The On-Call Hygiene Checklist
Five things that decide whether a rotation costs you a bad night or a bad month.
You instrument your services. This is the same instinct, pointed at the rotation itself — a two-minute self-audit. Run it at the start of a rotation, or after a week that hurt. Every unchecked box is a place the next incident gets more expensive than it needs to be. // an operational self-audit, not a wellness quiz
01 · HANDOFF
The rotation changes hands in writing, not in a hallway.
- Taking on-call: you have a written list of open incidents, known-fragile services, and in-flight changes or freezes for the week.
- Taking on-call: you know who to escalate to per domain without looking it up mid-incident.
- Handing off: what fired, what you silenced (and why), and what's still smoldering is written down — not "it was quiet, you're good."
If handoff is verbal, the next responder rediscovers everything you already learned — at 3am, under load.
02 · ALERTS
You're triaging signal, not absorbing noise.
- More than half of last week's pages were actually actionable — something you had to do, right then.
- No single alert paged you more than twice for the same non-actionable reason.
- Every alert that pages a human passes the test: if this fires and nobody looks for four hours, does a user notice? If no, it's a dashboard, not a page.
If most pages aren't actionable, you're not on-call — you're the alert system's noise sink. That's a tuning backlog, and it's documentable.
03 · DEBRIEF
The human cost is logged, not just the root cause.
- Your postmortems — or a private note — capture responder cost (sleep lost, time-to-calm, next-day impact), not only the technical timeline.
- After a bad night you write down the one thing that made it harder than it needed to be, before you rationalize it the next morning.
The RCA logs what the system did. If nothing logs what it cost the operator, that cost stays invisible — until someone quits.
04 · RECOVERY
Your off-week is protected like an SLA, not a suggestion.
- There's an explicit expectation after a rough on-call night — late start, no meetings, or comp time — and you actually take it.
- Your off-rotation weeks are genuinely off: you're not the silent backstop for whoever is on.
Recovery you don't schedule is recovery you don't get. On-call is a load; it needs a defined offload.
05 · RECURRENCE
You're tracking patterns across incidents, not just surviving each one.
- You'd notice if the same service paged you 3+ times this quarter — because it's written down somewhere, not just a feeling.
- Before any rotation or headcount conversation, you can point at a log instead of arguing from memory.
One bad night is noise. The same service, the same class of alert, the same missing runbook — repeated — is a documented operational pattern with a business cost attached. That's the version leadership acts on.
Want it as a one-page PDF? Print it, pin it by your desk, hand it to your rotation. Enter your email and it's yours — plus the occasional short note on on-call ops. No spam, unsubscribe anytime.
Get the PDF →