How to Design a Sustainable On-Call Rotation
Structural choices that make on-call sustainable: rotation size, escalation policy, alert ownership, and what to measure to know it's working.
The hidden cost of bad on-call design
Unsustainable on-call is one of the leading causes of senior engineer attrition. When engineers are paged constantly, sleep-deprived, and given no time to fix the underlying problems causing incidents, they leave. The institutional knowledge walks out the door with them.
How to size your rotation
The minimum viable on-call rotation is four people — enough for a weekly rotation where each engineer is on-call no more than once per month. Below that, burnout accumulates quickly. If your team is too small, that is a signal to either hire, scope down system ownership, or bring in external support.
- Aim for 4–8 people in any single rotation
- Limit primary on-call shifts to 1 week per 4 weeks
- Keep secondary/shadow rotations for knowledge transfer, not extended coverage
- Track hours paged per shift and report it to engineering leadership monthly
Escalation design that actually works
Good escalation is not just a PagerDuty configuration. It is a clear, documented agreement about what each tier is responsible for, how long they have to respond, and when to escalate. Every service should have a documented escalation path before it goes on-call.
Compensation and recognition
On-call engineers are providing a real business service outside of normal working hours. That deserves real compensation. Whether you use per-incident pay, per-shift stipends, or comp time, the policy should be explicit, fair, and respected by leadership.
The feedback loop: incident → improvement
Every incident that pages a human is either a signal to fix the system or a signal to remove the alert. If you are not systematically improving the system after incidents, your on-call load will never decrease. Dedicated reliability work time — at least 20% of engineering capacity — is how sustainable on-call becomes possible.