pillars of data observability:
Freshness: Is data up-to-date? When was the last successful update?
Volume: Is the data size expected? Row counts, file sizes, partition counts.
Schema: Did structure change? New columns, type changes, dropped fields.
Distribution: Are values normal? Statistical properties of columns over time.
Lineage: Where did data come from? What depends on it?
Each pillar catches different failure modes. Schema changes break pipelines immediately. Distribution shifts cause silent data quality degradation.