The blameless post-mortem principle
The most important rule: post-mortems are about systems, not people. When an incident happens, it's almost never because one person made a careless mistake. It's because the system allowed — or even invited — that mistake. A deployment process that doesn't require a second review, a monitoring setup that doesn't alert until 30 minutes after an outage starts, a test suite that doesn't cover that edge case.
If your post-mortem ends with "we told the developer to be more careful," you haven't run a post-mortem. You've had a performance review. The next incident will happen for the same reason.
When to run a post-mortem
- Any outage that lasted more than 30 minutes in production
- Any incident that affected paying customers
- Any data incident (data loss, data exposure)
- Any incident that breached your SLA
- Optionally: severe bugs that required an emergency hotfix
For smaller incidents or near-misses, a brief written summary (5 minutes, no meeting) is sufficient. Reserve the full post-mortem process for significant incidents.
The post-mortem template
⚠️ Action items without due dates and owners get done 0% of the time. Every action item must have one named owner and a specific date — not "next sprint" or "soon."
Running the post-mortem meeting
The meeting should happen within 48 hours of the incident being resolved, while details are fresh. Format:
- Walk the timeline (10 min) — the author presents what happened, in order, factually
- Root cause discussion (15 min) — ask "why" 5 times until you reach a systemic cause, not a personal one
- What went well (5 min) — don't skip this; it anchors the blameless culture and identifies things worth preserving
- Action items (10 min) — assign owner and date to each item before the meeting ends
Total: 40 minutes. More than this and people lose focus. Less than this and you haven't done the work.
The 5 Whys technique
Use this to find the real root cause instead of stopping at the surface:
- Why did the outage happen? → Payment calls were hanging
- Why were they hanging? → API timeout was removed
- Why was it removed? → Config change by an engineer
- Why was it not caught? → Config changes don't require review
- Why don't config changes require review? → We never added that to our process
Root cause: No peer review process for configuration changes. Action item: Add config change review to deployment checklist.
Track post-mortem action items in Resolvo
Log incidents, assign action items, track completion. Issue tracking built for engineering teams. Free to start.
Try Resolvo Free →