Blog

Escalation policy example: primary, secondary and manager chains with timeouts, acknowledgement and night rules

Copy-ready escalation chains for a small team: who gets paged first, how long to wait before the next level, what acknowledgement means and how nights differ from days.

A practical escalation policy example for a small team has three levels. The primary on-call engineer is paged first. If nobody acknowledges within 5 minutes, the secondary is paged. If there is still no acknowledgement after another 10 minutes, the engineering lead or founder is paged and the page repeats until someone owns it. Run one policy for business hours and a separate, stricter one for nights that only fires on critical alerts and allows a little more time to wake up.

This article gives you concrete chains you can copy, explains how to choose the timeouts between levels, defines what acknowledgement should actually mean and shows how to adapt the policy to teams of two to six people. Who is on call in a given week is a separate question, covered in our guide to on-call rotation models for small teams. Here we focus on what happens after the first page goes unanswered.

What an escalation policy has to define

An escalation policy is a set of rules that decides who gets notified next when an alert is not handled. It is different from a schedule: the schedule says who is primary this week, the policy says what happens when that person does not respond.

Every usable policy answers six questions:

  • Levels: the ordered list of targets, usually primary, secondary and a final fallback.
  • Targets: a schedule (whoever is on call) or a named person. Prefer schedules for levels 1 and 2 so the policy never needs editing when the rotation changes.
  • Channels: how each level is reached. Higher levels should use louder channels.
  • Timeouts: how long to wait for acknowledgement before moving to the next level.
  • Acknowledgement: what action stops the escalation and who can perform it.
  • Final behaviour: what happens at the last level. Repeat, widen to the whole team, or both.

If any of these is missing, the policy has a gap that will show up at 3 a.m.

An escalation policy example for a three-person team

The following is pseudo-configuration, not the syntax of a specific tool. It describes a business-hours policy for a team of three engineers plus a lead.

policy: prod-critical-business-hours
applies: Mon-Fri 09:00-18:00 (team timezone)
severity: critical, high

levels:
  - level: 1
    target: schedule:primary
    channels: [push, slack-dm]
    escalate_if_not_acked_after: 5m

  - level: 2
    target: schedule:secondary
    channels: [push, sms]
    escalate_if_not_acked_after: 10m

  - level: 3
    target: user:engineering-lead
    channels: [sms, push]
    repeat_every: 10m
    max_repeats: 3
    then: notify channel:#incidents

Level 1 uses push and a direct message because during the day the primary is at a desk. Level 2 adds SMS because the secondary is not expecting to be paged and may be in a meeting. Level 3 goes to a named person, not a schedule, because the lead is the person who can reassign work or pull someone out of a meeting.

The final step posts to a shared incident channel. At that point the alert has been ignored by three people for roughly 25 minutes, and visibility matters more than noise.

What this looks like during a real alert

14:02  API health check fails from 3 regions, alert fires
14:02  Level 1: primary (Anna) gets push + Slack DM
14:07  No ack. Level 2: secondary (Ben) gets push + SMS
14:09  Ben acknowledges. Escalation stops.
14:09  Anna is informed that Ben owns the alert
14:31  Monitor recovers, alert resolves automatically

Note the last two lines. The person who missed the page should be told that someone else took over, otherwise two people may start working on the same problem in parallel once Anna returns to her phone.

How long should the timeout between escalation levels be?

Timeouts are a trade-off between giving people enough time to respond and keeping total time to engagement acceptable. Work backwards from your reliability target: if a critical outage must have someone actively working on it within 15 minutes, the sum of your level 1 and level 2 timeouts must stay below that.

Reasonable starting values for a small team:

  • Level 1 to level 2, daytime: 5 minutes. Someone at a desk can read a push notification and acknowledge in under a minute.
  • Level 1 to level 2, night: 10 minutes. Waking up, finding the phone and understanding the alert takes longer, and escalating too fast wakes two people for every page.
  • Level 2 to level 3: 10 minutes in both policies. By then the alert has already been missed once, so the secondary deserves time to look properly.
  • Repeat at the last level: every 10 minutes, capped at three or four repeats.

Avoid timeouts under 3 minutes. They do not give anyone a realistic chance to respond and effectively turn a sequential policy into paging everyone at once. Also avoid timeouts above 15 minutes for critical alerts, because a single missed page then costs more downtime than most small SaaS products can afford.

What acknowledgement should mean

Acknowledgement is the action that stops escalation. Teams often treat it as a reflex, tapping the button to silence the phone. That makes the policy useless, because the alert is now owned by someone who is not working on it.

An acknowledgement should mean exactly one thing: I own this alert and I am looking at it now. Write that sentence into your on-call guidelines.

Some rules that keep acknowledgement honest:

  • Acknowledge is not resolve. The alert stays open until the underlying check recovers or someone resolves it deliberately.
  • Add an acknowledgement timeout. If an acknowledged critical alert is still open after 30 to 45 minutes without an update, re-notify the owner and optionally the secondary. This catches people who acknowledged and fell asleep again.
  • Make acknowledgement explicit. An emoji reaction in a chat channel is not acknowledgement. Use a button, a reply or an app action that the alerting system records with a name and timestamp.
  • Can't handle it? Reassign, don't ignore. If the primary is on a train without a laptop, they should hand the alert to the secondary immediately instead of letting it time out.

