The two definitions

MTBF — mean time between failures — measures how long a repairable machine runs, on average, between one failure and the next:

MTBF = total operating time ÷ number of failures

MTTR — mean time to repair — measures how long it takes, on average, to get a failed machine running again:

MTTR = total repair time ÷ number of repairs

A short worked example. A packaging machine is scheduled to run 600 hours in a month. It fails three times, and the three stoppages last 2, 4 and 6 hours. Operating time is 600 − 12 = 588 hours, so MTBF is 588 ÷ 3 = 196 hours. MTTR is 12 ÷ 3 = 4 hours. A higher MTBF is better; a lower MTTR is better.

The maintenance glossary has the short definitions of these and related terms.

What each number is for

MTBF is a reliability measure: it tells you whether a machine, or a type of machine, fails more or less often over time. It is the number to watch after a design change, a new preventive-maintenance plan or a change of supplier.

MTTR is a maintainability and response measure: it tells you how quickly the team restores production. It moves with things a maintenance team controls directly — how fast someone takes the job, whether the part is on the shelf, whether the drawing is at hand.

Together they describe availability. A machine that fails rarely but takes a day to fix and a machine that fails weekly but is back in ten minutes can lose the same amount of production. You need both numbers to see which problem you have.

Where the numbers go wrong

Every pitfall comes back to the same two timestamps: when the machine stopped and when it was running again.

  • Repair time measured to the end of the paperwork. If a job is only “done” when the report is written, the report-writing time ends up inside MTTR. A repair that took forty minutes looks like three hours because the report was finished after the shift.
  • Stops that never became records. A five-minute stop fixed with a reset is still a failure. If short stops are not recorded, MTBF looks much better than the machine really is.
  • Planned stops counted as failures. Preventive maintenance, changeovers and stops for no orders are not failures. Mixing them in makes MTBF meaningless.
  • Waiting time hidden inside repair time. Two hours waiting for a part and twenty minutes fitting it are different problems. If the record does not show the wait and its reason, MTTR cannot tell you where to act.
  • One average for the whole plant. Averaging a press, a conveyor and a robot cell produces a number that describes none of them.

The data discipline both numbers need

None of this requires sophisticated software. It requires that the shop-floor workflow captures the right facts at the moment they happen:

  1. A breakdown is recorded against the machine when it happens, not reconstructed at the end of the shift.
  2. Someone takes ownership, and that moment is recorded.
  3. Waits are recorded with a reason — waiting for a part, for production, for an electrician.
  4. “Machine running” is recorded when the machine runs, as its own fact.
  5. The report is completed when there is time, without moving the running timestamp.

That last point is the one most systems get wrong. In UR Maintain, reporting the machine running is a fact about the equipment and does not close the job, so a line can be back in service at 03:00 and the report written after the shift. We wrote more about this in why breakdown reporting should not slow the repair.

A note on where UR Maintain is today

In UR Maintain those facts — the stop, the owner, waits with reasons, machine running and the closed report — are recorded on the job, and the manager dashboard builds the downtime figures from them. The point of this article stands regardless of the tool: get the timestamps right first, and the formulas take care of themselves.

MTBF, MTTR and the related figures

Two neighbours are worth knowing because people mix them up:

  • MTTF — mean time to failure — is used for parts that are replaced, not repaired: a bearing, a sensor, a fuse. It is the average life of a non-repairable item.
  • Availability combines the two main figures: availability = MTBF ÷ (MTBF + MTTR). In the example above, 196 ÷ (196 + 4) = 98%. It is one of the three factors in OEE, which is why production managers care about maintenance numbers at all.

Some teams also split MTTR into mean time to respond (until someone takes the job) and mean time to restore (until the machine runs). If your system records the moment of ownership separately, that split comes for free and tells you whether the problem is people availability or repair difficulty.

How much data you need before the numbers mean anything

A single month with two failures produces an MTBF that will swing wildly next month. As a rule of thumb, look at a machine's figures over enough time to include several failures, and compare like with like — the same machine over time, or machines of the same type. Treat the first months after introducing a CMMS as a calibration period: the numbers will move simply because recording got better, not because the machines changed.

Using the numbers well

  • Track MTBF and MTTR per machine or per machine type, and watch the trend rather than a single month.
  • Look at MTTR in parts: time to take the job, time waiting, time repairing. The biggest share is usually where the cheapest improvement is.
  • Pair MTBF with the failure causes on the job history. A falling MTBF with the same cause every time is a different conversation from one with a new cause every time.
  • Do not set targets on numbers the team does not trust. Fix the data first.