A good MTTR is lower than your own baseline for the same failures, and low enough that repairs fit your loss budget. We found no neutral benchmark that holds across plants, since MTTR depends on the machine, the failure mix and what the clock includes. Set the target from your own repair times: break the MTTR into phases, then remove the waiting.
First, which MTTR do you mean?
The same repair can give two honest numbers. The IEC vocabulary defines MTTR as mean time to restoration, and it marks "mean time to repair" as deprecated.
Restoration runs from the failure until the item works again, so it includes waiting for people and parts. A name like "Mean Time to Repair or Replace", SMRP's metric 3.5.2, can sound like hands-on time only, and that slide deck lists just the name, so write down which clock your plant uses.
For OEE, use restoration, because the line loses output from the moment it stops. The formula is on how to calculate MTTR.
Before you trust any MTTR benchmark
You will find single figures online. Ask three questions before you repeat one.
- Where does the clock start and stop? Hands-on time and restoration can differ by hours.
- Which failures are in it? A plant that counts 5-minute resets will show a far lower MTTR than one that counts only major repairs.
- Which machines? A sensor swap on a conveyor and a gearbox on a press have little in common.
We found no neutral published MTTR range that answers all three. So this page gives you a method and no range.
Why the mean misleads
Example: a line logs 14 failures in a month. The times to restore, in minutes, are 25, 30, 35, 40, 45, 50, 55, 60, 70, 80, 95, 120, 180 and 375.
The total is 1,260 minutes, so MTTR is 1,260 / 14 = 90 minutes. Half of the repairs took under 57.5 minutes, though.
Without the one 375-minute repair, the mean falls to 885 / 13 = 68.1 minutes. Report the mean, because it matches the lost time, but look at the median and the longest repairs as well.
Long repairs often have a different story, such as a part that was not in stock. Short ones point to a different job: diagnosis and fast response.
Set the target phase by phase
A single MTTR target gives nobody a job. Split the same 90 minutes by phase, using timestamps or a two-week study at the line.
Only 18 of the 90 minutes are hands-on repair. A faster wrench cannot fix the other 72, and most of them are waiting or searching.
| Phase | Today | Target | How |
|---|---|---|---|
| Notice and call | 8 | 4 | Clear signal, and a set time before calling |
| Wait for technician | 17 | 10 | On-call rule per shift |
| Diagnose | 20 | 12 | Fault guides at the machine |
| Get parts | 20 | 8 | Common spares near the line |
| Hands-on repair | 18 | 18 | Leave it alone for now |
| Test and restart | 7 | 7 | Do not cut the checks |
| MTTR | 90 | 59 |
A simple check on each phase target: could your best quarter of repairs already hit it? In this log, the fastest four took 40 minutes or less, so 59 is not a fantasy.
Set separate targets for long failures. A planned gearbox swap will never hit 59 minutes, and it should not drag down the target for a sensor fault.
No timestamps yet? Run a two-week study
Tape a sheet to the machine with five columns: stop, call made, technician here, fault found, line running. Ask whoever is there to write the clock time, and accept rough answers.
Twenty failures across a few machines are enough to see which phase is largest. Move the times into a downtime log so the baseline stays with the line.
What the target is worth in OEE
Example: this line has 420 planned hours a month. Failures take 14 x 90 = 21 hours today, and availability from failures is 399 / 420 = 95.0%.
At an MTTR of 59 minutes, failures take 13.8 hours and availability is 96.7%. That wins 7.2 hours a month with the same number of failures.
At 88% performance and 97% quality, OEE moves from 81.1% to 82.6%, or 1.5 points. Put euros on it with the downtime cost calculator.
In a NIST survey of US discrete manufacturers (2016 data), unplanned downtime took 7.8% of planned time overall. That is one sector and it covers more than failures, so it gives scale, not an MTTR.
Check MTBF at the same time
A low MTTR can hide a bad habit. Quick resets give a fine MTTR while the same fault comes back every day.
So read MTTR beside MTBF. A falling MTTR with a falling MTBF means the team is getting good at repeating the repair, so see how to improve MTBF and what is a good MTBF.
To carry out the target, use the MTTR reduction plan. For the first minutes of a stop, see first response to a line stop.
Diagnosis is often the phase you can shrink the most. Where cameras are fitted, Fabrico can stream video of a stop on request next to its machine data, so the technician can review the stop. See how Fabrico uses cameras for OEE.
What people ask next
What is a world class MTTR?
We could not find a neutral figure we would stand behind. A gearbox and a sensor fault share no sensible target, and the clock start differs by plant. Use your own repair times, split by failure type.
Does MTTR include waiting for parts?
Under the IEC meaning of mean time to restoration, yes. It runs from the failure until the item works again. Some plants track hands-on time too, as long as it has a separate label, as this guide shows.
Should MTTR be measured per machine or per plant?
Per machine and per failure type. A plant average mixes sensor swaps with gearbox changes and points nowhere. See the downtime Pareto chart for how to rank them.
How do MTTR and MTBF affect OEE?
Together they set availability from failures: MTBF / (MTBF + MTTR). That is one part of the availability factor in OEE. Test your own numbers in the MTBF and MTTR calculator.
How much can MTTR realistically come down?
Measure the phases first, since the waiting share decides it. If most of your MTTR is hands-on repair, there is less to gain from response work. The MTTR reduction plan shows how to size each fix.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.