Glossary
Escalation policy
An escalation policy is the set of rules that decides who is paged next, how, and after how long, when an alert is not acknowledged.
Also called Escalation rules, Escalation chain.
An escalation policy answers one question: what happens when the first person paged does not respond? It is an ordered list of steps. The first step pages someone straight away, and each later step fires only if nobody has acknowledged by the time it is due. An acknowledgement at any point stops the rest of the policy, so a page that is answered quickly never reaches the later steps.
The waits between steps are the part worth thinking about. Too short, and the whole team is woken for something the primary was already handling but had not yet acknowledged. Too long, and a missed page sits unanswered while the outage runs on. A few minutes between the first steps suits most teams, and higher severities deserve shorter waits than low ones rather than one uniform delay.
Escalation is also where channel choice does real work. A push notification is easy to sleep through and a phone call is not, which is why a call often comes at the second step rather than the first. Moving to a louder channel, or to a different person, is what makes each step add something the step before it did not.
One rule is not optional: the policy must end somewhere definite. A chain that runs out of steps and stops silently leaves an incident that nobody is handling and nobody knows is unhandled. That is worse than having no policy, because everyone believes the system has it covered. Ending on a person who is always reachable, or repeating the policy from the top, closes the gap.
| Step | Who is paged, and how | When it fires | Waits before the next step |
|---|---|---|---|
| Step 1 | The primary on-call, by push notification and SMS | Straight away | 5 minutes |
| Step 2 | The primary on-call again, by phone call | 5 minutes after the first page | 10 minutes |
| Step 3 | The secondary on-call, by push notification and phone call | 15 minutes after the first page | 15 minutes |
| Step 4 | The engineering manager, by phone call | 30 minutes after the first page | None: this is the last step |
Questions
What is an example of an escalation policy?
A common shape has 4 steps. Step 1 pages the primary on-call by push notification and SMS straight away. If nobody has acknowledged 5 minutes later, step 2 phones the same person. After another 10 minutes, step 3 pages the secondary on-call, and after another 15 minutes, step 4 phones the engineering manager. The manager is reached only if nobody has acknowledged within 30 minutes of the first page.
What are escalation policies used for?
Making sure a page is answered even when the first person cannot answer it. People sleep through notifications, lose signal and leave their phones in another room. An escalation policy turns one missed page into a short delay rather than an unanswered outage, and it records who was tried and when.
How long should each escalation step wait?
Long enough for someone who is awake and near a laptop to acknowledge, and no longer. For most teams that means a few minutes at the first steps, then longer waits once the page reaches people who need time to get to a computer. Use shorter waits for your highest severity than for routine alerts.
What is the difference between an escalation policy and an on-call schedule?
The schedule says who is on call at a given moment. The escalation policy says what to do with an alert: who to page first, which channel to use, how long to wait and who comes next. A step in a policy often points at a schedule, so the person it pages changes as the rota turns.
What should happen when an escalation policy runs out of steps?
Something deliberate. Either the last step reaches someone who is always reachable, such as a manager or an incident commander, or the policy repeats from the top a set number of times. What must not happen is a quiet stop, because then an alert nobody acknowledged looks exactly like one somebody is handling.
In Uptimely
Uptimely runs an on-call policy's escalation rules in order. The first rule pages at once, and each rule waits its own delay before the next one fires. Escalation stops when the incident is acknowledged or resolved, in the dashboard or with the Acknowledge button in Slack, or when the person paged presses 1 during the phone call. A policy can repeat from its first rule if nobody acknowledges, which is off by default. Every page is logged, first as queued and then as sent or as an error, and when the last rule has paged the run is marked complete.
Uptimely runs on-call escalation on every plan, including the free one, and logs every page it sends so a missed one is visible.
Start monitoring free →Read next
Part of the Uptimely monitoring glossary — plain definitions of the vocabulary uptime monitoring, incident response and status pages are discussed in. Something here read wrong to you? Email support@getuptimely.com.