On-call compensation: paying for the pager without breaking the budget

Q: Can we just declare engineers exempt and avoid on-call pay?

In the US, exempt-status alone doesn't eliminate retention risk and in CA may not eliminate legal risk. In the EU, exempt-status is less protective. The trend is toward explicit on-call compensation.

Q: What about engineers who volunteer for extra on-call?

Pay them at the same rate as the rotation. Don't let 'volunteers' become a hidden labor pool — it creates inequity and obscures the real reliability problem.

Q: Should managers be on-call?

Managers should be on-call for incident command (coordinating response), not for first-line debugging. Including managers in the primary rotation is unusual above team-lead level.

Q: How do we handle on-call across timezones (follow-the-sun)?

If you have engineering in 2+ timezones, follow-the-sun rotations dramatically reduce off-hours pages. Comp typically reduces to a smaller stipend since night pages are rare. This is the lowest-burnout model.

60-Second Summary

On-call (carrying a pager that wakes you at 3am for production incidents) is real work. Most US tech companies pay nothing extra for it. Most European tech companies must pay for it by law.
The four models: (1) no extra pay (relies on exempt-status assumption), (2) flat weekly/shift stipend ($150–500/week), (3) per-incident pay ($50–200/incident off-hours), (4) productized — on-call is a measured workload that reduces other expectations.
Unpaid on-call has legal exposure: in California and several EU jurisdictions, on-call time may be 'controlled standby' and legally compensable. EU Court of Justice ruling C-580/19 is the reference precedent.
The HRBP's job is to design a defensible policy, audit rotation fairness (no engineer doing more than 1 week in 4), track on-call burnout signals, and ensure the comp model matches the actual burden.

On-call is the most under-managed piece of total rewards in tech. Engineers get woken at 3am to fix production. They lose weekends to incident response. In most US tech companies, they get paid nothing extra for it — the assumption is 'exempt salary covers it.' That assumption is increasingly fragile (legally and culturally) and is the single biggest hidden retention risk in modern engineering orgs.

What 'on-call' actually means

An on-call rotation is when one engineer at a time carries primary responsibility for responding to production incidents — usually paged via PagerDuty, Opsgenie, or similar. The engineer must be reachable, sober, and able to be at a laptop within ~15 minutes, 24/7, for the duration of their shift (typically 1 week). They may be paged 0 times a week or 30 times a week. They sleep next to a phone. They can't drink at dinner. They can't go hiking. The pager rules their week.

The burden is highly variable across teams. A well-instrumented platform team might page their on-call 2–3 times in a week, mostly during business hours. A poorly-instrumented payments service might page 15+ times, several at 3am. Comp policy must be sensitive to actual burden, not just rotation slot count.

The four comp models

Model	How it works	Pros	Cons	Used by
No extra pay	On-call expected as part of exempt salary	Simple; no budget impact	Legal exposure in CA / EU; retention risk; resentment	Most US startups, some big tech
Flat stipend	$150–500 per on-call week, regardless of incidents	Predictable; recognizes burden	Doesn't differentiate quiet vs busy weeks	Stripe (~$500/wk), many mid-stage startups
Per-incident	$50–200 per incident outside business hours	Aligns with actual burden	Incentivizes 'let it page'; complex tracking	Some platform/infra teams
Productized on-call	On-call counts as your primary workload that week (no sprint commitments)	Best for burnout; respects time	Reduces team velocity by 15–20%; some engineers prefer cash	Honeycomb, some SRE teams

The hybrid that's becoming standard

By 2026, the emerging norm at well-run engineering orgs is: (1) productized on-call (no sprint commitments during the week you're on), PLUS (2) a flat stipend ($200–400/week), PLUS (3) explicit time-off for any night that included a page (you take the next day, no questions asked). This costs more than 'no extra pay' but materially reduces senior engineer turnover.

Legal exposure of unpaid on-call

Plain-English summary by jurisdiction. This is not legal advice; loop in employment counsel before changing policy.

Jurisdiction	Status of unpaid on-call	Risk level
US federal (FLSA)	Exempt employees generally covered; non-exempt must be paid if 'engaged to wait'	Low-medium for exempt salaried engineers
California	Stronger 'controlled standby' test; courts have ruled compensable when restrictions are significant	Medium-high; document constraints carefully
EU broadly (CJEU C-580/19)	'Stand-by time' is working time when the worker's ability to use the time freely is significantly restricted	High; assume on-call time may be working time
Germany	On-call (Rufbereitschaft) is compensable; rates set by collective agreement	High; explicit pay required
UK	Working Time Regulations + case law: on-call at-home may not be working time, but active response is	Medium
India	Less developed case law; common practice is unpaid for IT/ITES exempt-equivalent roles	Low-medium today; trending higher

The CJEU ruling is the one to know

EU Court of Justice case C-580/19 (Stadt Offenbach am Main, 2021) ruled that on-call time qualifies as 'working time' when the response constraints significantly restrict the worker's ability to use that time freely. This is the basis for several follow-on cases and is the legal foundation for EU on-call pay claims. If you operate in the EU, your on-call policy needs to address this explicitly.

Rotation design: the people side

Minimum 4 engineers in a rotation. 3-person rotations are 1-in-3 weeks on call — that's an exit-driver.
Maximum 1 week in 4 on call. Hard cap. 1 in 5 or 6 is healthier.
Same person should not carry primary AND secondary the same week. Secondary backup is a real burden.
Holidays and weekends rotated explicitly. Don't let the most junior person always get Christmas.
Recent incident-heavy weeks earn a 'skip your next rotation' credit.
New hires don't go on primary on-call for at least 90 days. Shadow rotations first.

Burnout signals to monitor

Signal	Threshold	Action
Pages per on-call week (median)	>5	Reliability investment; don't compensate around it
Pages between 10pm–7am	>2 per week	Capacity / instrumentation problem; escalate to eng leadership
% of on-call weeks with ANY 3am page	>25%	Reliability emergency; pause feature work
Voluntary attrition rate on a team	>15% annually	On-call burden is usually a contributor; survey
Use of post-page recovery day	<50% of eligible	Engineers don't feel permission to take it; manager culture issue

The HRBP's playbook

Setting up on-call comp policy

1
1. Audit current state
Pull rotation data from PagerDuty/Opsgenie. How many people per rotation? Pages per week, distribution by hour, % off-hours. Talk to 5–10 engineers across teams about subjective burden.
2
2. Map legal exposure
List jurisdictions where engineers are on call. Get employment counsel input on each. Document the legal floor.
3
3. Choose a model + budget it
Most companies arriving at a real on-call comp policy land on flat stipend + productization. Budget impact: typically 0.3–1.5% of engineering payroll. Get CFO buy-in before announcing.
4
4. Document and announce
1-page policy: rotation rules, comp model, post-page recovery, escalation. Announce 4+ weeks before effective date with engineering leadership co-sign.
5
5. Audit annually
Per-team page rates, rotation fairness, comp policy uptake (e.g., are people actually taking their post-page recovery day?). Adjust policy based on data, not anecdotes.

FAQ

Frequently asked questions

Can we just declare engineers exempt and avoid on-call pay?

In the US, exempt-status alone doesn't eliminate retention risk and in CA may not eliminate legal risk. In the EU, exempt-status is less protective. The trend is toward explicit on-call compensation.

What about engineers who volunteer for extra on-call?