Operations
Runbook
A step-by-step operational document telling an on-call engineer exactly how to diagnose + resolve a specific class of incident. Reduces mean-time-to-recovery and gets first-time responders unblocked.
By Priya Ranganathan · Last updated July 22, 2026
In plain English
'When the payments queue backs up, do X, then Y, then Z. Escalate to on-call-lead only if step Z fails.' Written once by the person who last resolved it; consulted every time it recurs.
Example
Stripe's public API status page links to a customer-facing runbook for common webhook signature failures. Internally, every alert routes to a Slack channel with a pinned runbook link.
Why it matters
Companies without runbooks re-derive the same fix on every incident, and lose it when the person who knew retires. Runbook-first culture halves MTTR within a quarter of adoption.
Common mistakes
- Writing runbooks that describe the system instead of the fix — read the code, read the runbook
- Not linking alerts to runbooks — the on-call engineer wastes 10 minutes searching
- Not tagging runbook age — a 3-year-old runbook is worse than none (misleads confidently)