Automatically spotting values that break from the expected pattern, instead of relying on fixed manual thresholds.
Reviewed by Francisco Ferreira ·
Anomaly detection is the automatic identification of data points that don't fit the established pattern — a row count, null rate or metric value that lands outside what the history would predict. The alternative, hand-set thresholds, ages badly: a rule that made sense last year fires false alarms today.
Done well, it adapts to context — rhythm, seasonality, gradual growth — so it flags the genuine break and stays quiet on normal variation. The goal is high signal: alerts you actually trust, instead of a noisy channel everyone learns to ignore.
Tabkeel runs anomaly detection on each monitored table against its learned baseline, so you get told about the real problem and not the noise.