Here the FireHydrant company writes that MTTR (mean time to recovery) metric is overrated.

I tend to agree with their justification: not all incidents are the same. Not all systems are equally critical as well as not always the cause or rather a set of causes is easy to mitigate.

However, it’s still important to measure the quality of your incident response somehow. The propose two metrics instead:

- Mean time to detection - basically, how long it takes for you to spot a problem. I would also ask an additional question here: was a problem reported by your observability systems, or by your colleagues or customers?

- Mean time to retrospective - or mean time to postmortem if you wish. I think this is a good one. It’s true that some details may fade away from memory, so sooner you do a retrospective - the better! Also, it’s f you actually measure this time, it can help you to prioritize meetings and put postmortems as a higher priority comparing to, for example, a team standup.

#observability