Data lineage
Data lineage is the documented path a data item follows from its original source through every system, transformation and control until it reaches a report, model or decision. It shows where a number came from and what was done to it. End-to-end lineage is a core expectation of the ECB for risk data.
Roles that need it
What it is
Data lineage is a map of how data moves and changes. It records the source system where a value is first captured, each hop between systems, every transformation (a join, a filter, a currency conversion, a manual adjustment) and the controls applied along the way. The end point is the report, regulatory return or model that consumes the value.
Lineage can be documented at different depths. Business lineage shows the flow between applications in plain terms. Technical lineage goes down to tables, columns and code. Most banks combine both, often in a lineage tool fed by metadata harvested from databases and ETL (extract, transform, load) jobs.
Why it matters
When a regulator, auditor or manager asks why a figure looks wrong, lineage is how the team finds the answer quickly. Without it, investigations depend on the memory of a few people.
The Basel Committee’s BCBS 239 principles ask banks to aggregate risk data accurately and completely, also under stress. The ECB guide on risk data aggregation and risk reporting (2024) goes further: it expects complete, up-to-date lineage for critical data elements, from the report back to the point of capture, including manual steps and controls. Lineage also supports change: before a system is replaced, lineage shows which reports will be affected.
Worked example
Illustrative numbers. A liquidity report shows deposits of 1,200 million euro, while the general ledger shows 1,260 million. The lineage for the deposits figure shows four hops: core banking system, data warehouse, a reporting mart and the report engine. Between the warehouse and the mart, a filter excludes accounts flagged as “dormant”. That filter removed 60 million euro of balances (1,260 minus 1,200). The data owner confirms that dormant accounts must be included, the filter is corrected, and the lineage record is updated with the new rule and its approval date.
Who uses it
Data governance analysts build and maintain lineage and link it to the business glossary. Regulatory reporting analysts use it to explain movements and reconcile returns to the ledger. Business analysts use lineage to scope the impact of a change. Data analysts check it before trusting a dataset. Regulatory consultants assess lineage coverage against supervisory expectations. IT auditors test whether documented flows match what systems actually do.
Common confusions
Lineage is often confused with a data flow diagram drawn once for a project. A diagram is a snapshot; lineage must be kept current as systems change. It is also different from data provenance, which focuses on the origin and history of a single record. Finally, lineage shows where to look and which controls exist; measuring whether the data is right is the job of data quality rules.