Postmortem Metrics is the analytics layer over the postmortem process itself. Are postmortems being written? Are they on time? Do they have meaningful action items? Do those action items get done? The page surfaces the answers across teams so the practice does not slowly degrade into "we wrote it but never read it."
Not every incident warrants a postmortem. Sev-1 always; sev-2 if customer-impacting; sev-3 only if learning value. The page shows coverage: percent of qualifying incidents with a published postmortem. Below 80% triggers a review. The qualifying-incident definition is configurable per tenant.
A postmortem written 3 weeks after the incident is mostly archaeology. The page tracks time-to-publish: median, p95, longest-overdue. Default target is 5 business days. Overdue postmortems are visible per team, with the responsible engineer pinged automatically.
Action items in postmortems often get written and then never done. The page tracks completion: percent of action items completed, distribution by owner, and a list of stale items (open more than 30 days). Stale items resurface in the team's weekly report so they do not get archived in silence.
Each team has its own slice: coverage, time-to-publish, action completion. Side-by-side compare across teams identifies practices that work (and could spread) and teams that are slipping (and could use support). The view is informative, not punitive, shared organizationally rather than as a leaderboard.
Subscribe to Nova AI Ops on YouTube for demos, tutorials, and feature deep-dives.
Postmortems should be a habit, not a chore. Postmortem Metrics is the page that keeps the practice honest.