Detect unusual points, shifts, and patterns with statistical rules, control limits, seasonal baselines, and machine-learning methods. Tune alerts around operational cost and investigate whether anomalies are errors, attacks, or real events.