MTTR y MTBF are the two metrics every maintenance function eventually adopts, and they are the two most often misread. Mean time to repair measures how long it takes to get a machine running again once it has stopped. Mean time between failures measures how long it runs before stopping. They point in opposite directions: a function can improve MTTR by getting very good at breakdowns, which is not the same as having fewer of them. This page sets out what each measures, what each hides, and why reading either alone produces the wrong decision.
MTBF is the one that measures the schedule
Total running time divided by number of failures. If preventive work is doing anything, MTBF rises, and it rises slowly, which is what makes it an honest measure. It is also easily flattered by counting only the failures somebody logged, which is the argument for raising a work order every time.
MTTR measures response, and can improve while things get worse
Total repair time divided by number of repairs. A team that is superb at breakdowns has an excellent MTTR and may be doing no preventive work at all. Falling MTTR beside falling MTBF is a function getting better at firefighting, which is exactly the state preventive maintenance exists to leave.
Read them together, per asset, over quarters
Site-wide monthly figures move on mix and noise. Per asset, per quarter, the pair tells a story: rising MTBF with steady MTTR is a schedule working; falling MTBF anywhere is a machine or an interval to look at now.
Neither is a target to hit
Both are diagnostic. A site managed to an MTTR number will start closing jobs early and recording repairs generously, and the metric will improve while the plant does not. On the worked example this site publishes, the honest measure of the same thing is the reactive share and its trend.
Questions people ask about mttr y mtbf
What is a good MTBF?
There is no cross-site answer; it depends entirely on the asset and its duty. Your own asset's trend across quarters is the only comparison that means anything.
Should MTTR include waiting for parts?
Track both: repair time and total downtime. If they diverge, the constraint is the stores rather than the technicians, and that is a different fix.
Do we need software to calculate these?
No, if work orders record start and finish honestly. The arithmetic is trivial; the honest timestamps are the hard part.