MTTR: what it measures, and what it hides

MTTR is mean time to repair: total repair time divided by the number of repairs. If five breakdowns took 20 hours of repair work in total, MTTR is 4 hours. It measures how quickly you recover from a failure, not how often failures happen.

The formula

MTTR equals total repair time divided by the number of repairs, over the same period.

A worked example

A conveyor failed four times last quarter. The repairs took 2 hours, 3 hours, 1 hour and 10 hours. Total is 16 hours across 4 repairs, so MTTR is 4 hours.

Notice what happened there. Three repairs were quick and one took a day and a half, and the mean sits at a value that describes none of them. Report the median as well, which here is 2.5 hours, and the pair tells the true story: usually quick, occasionally terrible.

Four letters, four different measures

MTTR is used for at least four things and people argue past each other constantly. Say which one you mean.

• •Mean time to repair: the spanner time only
• •Mean time to recovery: from failure to working again, including waiting
• •Mean time to respond: from the alarm to somebody arriving
• •Mean time to resolve: everything, including the follow-up work

The gap between repair time and recovery time is where your improvements live. If the spanner time is 2 hours and the machine was down for 11, the problem is not repair skill. It is parts, access or the time it took anybody to notice.

What a single MTTR number hides

• •Which failures: a mix of easy and catastrophic averages into a meaningless middle
• •Waiting time, unless you deliberately include it
• •Whether the fix held, which repeat failure rate answers and MTTR does not
• •Shift patterns: a failure at 2am on a site with no night cover has a long MTTR for reasons nothing to do with maintenance

How to improve it

MTTR falls through preparation rather than through speed. The critical spares on the shelf, the manual and wiring diagram attached to the asset, isolation points documented, and somebody trained on that machine who is not on holiday.

Track it per asset, not per site

A site-wide MTTR is an average of unrelated things and moves for reasons nobody can act on. Per asset, and per asset class, it becomes a shopping list: these three machines take too long to fix, and here is which of the four causes it is.