The new hire's first page should not be their first page.

The new hire's first page should not be their first page.

You know the version I mean: senior engineer finally admits they can't take the pager forever, hands the rotation to the newest person, and offers a wiki link and the words "you'll figure it out." Three months later either the new hire has quietly become a second Dave (the person nobody can replace), or they've frozen at 2AM on a page nobody taught them to read. Both outcomes were decided the day onboarding was defined as osmosis.

We just published a one-page on-call onboarding checklist that replaces osmosis with a written path — read-only observer to certified responder in two to three rotations. The short version:

Four documents, not forty. The reading list that actually gets read is short: the incident response plan, the severity matrix, the first-30-minutes runbook, and the current rotation schedule. Everything else arrives when a page forces it. Day-one access is read-only — dashboards, alert history, past incident timelines. Read access is cheap and safe; it's also the difference between shadowing and spectating.

Shadow rotation: they take the timeline, not the wheel. Every page pairs, triage happens out loud, and the new hire writes the incident timeline — timestamps, who did what, what the alert actually said. Then a ten-minute debrief within 24 hours: what surprised you? The surprises are the syllabus. If they were shocked that a disk-full page resolves with a log rotate instead of a server rebuild, that gap becomes a runbook fix before the next rotation. Good onboarding improves the docs for everyone.

The drill flips the roles. One 45-minute tabletop where the new hire drives a first-30-minutes scenario while a senior observer stays silent. The pass condition: unprompted, they classify severity, open the timeline, draft the first customer communication, and state the escalation threshold. Drill the social parts on purpose — waking the boss, paging a person instead of a channel, saying "I don't know yet," and deciding to stop investigating and start escalating. The first real escalation at 2AM should be muscle memory, not a moral crisis.

Keys and break-glass get a hands-on walkthrough in daylight. If emergency access is a sealed kit, the new hire walks it once: where it lives, when it may open, what gets logged immediately after. The classic break-glass failure is a credential nobody has ever typed. Same rule for the password vault and the API keys the rotation depends on — plus the five-minute leak speech now, before an incident makes it urgent.

Certification is a gate, not a feeling. A dozen checkboxes with a sign-off line: reading list done, one shadow rotation with timeline, one handoff written, one drill passed, break-glass walked, severity matrix recited. When the boxes are checked, the next rotation is theirs — no vibes, no "whenever you feel ready." Cap the shadow period at two rotations, because an open-ended "shadow until ready" quietly becomes never, and that's how teams end up with a single point of failure named Dave. And the first solo shift is still staffed — by a phone. Calling is free; breaking things alone is expensive.

The honest version for a two-engineer team: this is one shared doc and a calendar invite, and it still matters — because it's how you buy your first real day off. The cheapest insurance a small team can own is a second person who knows where the rollback lives.

Full checklist (the certification gate template included): On-Call Onboarding Checklist for New Hires

If you're building the surrounding system: the 2AM handoff test · on-call handover template · severity matrix, 3 levels

Toolkits: Ops Starter Kit — $14 · First 30 Minutes runbook — free