Data Observability

Data observability is the practice of monitoring the health of data itself in production — not whether the pipeline ran, but whether what it delivered is right. The standard signals: freshness (did the table update on schedule?), volume (did row counts swing abnormally?), schema (did a column change or vanish?), and distribution (did nulls, duplicates, or value ranges drift?).

Observability tools learn baselines from history and alert on anomalies, using lineage to trace an incident to its upstream cause and its downstream blast radius. It complements, not replaces, explicit quality rules and data contracts: contracts catch the violations you anticipated; observability catches the ones you didn’t.