Anomaly detection
Analytics systems use historical baselines, statistical rules, or machine-learning models to flag unusual spikes, drops, or shifts. An anomaly is a signal for investigation, not proof of a problem: it may reflect a product release, marketing campaign, tracking issue, seasonality, fraud, or a genuine change in user behavior. Effective anomaly detection accounts for recurring patterns and gives teams enough context to diagnose the cause.
What is Anomaly detection?
Anomaly detection identifies metric values or behavior patterns that differ significantly from what is normally expected.
Why it matters
Analytics systems use historical baselines, statistical rules, or machine-learning models to flag unusual spikes, drops, or shifts. An anomaly is a signal for investigation, not proof of a problem: it may reflect a product release, marketing campaign, tracking issue, seasonality, fraud, or a genuine change in user behavior. Effective anomaly detection accounts for recurring patterns and gives teams enough context to diagnose the cause.
Example
Daily purchases suddenly fall 35% below their normal weekday range. An alert prompts the team to check checkout errors, payment providers, and tracking health.