Alerts

Get notified when a metric crosses a threshold, deviates from normal or goes stale. Conditions, anomaly detection, channels, digests and muting.

Updated Oct 5, 20265 min readEdit on GitHub

Alerts watch a measure so you do not have to. They evaluate after every sync of the underlying source and notify the right people with context: the current value, the expected range, and a link to the chart that explains it. Explore examples in the live demo.

Alert conditions

Condition	Example	Good for
Threshold	`churn_rate > 3%`	Hard limits and SLAs
Change	`signups` down 20% vs last week	Sudden drops or spikes
Anomaly	`organic_clicks` outside expected band	Seasonal metrics
Freshness	`revenue` older than 2 h	Pipeline health
Row match	Any vessel with AIS gap > 6 h	Event-style monitoring

Anomaly detection

Anomaly alerts fit a seasonal model (day-of-week and trend) on up to 18 months of history and flag values outside a confidence band. Sensitivity has three levels: low (99% band, fewer alerts), medium (95%) and high (90%). Kimo needs at least 28 data points before an anomaly alert becomes active.

Create an alert

1
Open a chart and click the bell
Kimo pre-fills the measure, filters and dimension from the chart.
2
Pick a condition and a scope
Optionally split by a dimension: one alert per country or per campaign instead of one global alert.
3
Choose recipients and channels
People, groups, Slack or Teams channels, email, PagerDuty or a webhook.

Webhook payload

1{2  "alert_id": "alr_5Tq9",3  "status": "triggered",4  "measure": "signups",5  "value": 312,6  "expected": { "low": 401, "high": 488 },7  "scope": { "country": "DE" },8  "evaluated_at": "2026-10-07T06:15:00Z",9  "url": "https://app.getkimo.com/alerts/alr_5Tq9"10}

Keeping noise down

Cooldown — after triggering, an alert waits (default 6 h) before firing again for the same scope.
Digest — bundle non-urgent alerts into one daily message at 9:00.
Mute — silence an alert until a date, with a reason visible to the team.
Resolve — alerts auto-resolve when the value returns inside the band, and post a follow-up.

Ownership and audit trail

Each alert has an owner and an optional runbook link that appears in every notification. The alert history keeps the last 13 months of evaluations: value, expected range, who was notified and who acknowledged. This makes post-mortems straightforward ("when did signups start dropping, and who knew?") and helps prune alerts nobody acts on. Kimo flags alerts that triggered more than ten times in a month without anyone opening the linked chart, so you can tune or delete them.