Severity level (SEV)
The number attached to an incident to say how damaging it is and how fast, loudly, and widely the team must respond.
See it
What it is
A severity level grades an incident by impact so the response matches the damage. A typical scale runs from SEV-1 for a critical, broad outage down to SEV-4 for a minor issue, with each level defining who is paged, who leads, how often updates go out, and which recovery target applies. PagerDuty publishes its internal incident response guide openly, and its severity definitions are the scale most small teams copy first.
Reach for the scale at incident declaration, then reassess as evidence changes. The useful part is not the number itself but the response contract attached to it: roles, escalation path, communication cadence, and authority to take risky recovery actions.
Gotcha: scales are local. Some organizations use SEV-0, some use P1, and some reverse the numbering, so never assume the label explains itself. Separate severity, which describes impact, from priority, which decides what work happens first, and record every change with a reason.
Ask AI for it
Define and implement a SEV-1 through SEV-4 incident scale. Give each level explicit impact criteria, response time, incident roles, Slack update cadence, stakeholder audience, and closure authority. Map SEV-1 through SEV-4 to PagerDuty Events API v2 payload.severity values critical, error, warning, and info, and route SEV-1 and SEV-2 through high-urgency escalation. Add a classifier that maps alert labels and impact inputs to a level, require a human override, store its inputs and reason, and add tests for promotion, demotion, and conflicting signals.