Data lineage, end to end.

What is data lineage?

Data lineage is the documented path a data element takes from its origin — operational system, third-party feed, manual upload — through every transformation, aggregation, and consumption point until it lands somewhere a human or system reads it. Lineage answers the question "where did this number come from" and, more importantly, "what breaks if I change this upstream."

Why it matters in 2026

Three forces have made lineage a default expectation rather than an advanced capability:

• Regulatory mandates. BCBS 239 requires risk-data lineage for global banks. The EU AI Act requires training-data lineage for high-risk systems. CCPA and GDPR practically require lineage to answer subject-access and deletion requests.
• Data product economics. When your analytics, reporting, and ML models are themselves consumed as products by other teams, you can't ship a breaking change without knowing who depends on what. Lineage is the dependency graph.
• AI grounding. Every modern data catalog (Atlan, Collibra, Alation, Select Star) now positions lineage as the substrate for AI-driven discovery. Without it, AI assistants over your data hallucinate confidently.

Where to start

If you're new to lineage as a practice, the canonical entry point is our practitioner-level walkthrough of implementing impact analysis without bogging the organization in enterprise paralysis. From there, the articles below cover the specific operational decisions you'll face — how deep to capture, what to do about downstream consumers you don't control, and which tools earn their license cost vs. which can be replaced by dbt + a Snowflake account map.