On-Call Policy: Expectations, Compensation, and Boundaries
Unwritten on-call is a machine for manufacturing resentment. The manager privately assumes "on-call" means answering anything, instantly, around the clock, because that is what the last place did. The admin privately assumes it means best effort for true emergencies, because that is only reasonable. Both assumptions feel so obvious that neither is ever said out loud. Then they collide at 2 AM, mid-outage, in the worst possible format: an argument about what the job is. Two exhausted people hold it while something is on fire. Nobody wins that argument. I have had it, and I can report that the relationship loses.
The fix is boring and powerful: write the deal down. A written on-call policy protects both sides. It protects the admin from infinitely elastic expectations, unpaid invisible labor, and the guilt-powered week that never ends. It protects the manager and the business from silent burnout and surprise resignations. It also protects them from discovering during an outage that everyone had a different understanding of who owed what. And it gives a new hire an honest answer to "what am I signing up for?" That is a fairness issue and a hiring advantage at the same time.
One caveat, and it matters: employment rules, working-time regulations, and contracts vary a great deal by country, industry, and agreement. And this guide deliberately stays out of all of that. What follows are the common approaches and the questions to settle. Run the resulting draft past whoever owns HR and legal questions in your organization before anyone signs it.
In this guide
- What writing it down actually changes
- Section 1: Scope — what on-call covers
- Section 2: Response expectations
- Section 3: Recognition and compensation
- Section 4: Boundaries
- Section 5: When a page goes unanswered
- Negotiating it — from either chair
- Keeping the policy alive
- The policy skeleton
What writing it down actually changes
A policy does not need to be long — two pages is plenty for a five-person team — and it does not need legalese. It needs to answer five questions plainly. A new hire, a manager, and a spouse should all be able to read it and reach the same understanding. What does on-call cover? How fast must someone respond? What does the person get for carrying it? Where are the limits? And what happens when a page goes unanswered? Those five questions are the five sections below. For each, this guide gives the questions to settle and the approaches teams commonly land on; the answers are yours to negotiate.
Why so much emphasis on written? Because unwritten norms are set by the most anxious moment, not the most reasonable one. The one time the admin answered a routine request at 11 PM becomes the precedent everyone silently expects. Writing the deal down replaces precedent-by-accident with agreement-on-purpose. It converts future disputes from "you always" and "you never" into "let's look at what we agreed and whether it still fits."
As for form: plain language beats formal language everywhere in this document. Store it where the on-call kit lives, so the person on shift is never more than a click from the deal they are working under. Date it, because you will revise it. Whether it becomes a formal HR document or stays a team working agreement is one of those local questions. But even the informal version should be seen and blessed by the manager. That is because a "policy" only the team has agreed to binds nobody when it is tested.
At Harbourview Clinic the on-call arrangement was two admins and a line in the job description that said "some out-of-hours work." The compensation conversation was put on the calendar three times in two years and moved three times. It moved once for a budget cycle, once for a restructure, and once because the pager went off during it. Time off in lieu lived in one admin's notebook and had reached several weeks before anyone with the authority to grant it had seen the number. What finally worked was a one-page draft brought to the fourth meeting, edited in forty minutes, and signed the same week. The notebook is now a shared table the manager can see, and the meeting has not moved since.
Section 1: Scope — what on-call covers
Scope is the pierce-list question: which conditions are serious enough to interrupt a human off-hours? The policy does not need to enumerate every alert. It points at the alert tiers your monitoring already implements, typically something like "confirmed outage of a business-critical system, security emergencies, and total loss of the internet link page the on-call person; everything else waits for business hours." Designing those tiers well is its own discipline (see Alert Fatigue). The policy simply makes them contractual: the pager tier is the deal.
Just as important is what on-call explicitly does not cover, because the exclusion list is what makes the rest credible. Common exclusions include routine user requests ("my mouse is broken", password resets that can wait) and project work. They also include non-critical tickets regardless of who is asking, and new-request intake generally. An executive's forgotten password at 10 PM is not an outage. A policy that says so in writing spares the on-call person from having to say it personally at 10 PM. There are questions to settle. Which systems count as business-critical (name them)? Do security alerts have their own faster lane? Who may declare an emergency outside the list (usually: the manager, knowingly spending a favor, not a loophole)?
Sample wording, to show how plain this can be: "On-call covers confirmed outages and severe degradation of the systems listed in Appendix A, suspected security incidents, and total loss of internet connectivity. It does not cover requests, questions, or non-critical issues, whoever raises them; those are handled next business day. The manager may ask for out-of-scope help in a genuine exception, and it counts as a favor, not a precedent." Four sentences, and the scope is no longer a matter of memory.
Section 2: Response expectations
This is where honesty matters most, because dishonest numbers get written by optimism and paid for by humans. The workable pattern is an acknowledge-within target: from page to a human saying "I have it." Commonly, it is in the 15-to-30-minute range for critical pages, longer or next-business-day for lower tiers. Targets like that are ones a person with a life can meet. They allow for a shower, a school run, being in the garden without a phone in hand. "Immediate response" is not a target, it is a wish. Nobody can guarantee it, so writing it down guarantees only that the policy will be broken in its first week.
