Detect anomalies in metrics and datasets — defining “normal,” reducing false alarms, and routing true issues to investigation. Use when building alerting on KPIs, fraud-like outliers, or data quality anomaly checks.
