# Escalation policy: [Name]

> Acknowledgement stops the chain. Unacked pages advance after the wait.

## Policy card
| Field | Value |
|-------|-------|
| Policy key | [stable-key] |
| Owner | [Team] |
| Reviewed | YYYY-MM-DD |

## Auto-routing (optional)
**Triggers:** Incident created · Incident updated

| Match | Value |
|-------|-------|
| Severity | Critical, High |
| Components | [API, Payments] |
| Tags | production |
| Teams | [Platform] |

*Leave empty for a catch-all. Split policies by severity to reduce alert fatigue.*

## Steps
| # | Target | Strategy | Wait | Notes |
|---|--------|----------|------|-------|
| 1 | Schedule: [Primary] | Notify one | 5 min | Current on-call |
| 2 | Schedule: [Secondary] | Notify one | 5 min | Backup |
| 3 | Team: [Platform] | Notify all | 10 min | Any responder |
| 4 | User: [Director] | Notify one | — | Final human |

**Targets:** On-call schedule · Team · User

**Strategy:** Notify one = round-robin · Notify all = parallel (first ack wins)

---

## Reference designs

**Production Critical**
Primary schedule (5m) → Secondary schedule (5m) → Team notify-all (10m)

**Business-hours Medium**
Match: Medium + tag `business-hours` → Primary schedule (15m) → stop

## Mid-incident
Use **Bring in** to add a user, team, or schedule without editing this policy.

## Test before prod
- [ ] Staging incident triggers correct policy
- [ ] Shift verification on for schedules
- [ ] Critical matches here; Low does not
- [ ] Waits feel right with real ack flow

## Avoid
Single step, no timeout · Same policy for every severity · Final step is one person with no backup