A postmortem that is a story is read once and filed. One that ends in a list of owned, dated actions changes something.
impact who was affected, for how long, doing what.
"11,400 users could not check out for 41 minutes."
timeline detection, escalation, mitigation, resolution.
with the gap between the first two called out.
cause contributing factors. plural. no single root.
actions each with an owner and a date. and each one
either prevents recurrence, or shortens detection,
or shortens mitigation. nothing else counts.
Time-to-detection is the number worth extracting every time, because it is usually the largest part of the outage and is the one nobody looks at. Insisting on plural contributing factors is what stops the document blaming a person or a deploy — a system that can be taken down by one mistake has more than one problem. Actions without an owner and a date are a wish list, and reviewing the previous postmortem’s actions at the start of the next one is what makes them real.