On-call rotations for global teams
On-call for a team spread across three or more time zones needs different shapes than a co-located team. The patterns that work.
Every payment platform ends up with 24/7 on-call. If the team is co-located, the shape is a weekly rotation with escalation. If the team is spread across three or more time zones, that shape doesn't work: someone is always being paged in the middle of the night.
Written August 2026 from a review of on-call rotations across two organisations.
What doesn't work
- Single rotation across all zones. Someone always has a bad week.
- Follow-the-sun without handoff discipline. Incidents span zones; without written handoffs, context is lost.
- "Primary in one zone, backup elsewhere" without clear escalation. Ambiguity about who owns the incident produces delays.
What works
- Multiple rotations, one per zone. Each zone has its own primary during its business hours. Incidents outside business hours page the primary in the currently-awake zone.
- Explicit handoff at zone transitions. A written status of every open incident is transferred to the incoming primary. Not verbal; not "you'll see the tickets"; a written status.
- Escalation across zones only for severity. A P1 incident at 3 AM in zone A pages the zone-A primary; a zone-B backup is only paged if zone A doesn't ack within the SLA.
- Runbooks localised to the incident, not to the zone. The runbook is the same regardless of who's on-call.
The other important thing
Time off after on-call is not optional. An engineer who took a Friday-night page owes themselves Saturday off. Managers who don't enforce this see burnout.
Global on-call is expensive to run well. The alternative — expecting engineers to be paged at 3 AM every week without accommodation — is more expensive in retention.