Business hours vs nights: two policies, not one

A single policy for all hours either pages too loudly during the day or too softly at night. Split it.

policy: prod-critical-night
applies: Mon-Fri 18:00-09:00, Sat-Sun all day
severity: critical only

levels:
  - level: 1
    target: schedule:primary
    channels: [sms, push-critical]
    escalate_if_not_acked_after: 10m

  - level: 2
    target: schedule:secondary
    channels: [sms, push-critical]
    escalate_if_not_acked_after: 10m

  - level: 3
    target: user:engineering-lead
    channels: [sms, push-critical]
    repeat_every: 10m
    max_repeats: 4

The differences from the daytime policy are deliberate:

  • Only critical severity. A slow endpoint or a certificate that expires in 14 days can wait until morning. At night those alerts go to a channel and are reviewed at 9:00.
  • Louder channels from level 1. Slack DMs do not wake anyone. Use SMS or a push app that is allowed to bypass the phone's silent mode.
  • Longer first timeout. 10 minutes instead of 5, for the wake-up time.
  • No shared channel at the end. Posting in a team channel at 3 a.m. helps nobody. Repeat to the last level instead.

Deciding which alerts are critical enough for nights is the hard part. If your night policy fires more than occasionally, the problem is usually alert quality, not the escalation chain. Our article on cutting alert fatigue with actionable pages covers how to filter that list.

Adapting the chain to your team size

The three-level example assumes enough people to have a distinct primary, secondary and lead. Smaller and larger teams need small adjustments.

Two engineers

Level 1 is the primary, level 2 is the other engineer, level 3 is the founder or whoever owns the business risk. Accept that the secondary is effectively always on call and keep the night policy very strict so that this does not burn people out.

Four to six engineers

Keep primary and secondary as schedules, and make level 3 a small schedule of leads instead of one named person, so a single manager is not the fallback for every night of the year. Consider separate policies per service if teams own different systems.

The manager question

The engineering lead belongs at level 3, not level 2. Putting a manager directly after the primary turns every missed page into a management event and discourages honest reporting of missed alerts. At level 3, the manager's job is not to fix the outage but to find someone who can.

How to set up an escalation policy step by step

  1. List the monitors that justify waking someone up and tag them as critical. Everything else gets high or low severity.
  2. Create two schedules, primary and secondary, with the secondary offset by one rotation so that last week's primary is this week's secondary.
  3. Write the business-hours policy with 5 and 10 minute timeouts, schedules as targets for levels 1 and 2 and a named fallback at level 3.
  4. Write the night policy with critical severity only, louder channels and a 10 minute first timeout.
  5. Define acknowledgement in one sentence and add an acknowledgement timeout of 30 to 45 minutes for critical alerts.
  6. Configure every person's phone so the paging app or SMS sender can break through do-not-disturb mode.
  7. Trigger a test alert for each policy and let it run through every level without acknowledging. Check timings and channels against your config.
  8. Repeat the test after every schedule change and at least once a quarter.

Step 7 is the one most teams skip. A policy that has never been tested end to end usually fails at the last level, because a phone number is outdated or a notification is filtered.

Common mistakes in escalation policies

  • Paging the whole team at level 1. It feels safe but creates diffusion of responsibility: everyone assumes someone else is on it.
  • Using the same channel at every level. If level 1 failed because Slack notifications are muted, level 2 on Slack will fail too.
  • No final step. A chain that ends after level 2 means an unacknowledged alert simply goes quiet.
  • The same person at two levels. During a rotation handover the primary can end up as their own secondary. Check for overlaps.
  • Escalating warnings. Only alerts that need a human now belong in an escalation chain.

Implementing the chain with m0nitor

In m0nitor, on-call schedules define rotations, temporary overrides for holidays and swaps, and do-not-disturb hours, and alerts are routed to whoever is on call. Alert workflows automate reactions to monitor events, such as notifying a channel, creating an incident or calling a webhook, which covers the routing of business-hours and night alerts to different targets.

Alerts can go to email, Slack, Discord, Microsoft Teams, Mattermost, Telegram, Pushover, SMS, PagerDuty and ilert, as listed on the alert channels and integrations page. If your team already runs timed multi-level escalation with acknowledgement in a dedicated paging tool, forwarding alerts there is one option. The policy design in this article applies regardless of which tool executes it.

FAQ

What is an escalation policy in on-call?

An escalation policy is an ordered set of rules that decides who is notified when an alert is not acknowledged in time. It defines the levels, the people or schedules at each level, the notification channels and the timeouts before moving to the next level.

How many levels should an escalation policy have?

For small teams, three levels are enough: primary on-call, secondary on-call and a final fallback such as the engineering lead or founder, with repeated notifications at the last level. More levels mostly add delay without adding people who can actually respond.

What is the difference between an on-call schedule and an escalation policy?

The schedule determines who holds the primary and secondary role at a given time. The escalation policy determines what happens when the person currently on call does not respond. Policies should reference schedules rather than names so they stay correct when the rotation changes.

What is a good escalation timeout?

Start with 5 minutes from level 1 to level 2 during business hours, 10 minutes at night, and 10 minutes from level 2 to level 3. Adjust so that the total time until someone is actively working on a critical alert stays within what your reliability target allows.

Product

  • Pricing
  • Features
  • Integrations
  • Security
  • Free network check

Features

  • Uptime monitoring
  • Status pages
  • Alert workflows
  • Incident management
  • Dependencies
  • AI assistant
  • Global infrastructure
  • On-call schedules
  • Enterprise SSO

Use cases

  • Websites for free
  • Cron jobs
  • SSL certificates
  • Databases
  • IP blacklists
  • GDPR compliance

Resources

  • What is a status page?
  • Documentation
  • API
  • MCP server
  • Monitoring glossary
  • Free tools
  • SLA calculator
  • SSL checker
  • DNS lookup
  • Blog
  • Changelog
  • About
  • Contact

Compare

  • Free uptime monitoring
  • UptimeRobot
  • Better Stack
  • Pingdom
  • StatusCake
  • Uptime Kuma
  • Hyperping
  • Atlassian Statuspage
  • Instatus
  • Cachet
  • Freshping