Every production incident is a data point. Organisations that run effective postmortems convert data points into improved systems.
The effective postmortem: a blameless timeline. What went well. What went poorly. Root cause analysis (the systemic cause, not the surface cause). Specific, assigned, time-bounded action items.
"We should improve our testing" is not an action item. "Before [date], [person] will add a smoke test for [specific failure mode]" is an action item.
The blameless culture: incidents caused by human error are system design failures. If blame is assigned to individuals, people hide incidents rather than report them. The postmortem loses its value.
— Dick Bassey | DevDick | 2025