Say plainly, in the policy's own words: response is not resolution. The commitment is that a qualified human engages promptly and works the problem, escalating for help when needed. It is not that any particular problem is fixed within any particular time. Small teams cannot promise fix times at 3 AM, and pretending otherwise sets the on-call person up to fail on their first hard night.
Reachability expectations belong here too, stated maturely and briefly. During a shift, stay within reach of your phone and within a reasonable distance of a laptop with a working connection. Keep yourself in a state where you could actually troubleshoot. That is the polite, sufficient way for a policy to address alcohol and similar topics without turning into a code of conduct. Adults handle this fine when the expectation is explicit and the shift has a visible end.
| Tier | Example | Acknowledge within | Then |
|---|---|---|---|
| Critical page | Business-critical system down, internet link dead | 15–30 min | Work it; escalate freely |
| Urgent, not critical | Degraded but functioning, redundancy lost | About an hour | Assess; fix or schedule |
| Morning queue | Disk trending full, one AP down overnight | Next business morning | Handle in working hours |
Section 3: Recognition and compensation
Carrying a pager has a cost even on nights when it never sounds. There is the planning around it and the drink not had. There is the pool not swum with the kids because the phone must stay dry and near. Sustainable on-call recognizes that cost explicitly. It is also the conversation most often scheduled, postponed, and scheduled again, which is why it belongs in a document rather than a meeting invitation. The common models, surveyed neutrally:
| Model | How it works | Tends to fit when | Watch out for |
|---|---|---|---|
| Stipend per shift | A flat amount for each week (or weekend) carried, paged or not | Pages are rare; the burden is mostly availability | Feels unfair when a shift turns into a firestorm — often paired with callout pay |
| Hourly / callout pay | Time actually worked on incidents is paid (often with a minimum per callout) | Pages are common enough that worked time is the real cost | Requires honest time logging; can accidentally reward noisy systems staying noisy |
| Time off in lieu | Off-hours incident time converts to time off; a rough night converts to a late start or a recovery day | Budgets are tight; flexibility is valued | Only real if the time can actually be taken — unspent lieu-time is a broken promise |
| Included in salary | On-call named in the role and reflected in overall pay | Small, senior teams with genuinely quiet pagers | Fairness erodes fast if page volume grows or falls unevenly across people |
Which model fits is a local question with three honest inputs. The first inputs are how often pages actually happen (measure, don't guess), and how critical the covered systems are to the business. The last is what the budget can genuinely sustain. Many teams combine models — a modest stipend for carrying the pager plus time off in lieu for nights that actually got rough. Whatever the model, one floor is worth defending as simply humane: a bad night buys recovery time. The person who was up from 2 to 5 AM does not present themselves at 9 AM out of pride. The policy says the late start is expected, not negotiated per incident.
Money is not the only recognition that matters, and the non-monetary kinds are free. They include schedule flexibility around shifts, and first pick of holiday weeks for whoever carried the worst ones. (The holiday-ledger mechanics are in Setting Up Your First On-Call Rotation.) Then there is visible credit. The manager who tells the company "the reason nobody noticed Saturday's outage is that Dana fixed it at 6 AM" is paying a real currency.
Section 4: Boundaries
Boundaries are where a policy shows whether it was written for the systems or for the people. The ones worth writing down:
- Swap rights. Shifts may be swapped freely between qualified people without manager approval, provided the swap is recorded in the authoritative schedule. Life happens; the policy should make accommodating it frictionless.
- Leave is leave. On-call is never scheduled during booked vacation, and booked vacation is never treated as silently-still-reachable. The deeper principle is that time off only restores people when it is engineered to be real. It is the same one behind how a solo admin takes a real vacation. It applies with less excuse on a team, because a team has other people.
- Load limits. Settle the maximum consecutive on-call weeks (common answer: one — no back-to-back weeks except by the person's own swap choice). Also settle the minimum gap between shifts the schedule should aim for. A rotation that math cannot support — two people, alternating forever — is a staffing finding, and the policy is the place it becomes visible.
- Life accommodations. Pregnancy, health conditions, caring duties, bereavement, and similar seasons of life change what a person can carry at night. The policy does not need to enumerate outcomes. It needs to promise the conversation: that accommodations will be arranged with decency and without career penalty. It must also name whom to talk to. Teams that handle this well keep people for a decade. The details are questions for management and HR, which is exactly where the policy should send them.
- A noise ceiling. Here is the quiet clause worth adding. If the after-hours page count stays above an agreed level for a sustained stretch, the team owes itself an alert review before anyone owes another rough week. This turns "the pager is out of control" from a complaint into a tripwire with a defined response. That response is the reduction work described in Humane On-Call.
Tip: add one small boundary that costs nothing and defuses the most common flashpoint. A page during dinner does not require an apology, and a missed non-critical notification never requires one. The policy's tone teaches people how to treat each other at 2 AM.
