How fast a service is consuming its error budget, the leading indicator that SLO-based alerting fires on.
Burn rate is the speed at which a service is depleting its error budget. A 1x burn rate means you're on track to use exactly your budget over the SLO window. A 14x burn rate means you'll burn the entire month's budget in two days. Multi-window burn-rate alerts (e.g. 1-hour and 6-hour rolling windows) are the modern replacement for static-threshold paging because they capture both fast-moving outages and slow-moving regressions in the same framework.
Static thresholds page at the wrong moment, often after damage is done. Burn-rate alerts page when the service is on a trajectory that will breach the SLO if it continues. That's earlier than 'we just broke the SLO' and later than 'one slow request' so the on-call signal-to-noise ratio is dramatically better.
See the part of the platform that handles burn rate in production.