Home → Leadership
Burnout and Load: Sustainable Pace, Not Heroes
Every team has one: online at 2am, fixes what nobody else can, saves the sprint on the last day. Everyone likes them, everyone thanks them. Then one day they resign, and the team spends three months not understanding what happened.
- A hero is not an achievement, it is a symptom. The problem is that something needs rescuing at all.
- Load that is not measured is not managed. Pages, interruptions and night hours are all countable.
- Burnout shows in behaviour before output. Going quiet is the earliest signal.
- "Take it easy" is not recovery. Recovery that is not on the calendar does not exist.
From the field: 60% of the pages on one person
Our on-call rotation was fair on paper: five people, weekly shifts. One person was permanently tired anyway. "The rotation is even," we thought, and moved on.
Then we pulled three months of alert history. Sixty percent of the pages that woke someone at night did not fall in that person's shifts — whoever was on call, the hard alerts got escalated to them. The rotation was fair; the actual load was not.
The second finding was more uncomfortable. Most alerts came from the same three services, and the permanent fixes had been sitting in the backlog for six months. Because every rescue succeeded, the fix never became urgent.
What we changed
- Night pages were billed to the next day. For every alert that woke someone, no sprint work was expected the following day. This made the load visible for the first time.
- Handover became a one-hour meeting. What happened, what needs fixing. The items went straight into the sprint, not the backlog.
- The "who do we call" list was abolished. The on-call engineer handled hard alerts too, with a joint look the next morning if needed. Two painful weeks; after three months, a team that knew twice as much.
Four months later night pages had more than halved. Not because of a clever fix: the repair work was no longer on a someday list, it was inside the sprint.
Burnout shows in behaviour first
Managers usually notice burnout when output drops. But output is the last thing to go; people cut from themselves first.
- No longer joining debates
- Review comments down to one line
- Never volunteering for new work
- Not taking leave — and staying online when they do
- "I don't mind either way" becoming frequent
- Deadlines slipping
- Error rate climbing
- Skipping meetings
- A resignation letter
What the early signals share is that none of them is a complaint. Nobody says "I am exhausted"; they simply take up less space. That is why the question in a 1:1 should not be "are you too busy?" but "how many evenings in the last two weeks did you open the laptop after hours?"
Four countable things
| Measure | Where to look | What it tells you |
|---|---|---|
| Pages that woke someone | Alert history | Real on-call load, not the rota |
| Uninterrupted hours | Calendars | Whether focus time still exists |
| Out-of-hours activity | Commit and message timestamps | Quiet overwork |
| Unused leave | HR | Accumulated fatigue |
You collect these to make load visible, not to monitor individuals — and you have to say so to the team. Otherwise the measurement itself becomes a new source of pressure.
About crunch
Every team has intense weeks; that is not the problem. The problem is intensity that never returns to normal. A healthy crunch has three conditions:
- It has an end date and the date is real. Say "this month is heavy" and let it run three months, and nobody believes you again.
- The reason is stated. Someone who knows why you are rushing works differently; someone who does not just gets tired.
- Recovery is on the calendar. "We'll rest afterwards" is not recovery; "the sprint after launch runs at half capacity" is.
A manager who messages at 11pm teaches the team what hours look like, even while writing "no rush, look tomorrow". Scheduled send is a small setting with a large cultural effect. And a manager who stays online during their own leave has banned the team from taking leave — without banning anything.
Checklist
- Whose night was interrupted last month, and how often?
- Which three services produce most alerts? Are their fixes in the sprint or the backlog?
- How many people, if they took a week off, would stop work?
- Who has not taken leave in three months?
- Who is sending out-of-hours messages — including me?
- Did the last crunch end when it was supposed to? Did recovery reach the calendar?
- Has anyone gone quiet? (Silence is the easiest signal to miss.)
Conclusion
Sustainable pace does not mean working slowly. It means a team can hold the same pace next quarter too. A team running on a hero looks fast, but its speed is capped by one person's endurance.
The simplest measure: what happens when the most critical person takes two weeks off? If the answer is "nothing", you have built the right things.