs are consistent.

Availability percentages hide scale. At 99.63 percent the chiller was down 32 hours a year; for a theatre air-handling unit that could mean cancelled operating lists, while for an office lift it may barely matter. Set targets by criticality and express them in hours as well as percentages, so a target of 99.5 percent on a 24-hour asset is read as no more than about 44 hours of outage a year (0.5 percent of 8,760 hours = 43.8 hours).

Calendar time or operating time

Many maintenance systems simplify MTBF to calendar time: days in the period divided by the number of corrective work orders. For the chiller that gives 365 / 4 = 91.25 days, close to the 2,182 hours (90.9 days) of the operating-time method, because the chiller runs continuously.

The two methods diverge for assets with part-time duty. An outpatient X-ray room used 2,500 hours a year with 5 failures has an operating-time MTBF of 2,500 / 5 = 500 hours, but a calendar MTBF of 365 / 5 = 73 days. Neither is wrong; they answer different questions. Calendar MTBF suits planning and comparisons between similar assets with similar duty; operating-time MTBF, which needs run-hour meter readings, suits reliability engineering and comparisons across different duty cycles.

Whichever method you choose, apply it to every asset in the comparison and keep it from year to year. A change of method should be recorded with the date it took effect, so a step change in the trend is not mistaken for a change in the equipment.

PPM compliance and the planned-to-reactive ratio

PPM compliance shows whether the preventive plan is delivered. A common industry convention counts preventive work orders completed within their scheduled window as a share of those scheduled in the period; this is not an EN 15341 formula. Simpler systems divide completed work orders by those raised. The two can differ widely: a schedule with 12 work orders, 11 completed but only 8 on time, scores 91.7 percent on completion and 66.7 percent on time. Always label which one a report shows.

The planned-to-reactive ratio shows where the time goes. If technicians logged 3,000 hours in a quarter, 1,950 of them on preventive work and 1,050 on corrective work, planned work is 1,950 / 3,000 = 65 percent. A rising share of planned work alongside a rising MTBF is the strongest sign a maintenance programme is working.

Backlog, cost and downtime

Five further KPIs complete a basic set. Each needs only data a work-order system already holds, provided estimates and costs are entered.

• Backlog in weeks = estimated hours of open work / weekly productive capacity. With 640 open hours and 5 technicians at 32 productive hours each (160 hours a week), backlog is 640 / 160 = 4.0 weeks.
• Maintenance cost as a share of replacement asset value = annual maintenance cost / replacement value. With 180,000 euros spent on assets worth 6,000,000 euros to replace, the ratio is 3.0 percent.
• Downtime percentage = hours out of service / hours in the period. A device out of service for 36 hours in a 720-hour month has 5.0 percent downtime.
• Repeat failures = assets with more than an agreed number of corrective work orders in 90 days, listed by asset and failure code.
• Cost per work order by type = total labour, parts and external cost / number of work orders, split into preventive and corrective.

Data hygiene: what the KPIs depend on

Every KPI above is only as good as the work orders behind it. These rules matter more than the choice of dashboard.

• Record the failure or request time, the start time, the finish time and the return-to-service time separately
• Type every work order correctly: corrective, preventive, calibration, installation or acceptance
• Link every work order to a single asset, never to a room or a free-text description
• Use coded failure, cause and action fields so repeat failures can be counted
• Close work orders when the work is done, not in a batch at month end
• Enter estimated hours on open work so backlog can be calculated
• Capture labour, parts and external cost on each work order
• Read run-hour meters monthly on assets where operating-time MTBF matters

KPI sets for hospitals and for municipalities

A hospital's clinical engineering and estates teams usually report IPM completion (with an on-time view for high-risk devices), equipment location rate, downtime for critical devices, MTBF and repeat failures by model, response time to clinical requests, and maintenance cost by department. The WHO suggests an IPM completion goal above 90 percent, and over 95 percent for the highest-priority devices, as guidance rather than a legal threshold.

A municipality reports on a wider and older asset base: roads, water, buildings and fleets. ISO 55000:2024 defines asset management as the coordinated activity of an organisation to realise value from assets, normally balancing costs, risks, opportunities and performance. In Ontario, O. Reg. 588/17 required every municipal asset management plan to include proposed levels of service, lifecycle activities and a financial strategy by 1 July 2025. A municipal KPI set therefore adds condition grades, backlog in money as well as weeks, planned-to-reactive ratio, and cost per unit of service, alongside availability for critical assets such as pumping stations.

MTBF, MTTR and PPM compliance in Skyline Nexus ERP

Skyline Nexus ERP calculates its maintenance KPIs from CMMS work orders, and states the formulas plainly. MTTR is the average of completion time minus start time, in hours, over completed Corrective work orders, so it is a repair-time measure. MTBF is the number of days in the period divided by the number of completed Corrective work orders on the asset, the calendar method. PPM Compliance is completed work orders divided by all work orders raised from each schedule, a completion ratio rather than an on-time measure.

Each Skyline Nexus work order carries its type, started and completed timestamps, downtime hours, labour, parts and external cost, and coded failure, cause, damage and activity fields, which feed the Failure Pareto, Downtime Analysis, Asset Performance, Maintenance Costs and Department Costs reports, with PDF and Excel export for your own on-time or availability calculations. Start the free month or book a demo and compare the reports against a quarter of your own work orders.