Data Lineage

A number in a board report is usually the result of many steps: extracts from source systems, mapping, calculations, adjustments and aggregation. Lineage documents that route. It can be drawn at the level of systems, tables or individual fields, and tools can capture much of it automatically from code and data pipelines. Forward tracing shows what would be affected if a source changes, while backward tracing shows where a doubtful figure came from. It supports data quality investigations, impact analysis before system change, regulatory reporting and audit. Banking supervisors stress it in their principles on risk data aggregation, because a firm that cannot trace its risk numbers cannot show they are accurate. It also supports privacy duties, such as finding every place a person's data flows. Lineage is distinct from a catalogue, which describes what data exists rather than how it moves and changes.

A regulator questions a bank's reported liquidity figure. The reporting team follows the lineage backwards: the report pulls from a consolidation layer, which takes feeds from three treasury systems, one of which applies a manual currency adjustment. The team finds a wrong rate in that adjustment and corrects the figure and the feed.