Post-Mortem Best Practices That Actually Drive Change

After reviewing numerous post-mortems, it became clear that most are ineffective. They follow a similar structure: identify the issue, write a report, hold a meeting, list action items, and forget to follow up. The author describes this as the "Post-Mortem Nobody Learns From" phenomenon. The blameless culture, which aims to avoid punishing individuals, often becomes an excuse to avoid accountability.

It doesn't mean the person responsible for the bad deploy is unaccountable; it simply means they are not the sole focus of the investigation.

The author shares their own post-mortem template, which includes sections for incident details, impact analysis, timeline, root cause, contributing factors, what went well and wrong, action items, and lessons learned. The key to effective post-mortems, according to the author, lies in the action items. However, industry-wide, action items from post-mortems have only a 30% completion rate.

To combat this, the author proposes three rules. First, limit action items to a maximum of three per post-mortem. If there are too many, the problems need to be re-evaluated. Second, every action item must be turned into a JIRA ticket linked to the next sprint. If it's not vital enough for immediate action, it's not an action item. Third, review the completion of action items in the next post-mortem. This creates accountability without assigning blame.

The author introduces the concept of "Repeat Incident Rate," which measures the percentage of incidents with the same root cause as previous incidents. When they started tracking this metric, their repeat rate was 45%. After implementing the three rules, it dropped to 12% over six months. This is the real measure of whether post-mortems are effective. If you're seeking improved incident learning loops and pattern detection, the author suggests exploring their solution at Nova AI Ops.