MTTR and MTBF are the two numbers that describe reliability. MTTR, mean time to repair, measures how fast you recover after a failure. MTBF, mean time between failures, measures how long a system runs before it fails. One without the other tells half the story.
Teams often track only one of them, usually MTTR, because it is the number that shows up in incident reviews. But a team with a great MTTR and a terrible MTBF is still a team whose systems keep breaking. Both numbers belong on the same dashboard.
MTTR: how fast you recover
MTTR is the average time from when an incident starts, or is detected, until the service is restored. It is calculated by dividing total repair time by the number of incidents. A team that fixed five incidents in 30, 45, 60, 90, and 120 minutes has an MTTR of 69 minutes.
MTTR has three phases, and each one is a different problem. Detection is the time until someone notices. Response is the time until work starts. Resolution is the time until the service is back. A slow alert, a slow escalation, and a slow fix all add to the same number, but they need different solutions.
Improving MTTR is usually about process: better monitoring, clearer on-call rotation, runbooks, and practice. It is the faster half of reliability to improve, because it depends mostly on the team rather than on the hardware.
MTBF: how often things break
MTBF is the average operating time between one failure and the next. It is calculated by dividing total operating time by the number of failures. A server that ran 8,760 hours in a year and failed 3 times has an MTBF of 2,920 hours, or about 122 days.
MTBF is about the system, not the team. It improves when hardware is replaced, software is patched, or a fragile component is redesigned. It falls when a system ages, when load grows past its design, or when a bad change introduces a new failure mode.
Unlike MTTR, MTBF cannot be improved by working harder during an incident. It can only be improved by preventing the next one, which is slower, more expensive work — and the work that actually reduces the number of 2 a.m. calls.
How they combine into availability
Availability is approximately MTBF divided by the sum of MTBF and MTTR. A system with an MTBF of 2,920 hours and an MTTR of 4 hours is available about 99.86% of the time.
That formula shows why both numbers matter. Cutting MTTR from 4 hours to 2 raises availability to 99.93%. Doubling MTBF to 5,840 hours raises it to 99.93% as well. Either improvement gets you to the same place, so the right choice is whichever costs less.
It also shows the limits. If MTBF is only 100 hours, even an MTTR of 15 minutes leaves availability at 99.75%. No amount of fast repair can fix a system that breaks constantly. At that point the only real answer is to improve MTBF.
Note: The availability formula assumes failures are evenly spread and repair times are typical. Real systems have bad weeks and long-tail incidents, so treat the result as a planning estimate, not a guarantee.
Which one to improve first
Start with the cheaper win. If detection is slow because alerts go to an inbox nobody watches, fixing that costs almost nothing and cuts MTTR immediately. If a server fails monthly because its disk is dying, replacing the disk costs a few hundred dollars and doubles MTBF.
When both need work, fix MTBF first for systems that fail often, and MTTR first for systems that fail rarely but stay down a long time. A system that fails once a year and takes a day to fix needs a better recovery plan. A system that fails weekly and takes ten minutes to fix needs a better system.
Track both numbers by system, not just for the whole estate. A fleet average hides the one server that fails every week, and it hides the one service that takes a day to restore. The averages are for reporting; the per-system numbers are for fixing.
Try the tool
The fastest way to see where you stand is to calculate both numbers from your own incident history. The MTTR calculator turns your incident durations into an average repair time with a phase breakdown.
Run the MTBF calculator next: it turns operating hours and failure counts into the average time between failures. Together, the two numbers show which half of reliability needs the work.