Skip to main content
A rule that fires at 3 a.m. is only as good as the path to a person who can act. Most missed pages come from the path, not the rule: an unverified phone number, a schedule with a gap on Sundays, an escalation policy that nobody created. Build the path once, then prove it with a rehearsal.

​
Decide who answers what

Before configuring anything, write down for each service: the team that owns it, who is on call first, who comes next, and which severities justify a phone call at night. Group the people of each owning team in a team; schedules, escalation targets, and fallback routes all refer to teams.

​
Give every person a verified way to be reached

  1. An administrator configures the connectors under Settings → Notify → Connectors: an IM app, email, and phone and SMS for critical pages. Send test on each one.
  2. Each member binds personal methods under Account settings → Notify and sends a test to each.
  3. An administrator with org.members.manage Verifies each method after confirming that it belongs to the person. Unverified methods never deliver.
  4. Each member orders the methods for Alert events, with waits that escalate the channel: for example IM at once, SMS after 3 minutes, a phone call after 5 minutes. Later steps stop as soon as someone acknowledges.
Quiet hours stop notifications for a person, and they are not sent later. Turn on Allow critical alerts to bypass quiet hours for everyone on a schedule that receives critical pages.

​
Build schedules without gaps

Create one on-call schedule per team, in the timezone of the team:
  • a Weekly rotation with a handoff in working hours, when both people are awake;
  • for follow-the-sun or business-hours coverage, a first rotation restricted to working hours and a second rotation below it for the remaining hours;
  • every member, in the order of the rotation, and at least two people per schedule.
Check Next 7 days before saving, and look for Coverage gap in the schedule list and on Home. Record leave and swaps as temporary overrides ahead of time, and Export calendar so the shifts appear in the calendar of each person.

​
Create the default escalation policy first

Rules saved in the rule workbench use the policy named default; without it, their incidents page nobody. See The default policy. Turn on Repeat the whole policy until acknowledged with a small Max loops for critical services. Create further policies for teams with different needs, and point rules at them through the API with escalation_policy_id.

​
Add safety nets

  • Fallback routes deliver to a team channel or an organization address when no personal method of a recipient works.
  • Notify policies decide the template and the connectors of each event; the live preview shows who would receive an event without sending it.
  • The Delivery log shows every attempt, so a failing connector is found before an incident needs it.

​
Rehearse

Run a drill in working hours, and repeat it after every larger change:
  1. Fire a test rule on purpose with Trigger a rule.
  2. Check that the person on call receives the page on each channel, in the expected order.
  3. Wait without acknowledging, and check that the escalation reaches step 2.
  4. Acknowledge, and check that later steps and channels stop.
  5. Resolve, and review the Delivery log for failures.

​
Acceptance criteria

  • Every person on a schedule has at least two verified methods, one of them by phone for critical pages.
  • No schedule shows a coverage gap in the next 7 days.
  • A policy named default exists, and each step has a person or a schedule.
  • Fallback routes exist for every team that receives pages.
  • The last drill reached the person on call, escalated as expected, and stopped on acknowledgement.
Last modified on October 26, 2026