B8A Tech All articles
Engineering Leadership

The Rotation Is the Retention Plan: Redesigning On-Call Before Your Best Engineers Resign

B8A Tech
The Rotation Is the Retention Plan: Redesigning On-Call Before Your Best Engineers Resign

Photo by Photo by Vitaly Gariev on Unsplash on Unsplash

When a senior engineer resigns, the exit interview rarely says what actually happened. Compensation is cited because it is concrete and non-accusatory. Growth opportunity is cited because it is polite. What is rarely said — and what former colleagues all know — is that the engineer spent eleven of the last twelve months carrying a pager that went off four times a night, and decided that no equity refresh was worth another quarter of it.

On-call culture is the attrition driver the industry keeps declining to measure. And unlike compensation, it is almost entirely within a manager's control.

The Math of Unsustainable Rotations

Most broken rotations break for arithmetic reasons, not cultural ones. Consider a common configuration: a team of eight, primary and secondary coverage, one week per rotation. Each engineer is primary one week in four. That sounds tolerable — until you account for the follow-on work. Every incident handled during on-call generates post-incident tasks, and those tasks land during the engineer's 'off' weeks. The pager ends, the work continues.

Now layer in the sleep math. A rotation with four overnight pages per week is not one disrupted night — it is a week of degraded cognition, because interrupted sleep has a cumulative cost that a morning coffee does not repay. Engineers on such rotations are effectively working at reduced capacity for a quarter of their time while being measured as if at full capacity.

The pattern generalizes: rotations below six people reliably produce burnout, regardless of how quiet the system is supposed to be. Below that threshold, there is no recovery buffer. One bad incident cascades into a bad week, one bad week into a bad quarter.

The Follow-the-Sun Illusion

Follow-the-sun coverage is the most frequently proposed fix and the most frequently botched. Handing the pager across time zones sounds elegant; in practice, it often means handing pages to a team with less context, weaker tooling access, or a mandate to escalate anything nontrivial — which means the page simply travels back to the original team, now with a delay attached.

Follow-the-sun works only when each region can genuinely resolve the majority of its pages independently. If your escalation rate across regions exceeds 20%, you have not built global coverage; you have built a relay race where every leg drops the baton.

Designs That Actually Hold Up

Several structural changes consistently improve rotation health, and none require headcount:

Split primary and escalation tiers honestly. The primary should be able to resolve 90%+ of pages with documentation and standard access. If your primary pages require the person who wrote the service, that is a knowledge-sharing problem wearing an on-call costume.

Cap pages, then fix the causes. Track pages per shift per engineer. Above roughly two overnight pages per shift, the rotation needs intervention — not tougher engineers. Every page above the threshold should trigger a ticket against its root cause, and that ticket should have a real owner.

Pay the rotation in engineering time. Teams that treat on-call as a tax on top of full feature load lose engineers. Teams that explicitly reduce sprint commitments for on-call weeks keep them. The trade is honest, and honesty retains people.

Rotate the rotation design itself. Whoever designs the schedule should also carry the pager under it. Schedules designed by people exempt from them converge on cruelty with remarkable speed.

Reliability and Retention Are the Same Metric

The argument that sustainable rotations sacrifice reliability inverts reality. Exhausted engineers make slower incident decisions, write shallower post-mortems, and stop proactively hardening the systems they are too tired to think about. Burnout is not the price of reliability — it is the leading indicator of its decline.

If your organization is losing engineers and the exit interviews say 'compensation,' it may be worth asking a follow-up question: how many nights of interrupted sleep did this person have in their final six months? The answer is usually the retention plan you were already running — you just hadn't read it.

All Articles

Related Articles

Week One Is a Preview: How Onboarding Friction Quietly Sets Your Retention Trajectory

Week One Is a Preview: How Onboarding Friction Quietly Sets Your Retention Trajectory

The 3 A.M. Tax: Why Untended Alert Loads Are Quietly Reshaping Engineering Careers

The 3 A.M. Tax: Why Untended Alert Loads Are Quietly Reshaping Engineering Careers

The New Arithmetic of Make-vs-Buy: A Decision Framework for the Managed Services Era

The New Arithmetic of Make-vs-Buy: A Decision Framework for the Managed Services Era