On-call burnout: the signs and the fixes
On-call burnout shows up as slower acknowledgements, muted or ignored alerts, swap and opt-out requests, and people leaving. Leaders can see these signals in the data and in how the rota behaves. Fix it by cause: cut noise, lower frequency, make the schedule predictable, and give people time to recover.
Why leaders should care
On-call burnout is what happens when the load of carrying a pager outruns what people can sustain. This page treats it as a management problem, because the causes are things a leader can change: how many alerts fire, how big the rotation is, how predictable the schedule is, and whether anyone gets time to recover. It is not medical advice. It is about the business.
The business case is direct. An engineer who leaves takes knowledge of the systems with them, and replacing them costs months of hiring and ramp-up. A rotation that is tired responds more slowly, which customers feel as longer outages. And a reputation for punishing on-call makes hiring harder. Retention is the visible cost, and slower incident response is the hidden one.
Signals leaders can see
Burnout rarely announces itself. The signals are in the rota and the data, and most of them need no new tooling.
| Signal | Where to look | What it may mean |
|---|---|---|
| Slower acknowledgement | Time to acknowledge, by hour and person | Tired people, or alerts nobody trusts |
| Alerts muted or filtered by individuals | Personal notification settings, chat | Noise has driven people to defend themselves |
| The first step goes unanswered more often | Escalations caused by no answer | People are not reaching or not responding |
| Swaps and opt-outs increasing | The rota history | The schedule is unbearable for some |
| Load concentrated on a few people | Pages per person per shift | Unfair distribution, or a few hero responders |
| Quiet reviews, less discussion | Incident review attendance and notes | People have stopped expecting things to change |
| People citing on-call when they leave | Exit conversations | The cost has already been paid |
Read the signals in three cuts. By person, to see whether the load is fair. By hour, to see whether nights are the problem. By alert, to see whether a handful of rules cause most of the pain. Compare against the team’s own past months, not against another team, because every system has a different normal.
No single signal proves anything. A cluster of them, moving the same way over a few weeks, is a reason to act.
Fixes by cause
Fix what is actually wrong. Four causes account for most of it, and each has a different remedy.
| Cause | What it looks like | The fix |
|---|---|---|
| Noise | Pages that need no action, at any hour | Audit, delete, reroute and tune alerts |
| Frequency | The same people on call too often | A bigger rotation, another time zone, or fewer commitments |
| Unpredictability | No way to plan evenings, weekends or leave | A rota published weeks ahead, easy swaps, fair rules |
| No recovery | Back at full duty after a night of pages | Recovery time, and rest built into the policy |
Noise and frequency
Noise. Being woken for something that needs no action is the most corrosive part of on-call, because it teaches people that the pager is not to be trusted. The remedy is the work in the 30-day alert fatigue plan: audit what fires, delete what nobody acts on, reroute what can wait. Cost falls as noise falls, since fewer pages mean fewer interrupted nights and fewer slow days after them.
Frequency. If the same few people are on call most of the time, no amount of tidying will help. A rotation that is too small has to be made bigger, whether by hiring, by adding a time zone or by cutting what the team commits to. How a small team can share the load fairly is in on-call for a team of five. If a rotation is thin, count what each person actually carried before deciding.
Unpredictability and recovery
Unpredictability. Being on call is easier to bear when the schedule is known. Publish the rota well ahead, make swaps one message, and avoid changing it at the last minute. Give people some control: preferred weeks, blocked-out dates, a way to step out for a while without a negotiation. Fair rules for holidays and weekends, applied to everyone, matter more than the rules themselves.
No recovery. A night of pages should not be followed by a normal day. Give people a later start, or time back in lieu, or lighter duty the next day, as a rule and not as a favour that has to be asked for. Pay has a role here too, and the models are covered in on-call compensation. Recovery time is a customer benefit in disguise: a rested engineer fixes the next incident faster.
A monthly load review
Put the load on one page each month and read it together. Per person, show night pages, weekend pages, the hours of interrupted sleep if you can estimate them, and the share of the rota carried. Beside it, put the team’s noisiest alerts and the acknowledgement times. Ask three questions: is the load fair, is it going up, and what would make it smaller?
Assign one action from each review to a named person. The value is not the report. It is that leaders see the load early and act while it is still cheap to fix.
What leaders own
Some things cannot be delegated to the team, because only a leader can decide them.
- Time to fix noise. Engineers will not delete alerts without protected time and permission.
- Staffing the rotation. A rotation that is too small is a hiring or scope decision.
- Saying no. New services and new alerts need owners and runbooks, or they wait.
- Pay and recovery rules. Written down, applied fairly, and reviewed.
- Reading the signals. Someone senior looks at the load every month and acts on it.
What leaders should not offer instead is advice to cope better. Resilience training is not a substitute for a quieter pager.
An example review
This example is invented. A manager notices that acknowledgement times have risen over two months, that two engineers have asked to swap most of their weeks, and that one has stopped speaking in incident reviews. The load page shows that the two engineers carry most of the night pages, and that one alert accounts for much of the noise.
The manager acts on three levels. The noisy alert is fixed that week. The rota changes so that nights are shared more evenly, with recovery time after any night of more than one page. And the manager tells the team what was found and what is changing, and asks whether it is enough. Two months later the swaps have dropped, and the quiet engineer is back in the reviews. Nothing about it needed a new tool. It needed someone to read the signals and to own the change.
Frequently asked questions
What are the first signs of on-call burnout?
Usually behaviour, not complaints. Acknowledgement times creep up, engineers mute or filter alerts for themselves, swaps and opt-outs become frequent, and reviews get quieter. By the time someone says they are worn out, the signals have often been visible for weeks in the rota and the data.
Is this a wellbeing problem or an operations problem?
Both, and leaders own the operations side. Noise, rota size, unpredictable schedules and missing recovery time are design choices that a manager can change. This page is about those choices. It is not medical advice, and anyone who is struggling should be supported as a person, not only as a rota entry.
Will paying more fix it?
Pay is part of fairness, and it helps, but it does not repair a noisy or undersized rotation. People burn out from being woken for nothing and from having no predictable time of their own. Fix the load first and then make sure the pay recognises what remains.
How do we find out what is wrong without a questionnaire?
Read the data you already have: pages per person, night and weekend pages, acknowledgement times, swaps and opt-outs. Then ask the people on the rota directly, and act on what they say. A short conversation after a hard week tells you more than a form.
Should individuals be able to opt out of on-call?
Sometimes, for defined reasons and periods, and it should be easy to ask. A fair policy says when opt-outs are allowed, how the load is redistributed, and how the person returns. Blanket exemptions for some people, without that structure, move the burden to others and create resentment.