Template

Blameless Postmortem Template

A ready-to-use postmortem document for IT incidents, so the team learns what really happened and fixes the cause, not the person.

Baca dalam Bahasa Indonesia →

Published Setya

When the service is back, most people want to move on. The postmortem is where the team works out what to change, so the same incident does not happen again.

A good postmortem says plainly what happened and why, and names who will change what. It looks at the system and the process, not at who pressed the wrong button.

This template keeps the review structured, so it doesn't turn into a long meeting with no actions.

1. Summary

Summary

Incident: [short title, e.g. "Email unavailable for head office users"]

Date and duration: [date], [start time] to [end time] ([total duration])

Severity: [P1 / P2 / P3]

Postmortem owner: [name]

What happened: [three sentences: what failed, who was affected, how it was fixed]

2. Impact

Impact

Users affected: [who and how many, e.g. "about 300 users at head office"]

Services affected: [services and what users couldn't do]

Business impact: [missed transactions, delayed processes, customer complaints, cost if known]

SLA impact: [was any SLA breached, and by how much]

3. Timeline

Use one time zone and write facts, not opinions. The times in your incident messages are a good starting point.

TimeWhat happened
[hh:mm]First alert or user report
[hh:mm]Incident declared and team engaged
[hh:mm]Escalated to [team / vendor]
[hh:mm]Mitigation applied
[hh:mm]Service confirmed back to normal

4. Root Cause

Root Cause

Trigger: [the event that started the incident, e.g. "certificate on the load balancer expired"]

Root cause: [why the trigger was possible, e.g. "certificate renewals depend on one person's reminder"]

Why it wasn't caught earlier: [missing monitoring, unclear ownership, untested process]

Ask "why" until you reach a cause in a process or system that the team can change. If the answer ends with a person's name, keep going.

5. What Went Well and What Didn't

What went wellWhat could be better
[e.g. alert fired within 2 minutes][e.g. runbook was out of date]
[e.g. service desk informed early][e.g. vendor escalation took 40 minutes]

6. Actions

Every action needs one owner and a due date. Fewer actions that get done are better than a long list nobody follows.

ActionTypeOwnerDue date
[e.g. add certificate expiry monitoring]Prevent[name][date]
[e.g. update the runbook]Detect / Respond[name][date]

Running the Review

  • Invite the people who worked on the incident, the service owner and anyone who owns an action
  • Start by saying the review is blameless: the goal is to improve the system, not to find who is at fault
  • Walk through the timeline first, then the cause, then the actions
  • Share the final document with management and the service desk, and track the actions until they're closed

Keep going

Everything in Service Delivery →

Esc