On-Call Compensation Models for Engineering Teams

At a glance

On-call compensation is the set of rules that decide what a person receives for carrying a pager, and what they receive when that pager goes off. Nearly every policy is assembled from the same small set of parts: payment for availability, payment for active work, different treatment of nights, weekends and public holidays, a floor under short callouts, and recovery time after a broken night.

The parts are not complicated. The difficult work is deciding which parts apply to your rotation, how the work gets recorded, and who signs off on the answer.

This page describes the shapes of the common models and the questions each one forces you to answer. It deliberately names no rates, no amounts and no country's rules. Those belong to your employment counsel, your payroll function and the contracts you already have.

Why the policy is an operational document

Compensation is usually filed under people and policy, and then treated as if it had no effect on the system. It has several.

It prices disruption. When a night page costs the business something visible, the alert that fires at three in the morning for a condition that could wait until nine acquires an owner and a deadline.

It shapes reporting. If active incident work is uncompensated and unrecorded, teams stop recording it. The rotation then looks quiet in every report while the people on it are steadily worn down, and you lose the data you would need to argue for more headcount.

It determines who can do the job. An unpaid rotation that regularly destroys sleep is not equally available to everyone on the team. People with caring responsibilities, health conditions, or no slack in their week absorb it worst.

It is a retention question. Rotations are rarely the stated reason someone leaves, but a rotation that costs a person their weekends without acknowledgement is a standing reason to answer a recruiter.

These are the four arguments that travel outside engineering. Take them with the paging data attached — pages per person, per service, and per hour of the day — rather than as assertions.

The building blocks

Availability payment. A fixed amount for carrying the pager for a defined period, paid whether or not anything happens. It prices the constraint: staying reachable, staying sober, staying near a working connection, cancelling plans that cannot be interrupted. It is usually expressed as a flat sum per shift or per week, or as a proportion of base pay.

Standby by the hour. The same idea, metered. Instead of a shift-sized amount, availability is paid per hour carried. This tracks uneven rotations more honestly — partial cover, follow-the-sun handovers, someone picking up half a weekend — at the cost of more administration.

Active response payment. Payment for time actually worked during a page: triage, mitigation, comms, the write-up if you count it. This is the part that distinguishes a quiet week from a punishing one. It requires an agreed way to record time, which is the part easiest to leave undecided — decide it before the payment exists, not after.

Minimum callout period. A floor applied to each active engagement. Without one, a page that takes twenty minutes to resolve is recorded as twenty minutes, which ignores the hour of lost sleep on either side of it. A minimum callout period is the policy's acknowledgement that interruptions cost more than their duration.

Unsociable-hours treatment. Different handling for nights, weekends and public holidays, either as a higher rate or as separate premiums. This is where a policy expresses that not all covered hours are equally expensive to the person covering them.

Recovery time. Rest after a shift that was actually disrupted — a late start, a day off, or a defined block of protected time. Recovery time is a fatigue control rather than a payment, and it is the one building block that directly protects the next incident's response quality.

Comparing the models

How to decide

Work through these questions in order. Each one narrows the choice.

1. What do your contracts and local requirements already oblige you to do? This is the constraint, not the design. Answer it first, with people qualified to answer it, and treat everything below as the space that remains.
2. How disruptive is the rotation, measured rather than assumed? Count pages per shift, pages outside working hours, and the number of shifts in the last quarter that cost someone sleep. A policy built on an impression of the rotation will be wrong in whichever direction the impression is wrong.
3. Is the burden even? If one service or one time zone absorbs most of the out-of-hours pages, a single flat allowance pays the person carrying most of them exactly what it pays the person carrying none. Split the page counts by person and by service before you decide the burden is even.
4. What can you actually administer? An elegant model that requires precise time records nobody keeps will decay into an approximation, and the approximation is what people end up being paid on. Choose the model your payroll process and your team's habits can sustain.
5. What behaviour does this create? Every model has an incentive edge. Active response payment can discourage people from resolving quickly. Allowance-only can discourage reporting. Name the edge and decide how you will watch for it.
6. What happens when the rotation gets worse? Decide in advance what triggers a review: a page-volume threshold, a team-size threshold, a change in coverage window.

Writing it down

A compensation model that lives in a manager's head is not a policy. Six things need to be explicit.

• The covered period. Exact start and end times, the time zone they are expressed in, and how handover boundaries are treated.
• What counts as active work. Whether it starts at the page or at acknowledgement, whether the write-up counts, whether a false alarm counts.
• How time is recorded. The mechanism, the deadline for submitting it, and who approves it. If the mechanism is heavier than the payment is worth, people will stop using it.
• Definitions of night, weekend and holiday. Including which calendar applies for distributed teams.
• Swaps, overrides and cover. Who is paid when someone takes another person's shift at short notice.
• Recovery time. What triggers it, how much, who approves it, and whether it expires.

A worked example

The following is an illustrative example, not a description of a real company.

A platform team of five engineers covers a single production service, one week per person, primary only. Over a quarter they record thirty-one pages, of which nine arrive between midnight and six in the morning. The nine night pages are spread unevenly: one engineer absorbed four of them.

Under an allowance-only model, all five were paid identically for the quarter. The engineer with four night pages raised it at a retrospective; nobody had the data to argue about it, because active time was never recorded.

The team changes three things. They keep the weekly allowance for availability. They add an active response payment with a minimum callout period, so a short night page is recorded as a meaningful interruption rather than as the handful of minutes it took to resolve. They add a rule that any page between midnight and six triggers a late start the following day, approved by nobody — it is automatic.

The second-order effect is the interesting one. With night pages now carrying a recorded cost, the two alerts responsible for six of the nine night pages get an owner and a deadline. One is re-routed to a business-hours queue; the other gets a threshold change and an auto-remediation. The compensation policy did not fix the alerts. It made them visible enough to be worth fixing.

Common failure modes

Paying for availability but never for the work. The rotation looks compensated on paper while the worst weeks go unacknowledged.

No minimum callout period. Short night pages are recorded as short, and the policy systematically understates disruption.

Recovery time at the manager's discretion. Discretionary rest is taken by the people who feel most secure asking for it, which is not the same as the people who need it most.

Undefined swap rules. Someone covers a colleague's weekend as a favour, and the payment follows the schedule rather than the person who actually carried the pager.

Compensating instead of fixing. Payment is a fair exchange for disruption, not a licence to leave a rotation broken. If the allowance is the only thing keeping people on a loud rotation, the policy is masking an alert-quality problem.

Setting it once. Team sizes shrink, services get added, coverage windows widen. A policy with no review trigger drifts out of fairness silently.

What to do next

Today, without waiting for a budget conversation:

• Pull the last quarter's pages and count how many landed outside working hours, and per person.
• Write down what your current policy actually says. If nothing is written down, that is the finding.
• List which of the six building blocks you use, and which you have never explicitly decided.
• Take the questions in "How to decide" to whoever owns employment and payroll in your organisation, and get the constraints in writing before designing anything.
• Add a review trigger — a page-volume number, a team-size number, or a date.