Section 5: When a page goes unanswered
Every policy needs to say what happens when the acknowledge-within target is missed — and the answer that works is: the escalation path fires. The backup gets paged, then the wider chain, exactly as designed in Escalation Paths That Work When Tier One Is Asleep. The policy's job is to frame that machinery correctly. Escalation is the safety net that protects the business from the fact that humans sleep, get sick, and drive through tunnels. It is not a disciplinary trap, and a fired escalation timer is not an accusation.
Write the distinction explicitly. A missed page is a systems event. It gets reviewed like any incident. Did the escalation work? Was the channel loud enough? Was the timer sane? A pattern of missed pages is a different thing. It calls for a private, human conversation about whether the load, the channels, or the life circumstances need to change. It is not a public post-mortem. Policies that skip this distinction teach people to fear the pager. People who fear the pager start gaming it: acking without engaging, muting selectively, quietly resenting. The no-shame architecture is load-bearing here just as it is in the escalation design itself.
Negotiating it — from either chair
If you are the admin: bring data, not adjectives. "On-call is brutal" invites debate; "we averaged eleven after-hours pages per week last quarter, four of them between midnight and six" invites arithmetic. Your monitoring and alerting history already holds these numbers: page counts, after-hours share, worst weeks. And Alert Fatigue covers how to pull that picture together. Then use the draft-policy move: do not walk in asking "can we have an on-call policy?" Walk in with a one-page draft and ask "can we edit this together?" A blank page invites deferral; a draft invites line edits, and most of your reasonable defaults will survive unchallenged. It is the move that finally worked at Harbourview, on the fourth attempt.
If you are the manager: the case for engaging seriously is retention arithmetic. The person carrying your pager holds an irreplaceable map of your systems in their head. Burnout does not announce itself. It just resigns one day. And everything in the solo-admin burnout playbook scales up to small teams. It is the same quiet overload, multiplied by however many people are pretending to be fine. A fair stipend and a written recovery-time rule cost a rounding error against one departure. That departure costs you the recruiting, the months of ramp-up, and the outages during the gap. Paying attention here is not generosity; it is maintenance on your most critical system.
Keeping the policy alive
A policy written once and framed on the wall drifts out of truth within a year. Systems change, page volume changes, and people change. Book an annual review. Many teams pin it to the same meeting that reviews the rotation itself. Walk a short, fixed agenda against the year's actual numbers:
- Were the acknowledge targets met — and where they were missed, was it the humans, the channels, or the timers that were wrong?
- Is the page dose still sane, or has a noisy new system quietly doubled it since the deal was struck?
- Does the swap log show trades flowing both ways, or one person absorbing everyone's weekends?
- Does the holiday ledger balance across the year, or is it accumulating quiet unfairness?
- Has time off in lieu actually been taken, or is it piling up as a debt the business never intends to pay?
Twenty minutes with those five questions keeps the policy a living agreement — the same pilot-and-adjust spirit the rotation was born with. And when reality and document disagree, change whichever one is wrong; sometimes the honest fix is editing the target, not exhorting the people.
The policy skeleton
Here is the whole document as a fill-in exercise. Copy the table, replace the third column, delete what does not apply, and you have a draft worth bringing to the meeting:
| Section | Questions to settle | Your answer |
|---|---|---|
| Scope | Which systems are critical? What pages a human off-hours? What is explicitly excluded? Who may override? | |
| Response | Acknowledge-within target per tier? What counts as acknowledged? Reachability expectations during shifts? | |
| Recognition | Stipend, callout pay, time off in lieu, salary inclusion — or which mix? What does a bad night buy? | |
| Boundaries | Swap rules? Leave interaction? Max consecutive shifts? Accommodation route for health and life events? | |
| Escalation | What fires when a page is unacknowledged? How are missed pages reviewed — and how are patterns handled? | |
| Review | When is the annual review? Which numbers are examined? Who can propose changes mid-year? |
Where to go next
A written policy turns on-call from an ambient guilt-field into a deal. That deal has clear scope, honest response targets, real recognition, and defended boundaries. It has an escalation net that catches people instead of blaming them. It works best alongside its siblings. One is the rotation that supplies the schedule (Setting Up Your First On-Call Rotation). Another is the escalation path the policy leans on (Escalation Paths). The last is the design work that keeps the deal cheap to honor by keeping nights quiet (Humane On-Call). The rest of Escalation & On-Call for Small Teams fills in the picture. The resentment machine runs on silence. Write the deal down, and it has nothing left to run on.
Put this into practice with SysPeep Monitor
Every negotiation in this guide goes better with real numbers behind it. SysPeep Monitor watches your network and servers over SNMP, WMI, and SSH. Its thresholds, dependencies, and alert rules are designed to keep pages rare and real. Its dashboards and reports give you an honest picture of what actually fires, and when, to bring to the policy conversation. The free trial installs on any Windows machine.
Download Free TrialLearn About SysPeep