Alerts

Alerts watch your metrics and notify you when something needs attention. Create them from the Alerts screen.

Alert types

• Threshold — fire when a metric crosses a value (greater/less than, over a time window). E.g. "error count > 50 in the last 60 minutes".
• Change — fire when a metric moves by a percentage versus a baseline window. E.g. "page views drop > 20%".
• Error spike — fire on a sudden rise in errors.
• New error — fire when a previously unseen error appears.

Metrics

Page views, sessions, error count, and Core Web Vitals (LCP, CLS, INP, FCP, TTFB). Vitals alerts use the same thresholds as the Performance hub, so an alert and the dashboard never disagree about what counts as slow.

Channels

• Email — one or more recipient addresses.
• Slack — an incoming-webhook URL.
• Webhook — your own HTTP endpoint.

Options

• Cooldown — minimum time between notifications, to prevent alert fatigue.
• Segment — scope the alert to a specific segment.
• Enable/disable and send a test notification to verify your channel.

Setting thresholds that survive contact with reality

The failure mode for alerting isn't missing an incident, it's alerting so often that people stop reading. Before enabling, look at the metric's normal range over the last month and set the threshold outside it — then add a cooldown long enough that one bad afternoon produces one notification, not forty. An alert nobody acts on should be deleted, not tuned down.

If you just want a periodic summary rather than an interruption, use a digest instead. Alerts are for things that need someone to stop what they're doing.