Running on-call without burning people out

What changed when we treated pages as a budget we owed the team back: tighter alerts, real follow-the-sun, and a blameless review that people stopped dreading.

The budget framing did more work than any single policy. Every page that didn’t need a human at 3am was debt the team was carrying, and debt gets paid down — by tightening the alert, fixing the underlying flakiness, or deleting the check entirely. Within a quarter, the number of pages per week had dropped by more than half, not because we got better at ignoring them, but because most of them shouldn’t have existed.

Follow-the-sun only worked once we were honest that our “24-hour” rotation had really been one region covering three shifts. Splitting it across time zones that were actually staffed meant nobody carried a pager through their own night more than once every few weeks, and the handoff notes got sharper because they were written for someone about to start their day, not someone about to sleep.

The blameless review was the hardest habit to build and the one that mattered most. The format that worked: what happened, what the on-call engineer knew at each decision point, and what would have made the right call obvious sooner — never who should have known better. People started bringing their own near-misses instead of waiting to be paged into an incident, which is the actual signal that the culture had shifted.