# The Runbook Test

> A runbook is real if someone else can run it on the worst night of the quarter.

Source: https://christwentyfour.com/blog/the-runbook-test

Most runbooks are written by the person who needs them least, right after the fix, while the context is still in their head. They are accurate and nearly useless. The usual failure is skipped steps. To the author, they stopped being steps.

A runbook is real if someone else can run it on the worst night of the quarter without asking you anything.

## Test it with someone who has less context

Hand it to that person in a low-stakes window and watch without helping. Every question they ask is a defect. Fix it afterward. Reading it yourself does not count.

## Write for a tired stranger

That stranger is often you, six months later. Include exact commands, what success looks like, what to do on failure, who to wake, and a stop condition. "If this is not resolved in twenty minutes, escalate" beats an open list of things to try.

## Write it during the work

Nobody schedules a quarter of documentation. Ten minutes at the end of an incident, while it is fresh, beats an hour of memory later. The commands are already in your shell history.

## Rehearse the critical ones

Restores, failovers, and rollbacks get practiced on a schedule, where a mistake is cheap. An unrehearsed recovery is a guess.