Build multi-step escalation policies so alerts reach the right person, in the right channel, at the right time.
Escalation Policies
An escalation policy is an ordered list of steps. When an incident fires, LemWatch runs step 1, waits, runs step 2, waits, and keeps going until somebody acknowledges the incident or the policy runs out of steps.
When to use one
You want different people paged at minute 0, minute 5, and minute 15
You want email first, Slack second, SMS/webhook last
You run on-call rotations and need a fallback if the primary doesn't ack
Build a policy
Open Settings → Notifications → Escalation Policies
Click New policy, give it a name (e.g. "Production on-call")
Click Add step and fill in:
Step order — the position in the sequence (1, 2, 3 …)
Wait minutes — how long to wait before this step fires (step 1 is usually 0)
Channel — email, Slack, Discord, webhook, etc.
Recipient — an email address, or a webhook URL (see Step channels)
Save and attach the policy to a site under Edit Site → Notifications → Escalation policy
Step ordering & uniqueness
Each step needs a uniquestep_order within the policy. The editor auto-suggests the next number (max existing + 1) so adding a step after deleting one in the middle still works. If you manually enter a number that's already taken, the editor shows "Step #N already exists in this policy" instead of a raw database error.
Acknowledging stops escalation
The moment any teammate clicks Acknowledge on the incident, all remaining steps are skipped. The notification engine now also skips already-acked incidents at query time, so escalation work doesn't pile up. See Acknowledging incidents.
Editing an existing policy
Reorder steps by editing step_order values
Change channel or recipient on any step
Delete a step — the next save renumbers cleanly
Delete the whole policy — sites that referenced it fall back to their default channels