Q: How do I build an escalation policy that wakes the right person and not everyone?

Q: How do I build an escalation policy that wakes the right person and not everyone?

A: Good escalation is layered, with time gaps that give a fix a chance before it widens.

— Tier 1: alert the primary on-call only. Give them 5-10 minutes to acknowledge.
— Tier 2: no ack? Auto-notify the secondary on-call.
— Tier 3: still nothing after another 10-15 minutes? Now the team lead/manager.
— Reserve a 'critical' path that skips straight to multiple people for total-outage signals only.
— Always have a fallback contact so an alert never lands in a dead channel.

The key: severity decides the starting tier. A slow-response warning starts at tier 1 and stops. A full outage can start wider.

Follow-up: test the chain quarterly by letting a low-priority alert escalate on purpose. Untested escalation always has one broken link.

Got a question? Drop it in the comments.