Anomaly detection

Anomaly detection is a Databricks data governance capability, introduced February 2026.

Schema-level data quality monitoring that learns each table's historical patterns and automatically evaluates its completeness and freshness, consolidating every monitored table's health into one system table and UI.

• It judges freshness from the table's own commit history, so it can work out that a table has gone stale without reading a single row of it. 🔗
• Its completeness check counts rows and little else - the docs openly admit it ignores the fraction of nulls, zero values, or NaN. 🔗
• It arrived switched on per schema rather than per table, and spent its first six weeks unable to tell anyone what it found: in-UI alerts only followed in March 2026, still in Beta. 🔗

⚠ Limitations: Public Preview. Needs a Unity Catalog enabled workspace and serverless compute, plus MANAGE on the schema or catalog to switch it on. Views and foreign tables are not supported, and the completeness check does not take into account the fraction of nulls, zero values, or NaN. 🔗

Category

Data governance

Introduced

February 2026🔗

Also known as

Data Quality Monitoring Anomaly Detection, DQM anomaly detection

Verified

2026-08-03