Writing high-quality postmortem action items

WHAT THIS TESTS: Whether you understand that a postmortem's value is realized only through action items that actually get completed and prevent recurrence. The interviewer is checking that you can write items people will finish, not aspirational notes.

A GOOD ANSWER COVERS: A high-quality action item is specific enough that anyone can tell when it is done, for example 'add a timeout and circuit breaker to the fraud-check client' rather than 'improve resilience.' It has exactly one accountable owner, not a team in the abstract. It carries a priority and a due date so it competes fairly for engineering time, and it lives in the same backlog or tracker as normal work so it is visible and not forgotten. Effectiveness also depends on the type of fix: prefer items that eliminate the failure class (prevention), then items that detect it faster, then items that mitigate impact. A good postmortem produces a small number of high-leverage items rather than a long wish list nobody will finish.

COMMON WRONG ANSWERS: Vague exhortations like 'be more careful,' 'communicate better,' or 'add more monitoring' without specifics, owners, or deadlines. Generating a huge list that dilutes focus, or writing items no one is committed to completing.

LIKELY FOLLOW-UPS: How do you ensure action items are completed and not abandoned after attention fades? How do you prioritize across many postmortems? When is 'add an alert' the right fix versus a band-aid?

ONE CONCRETE EXAMPLE: After an outage caused by an unbounded retry storm, a poor item reads 'handle retries better.' A good item reads 'cap retries at three with exponential backoff and jitter in the order-service client, owned by Priya, P1, due next sprint, tracked as JIRA-1234,' which is specific, owned, prioritized, prevents recurrence, and is verifiable.