Mean time to repair (MTTR)

MTTR is the average time a machine is stopped per failure, from the failure until it runs again. EN 13306 counts the waiting; the hands-on part is mean repair time (MRT).

Waiting for a technician, a part, or a manual counts. Some textbooks use the same letters for the hands-on repair alone, so always say which time you report.


In the standard	Mean time to restoration (MTTR), EN 13306:2017, 11.5. Mean repair time (MRT), 11.4.
German	durchschnittliche Wiederherstellungszeit (MTTR), durchschnittliche Instandsetzungszeit (MRT)
In SAP PM	Mean Time To Repair (MTTR), shown in the app "Ausfälle technischer Objekte"
Also called	Mean time to repair; mean downtime (MDT), Gulati's name for the total including delays

Does MTTR measure restoration or repair?

EN 13306:2017 (Maintenance terminology) times one failure with two clocks.

Time to restoration (9.21) starts when the machine fails and ends when it runs again (restoration, 8.9). Waiting counts. Down time for other reasons, such as preventive maintenance, does not. Its average is MTTR (11.5).
Repair time (9.11) is fault diagnosis, the hands-on work of corrective maintenance, and the function check (8.4). Delays are not in it. Its average is MRT (11.4).

The gap between the two clocks is delay, which the standard sorts into logistic delay (9.13: waiting for a technician to arrive, for spare parts, or for information), technical delay (9.14: setup work around the job) and administrative delay (9.15 internal, 9.16 external). The clock starts at the failure, not at the call, so the minutes before anyone notices count too.

Gulati (Maintenance and Reliability Best Practices, chapter 6) keeps MTTR for hands-on repair and calls the total mean downtime. He observes that a CMMS records the start of the failure and the return to service, so the total gets reported as MTTR.

Example

Nordwerk, a fictional plastics plant, lost injection molding machine IMM-07 on a Tuesday night in January 2025. A hot-runner alarm stopped it at 22:30. Lea Brandt, the only electrician on the night shift, did not know the fault. Nobody was on call.

Time	What happens	Counts as
22:30	The alarm stops IMM-07	Clock starts
22:30 to 06:00	Nobody who knows the fault is at the plant	Waiting, 7.5 h
06:00 to 06:30	Sven Hartmann starts his shift and finds the cause	Repair
06:30 to 07:00	The spare part comes from the storeroom	Waiting, 0.5 h
07:00 to 09:00	Hands-on work	Repair
09:00 to 09:30	Function check; IMM-07 runs again at 09:30	Repair

Time to restoration: 11 h. Repair time: 3 h. Waiting: 8 h, which cost 3,360 € of margin at 420 € an hour.

We recommend reporting MTTR as EN 13306 defines it, showing the waiting inside it, and cutting the waiting before you speed up the hands. The Maintenance Manifesto agrees. "Maximize wrench time" asks to remove what keeps technicians from the machine. "The system is decision-memory" asks the system to hold what the last person learned.

How do you calculate MTTR?

MTTR = total time to restoration / number of failures
MRT  = total repair time / number of failures

Time to restoration, per failure: from the failure until the machine works again. Waiting counts. Stops for preventive maintenance do not. Ramp-up after restoration stays out (Annex C, Figure C.1).
Repair time, per failure: diagnosis, hands-on work, and function check. Waiting does not count.
Failures: those of one machine, or of similar machines, in one period. Write down which stops count as a failure.

This page means restoration unless it says MRT.

What is a good MTTR?

We found no public MTTR value that holds across machine types. EN 13306 is a vocabulary and gives none. Gulati and Wireman give no benchmark either. Gulati's 8 hours is a design requirement in a worked example.

We recommend comparing each machine with its own history, with one clock and one definition of failure. Plants count different stops as failures, so a comparison between plants says little.

When does MTTR mislead?

The clock may differ. Ask which two timestamps your CMMS subtracts. Work started to work done is close to repair time. Failure until the machine can work again is time to restoration. Reporting MRT as MTTR to the plant manager flatters the number: IMM-07's 3.0 h looks better than the 5.0 h the plant lived with. A rising MTTR with a steady MRT means more waiting.
Few failures make a shaky mean. One long failure moves it. Read single failures beside the average.

MTBF counts how often a machine fails, MTTR how long each stop lasts. Together they give availability.

Hub: Maintenance
Terms: Downtime, Corrective maintenance, Preventive maintenance
Pages: Uptime and downtime: the time model, Failure, fault, and breakdown, Why machine knowledge gets lost

Sources

EN 13306:2017, Maintenance: Maintenance terminology, CEN, 2017. Clauses 8.4, 8.9, 9.11, 9.13 to 9.16, 9.21, 11.4, 11.5; Annex C, Figure C.1. This page explains the standard in our own words; the standard is the authority. Buy it as DIN EN 13306:2018-02 from DIN Media.
Ramesh Gulati, Maintenance and Reliability Best Practices, 3rd ed., Industrial Press, 2021, chapter 6 (maintainability, MDT versus MTTR, design requirement of 8 hours).
Terry Wireman, Benchmarking Best Practices in Maintenance, Reliability and Asset Management, 3rd ed., Industrial Press, 2015. Every mention of MTTR checked; none gives a value.
Karl Liebstückel, Instandhaltung mit SAP S/4HANA, 6th ed., SAP PRESS, 2023, chapter 9.1 (app "Ausfälle technischer Objekte").
Maintenance Manifesto, 2025, principles 9 and 11.