Column-Level Lineage

Column-level lineage traces the journey of individual columns — not just whole tables — through every transformation between source and consumption. Where table-level lineage tells you that dataset A feeds report B, column-level lineage tells you that the national-ID field in the CRM flows into a staging table, is masked in one branch, and appears unmasked in a legacy export.

That precision changes what governance can do. Sensitive-data tracking becomes exact: when a column is classified as personal or confidential, the classification can be inherited automatically by every downstream column derived from it. Impact analysis becomes surgical: a planned change to one column lists only the dashboards and models that actually use it, not everything in the same table. Audit answers become defensible: 'show every report containing customer phone numbers' is a query, not a project.

For a Saudi Data Management Office, this granularity is what makes NDMO-aligned data classification enforceable in practice and PDPL accountability provable. Personal data rarely stays in one well-labeled table; it propagates through joins and derived datasets. Column-level lineage is the only reliable way to follow it, and the foundation for masking, access decisions, and breach-scope assessment at the field level.