Governance is the trust layer of the data platform. It answers four questions every consumer asks before trusting a number: what is this dataset, where did it come from, is it good enough, and am I allowed to see it? Every pipeline run, query and dashboard publish is observable through this system.
Responsibility in one sentence
Catalog datasets with ownership, track lineage, enforce quality rules with incident response, and apply row/column access policies with a full audit trail.
Could not build diagram for model "example-data-processing".
Searchable registry of datasets, columns, owners and tier tags.
Dataset-to-dataset graph built from pipeline lineage events.
Scheduled and on-demand quality checks with incident publishing.
Row/column policies, masking rules and immutable audit log.
The Catalog Repository stores ownership, tier and tag metadata. Stewards curate entries via the Data Catalog API.
Quality validation gates the nightly batch before gold tables update:
Access control is applied when an analyst publishes a dashboard:
Platform-wide guarantee
Governance aims for full audit coverage — every query, publish and policy change is recorded. When quality rules fail, downstream pipelines halt rather than propagate corrupt data.