On-Call Compensation Models: How Engineering Teams Are Handling It in 2026

On-call compensation arguments rarely start as compensation arguments. They start as someone quietly declining to join the primary rotation, or as a strong engineer interviewing out after a quarter of bad nights. By the time the stipend number reaches a spreadsheet, the trust problem is already there.

For engineering managers and HR partners at 20 to 100 person companies, the job is to pick a model that feels fair, survives a bad incident week, and does not create strange incentives. Industry practice is inconsistent. That does not mean you have to invent the wheel alone.

Why this is a retention issue

Pay is one signal. Fairness perception is another. Willingness to take primary is a third. If the same two people absorb every Sev-1 because everyone else treats on-call as optional pain, your compensation model has already failed even if the stipend looks market-average on paper.

Compensation also interacts with alert quality. Paying people more to endure a flappy pager is a short-term patch on a reliability problem. The money matters. The load shape matters more over a year.

The five main models

Flat weekly stipend. A fixed amount for each week of primary (and sometimes secondary) duty, regardless of pages. Simple to administer and common. It does not track burden, so a quiet week and a brutal week pay the same.

Per-incident payout. Compensation scales with pages or incidents. Better alignment with actual pain, worse incentive design if people can influence what gets declared an incident. Also creates payroll noise and disputes about incident boundaries.

Time off in lieu (TOIL). Hours or days back after overnight work or heavy incident load. Preferred in some regions and cultures. Requires tracking discipline and can collide with delivery schedules if many people bank time at once.

Hybrid stipend plus threshold. A base amount for carrying the pager, plus extra payout when pages or night incidents cross a threshold. Captures quiet-week fairness and bad-week burden without paying for every trivial alert.

Inclusive compensation. On-call is treated as part of the role and reflected in base salary or level expectations, with no separate payment. Clean accounting, weak visibility, and frequent fairness fights unless expectations and rotation size are truly shared.

Tradeoffs in practice

Flat stipends win on clarity and lose on equity during incident spikes. They work best when alert volume is moderate and rotations are even.

Per-incident models win on perceived fairness during chaos and lose when definitions get gamed or when a platform outage generates fifty pages that are really one event. If you use this model, define what counts before the outage, not during payroll.

TOIL respects recovery and can be culturally expected, but unrecovered TOIL becomes a hidden debt. If people never actually take the time, you have unpaid overtime with better branding.

Hybrids are operationally honest for many mid-size teams. They cost more process to run and still need clean incident definitions.

Inclusive models work in teams where nearly everyone rotates and levels already price in the duty. They fail in teams where only a subset carries the pager while "inclusive" pay is company-wide.

What is common in APAC and Singapore

US and many EU companies lean toward stipends or per-incident cash. APAC practice is more mixed. In Singapore and parts of Southeast Asia, TOIL and inclusive compensation show up often, shaped by labor norms, PME employment patterns, and a preference for time recovery over small cash add-ons.

That does not mean cash stipends are rare in the region. Growing startups that hire globally often import US-style stipends, then discover local expectations differ. The workable approach is to ask the team what feels fair, check legal constraints, and write the policy down before the next rotation starts.

Singapore Employment Act context

Singapore employment rules treat different categories of employees differently, and PME-category engineers are not handled like every overtime-eligible role. Whether on-call standby counts as working time, and how night response should be compensated, depends on contract terms, company policy, and advice you should not take from a blog post alone.

Treat the following as operational guidance, not legal advice. Verify Singapore labor details with qualified HR or legal counsel before you publish a policy. Document what standby means, what response expectations are, how TOIL is requested, and how exceptions work for prolonged incidents.

What engineers care about beyond money

Ask a tired primary what would make next quarter better and the first answers are often not a larger stipend. They want fewer junk pages, a secondary who actually backs them up, permission to take the morning after a hard night, and a rotation that does not punish people for taking leave.

Predictability beats generosity. A moderate stipend with sane alert volume retains people longer than a large stipend attached to a pager that never sleeps.

Recovery time is part of compensation. If your policy pays cash and still expects standup attendance after an overnight restore, you have not finished the design.

Where to start for the first time

For a team setting this up for the first time, start simple. Use a flat weekly stipend or a clear TOIL rule for primary, define secondary expectations, and measure pages per shift for a quarter. Publish the policy, including how incidents are counted if money depends on them.

Revisit after you have data. If one week is routinely twice as hard as the median, move toward a hybrid threshold. If almost nobody takes TOIL, either force the recovery or convert unused recovery into another form of compensation. If only senior people can handle primary, fix enablement before you throw money at a bus-factor problem.

Write the human rules next to the scheduling rules. Compensation is how you prove the rotation is a shared duty, not a volunteer punishment. Get that right and the spreadsheet debate gets quieter.