An on-call escalation policy is the rule that decides who gets alerted about an incident, in what order, and how long to wait before trying the next person. It runs until someone acknowledges. A typical one: call the primary engineer now, call the backup after 5 minutes, call the team lead after 10 more, then start over. Without a policy, an alert goes to whoever happens to see it, which at night usually means nobody.
This post explains how escalation policies work, using Edesy OnCall as the example, and shows three policies that suit small teams.
The parts of a policy
| Part | What it means | Edesy OnCall limits |
|---|---|---|
| Step | One rung of the ladder: who to alert, and how | 1 to 10 steps |
| Contacts | The people alerted in that step, all at once | One or more per step |
| Channels | How each contact is alerted | Call, email, or both |
| Wait | Minutes before this step fires, after the previous one | 0 to 1,440 minutes (1 minute minimum after step 1) |
| Repeat | How many times the whole ladder starts over | 0 to 10 times, 5-minute pause between rounds |
| Acknowledge | The response that stops the ladder | Press 1 on the call, email button, or dashboard |
Each monitor points at one policy. Several monitors can share a policy, which is the usual set-up: one "Production" policy for every critical check.
What happens during an incident
Take this policy: step 1 calls Asha (wait 0), step 2 calls Ravi and emails both (wait 5 minutes), repeat 1.
- 02:14 The checkout API fails twice in a row. An incident opens.
- 02:14 Step 1: Asha's phone rings. "This is an alert from Acme. Checkout API is down since 2:14 AM. Reason: HTTP 503. Press 1 to acknowledge this alert. Press 2 to escalate it to the next person." She is asleep and doesn't answer.
- 02:19 Step 2: Ravi is called and both get an email with an acknowledge button.
- 02:20 Ravi presses 1. Escalation stops. The timeline records "Acknowledged by Ravi via phone (pressed 1)".
- 02:41 The fix is deployed, the next check passes, the incident resolves, and both get a "back up" email.
If Ravi had missed it too, the ladder would have started over at step 1 after a 5-minute pause, once (repeat 1). After that, the incident stays open on the dashboard but nobody else is called.
Pressing 2 is for "I can't take this one": the next step fires straight away instead of waiting.
Acknowledge is not resolve
These two get confused often:
- Acknowledged means a human has taken the incident. Calls stop. The problem may still be there.
- Resolved means the problem is gone: the check passed again, your monitoring tool sent "resolved", or someone pressed Resolve in the dashboard.
Separating them stops two people fixing the same thing, and stops the phone ringing while you're already working on it. See acknowledge and resolve.
Three example policies
1. Solo founder
| Step | Wait | Who | Channel |
|---|---|---|---|
| 1 | 0 min | Founder | Call + email |
| 2 | 10 min | Co-founder or a trusted freelancer | Call |
Repeat: 2. The Free plan's 2 contacts cover this.
2. Small dev team (3 to 5 people)
| Step | Wait | Who | Channel |
|---|---|---|---|
| 1 | 0 min | This week's on-call engineer | Call |
| 2 | 5 min | Second engineer | Call + email |
| 3 | 10 min | Engineering lead | Call |
Repeat: 1. Since OnCall has no rotations, update step 1 when the week changes. Pin it to your team calendar so it doesn't get forgotten.
3. Agency watching client sites
| Step | Wait | Who | Channel |
|---|---|---|---|
| 1 | 0 min | Developer on duty | Call |
| 2 | 5 min | Second developer | Call |
| 3 | 15 min | Account manager |
Use one policy per tier of client, or one per big client with their own contact in a later step. More in agency client site monitoring.
Mistakes that leave alerts unanswered
- One-person policies. A phone on silent, a flat battery or a flight means no response. Always add a second step.
- Unverified phones. OnCall skips calls to a phone number that hasn't been verified, which is done with a short call that reads out a 4-digit code. Verify everyone and send a test call. See contacts and verification.
- Waits that are too long. A 30-minute wait on step 1 means 30 minutes of downtime before the backup hears about it. Five minutes is a good default.
- Unknown caller ID. People ignore calls from numbers they don't recognise. Ask everyone to save the alert number and allow it through Do Not Disturb.
- No email address. If an alert call fails outright, OnCall emails the contact instead, but only if an email is saved.
- Empty wallet. Alert calls are billed per minute from your Edesy wallet. If it runs out, nobody gets phoned. Keep it topped up.
- Monitors with no policy. An incident opens but nobody is paged. Check every monitor has one.
Night-time and language
Each contact picks English or Hindi, and their calls are spoken in that language. Between 21:00 and 09:00 IST, a workspace on the standard calling window still gets the call, as a plain spoken alert without press 1; contacts acknowledge by email or in the dashboard. Edesy support can enable 24x7 press-1 alert calls. See night mode.
What OnCall doesn't do
OnCall has escalation policies, not schedules. There are no rotations, shift calendars, overrides or maintenance windows, and alerts go by phone call and email only, not SMS or chat. If you need rotation management, see our honest PagerDuty alternatives comparison.
Plans
Contacts per workspace: 2 on Free (Rs 0), 5 on Starter (Rs 499/month), 25 on Pro (Rs 1,499/month). Prices exclude 18% GST, and alert calls are billed per minute from the wallet. See pricing and the escalation policies doc.
FAQ
What is an on-call escalation policy?
It is an ordered list of who to alert about an incident and how long to wait before moving to the next person. It runs until someone acknowledges the alert or the ladder runs out.
How many steps should a policy have?
Two or three is enough for most small teams: the primary person, a backup, and a lead or founder as the last step. Edesy OnCall allows up to 10 steps and up to 10 repeats of the ladder.
Does acknowledging fix the incident?
No. Acknowledging only stops further calls and records who took it. The incident resolves when the check passes again, when your tool sends a resolved alert, or when someone resolves it in the dashboard.
Does Edesy OnCall support rotations?
No. OnCall has escalation policies but no on-call schedules or rotations. A policy calls the same people in the same order every time; edit it when your rota changes.
Build your first escalation policy free at oncall.edesy.in.