A board report says that 98% of high-risk customer reviews are complete. That number is useful only if the bank can explain which customers were included, what “complete” means and whether every relevant system contributed. Data lineage makes those questions answerable by connecting the reported metric to its sources and transformations.

The Basel risk-data principles apply to systemically important banks and support reliable aggregation and reporting. The Committee's January 2026 implementation newsletter discusses continuing challenges, including lineage, and expressly does not create new supervisory guidance. Using these disciplines for a wider compliance dashboard is a practical extension; the bank must assess its own supervisory scope separately.

Define the metric as a controlled object

Create a short specification for each material metric. State the business question, numerator, denominator, reporting date, exclusions, treatment of missing values and responsible owner. Define status transitions. A case marked “work completed” may still await quality assurance, approval or customer contact; whether it belongs in the numerator should not depend on local convention.

Record the version of the definition. When the bank changes a rule, either restate comparative figures transparently or explain why the series is no longer directly comparable. A sudden improvement caused by a narrower denominator should not be presented as a control improvement.

Follow the transformation chain

Map the source systems, extracts, joins, filters, calculations and final presentation. Identify the keys that connect customers, accounts, legal entities and cases. Look for one-to-many relationships that can duplicate counts and inner joins that can silently remove unmatched records. These technical details matter because they change the management conclusion.

Include spreadsheets and manual adjustments. A lineage diagram that stops before the final workbook misses the place where important judgements may occur. Retain adjustment reasons, authorisation and reconciliation to the unadjusted result. The aim is not to eliminate all human intervention but to make consequential intervention visible and reviewable.

Reconcile both totals and exceptions

A total-count reconciliation is necessary but insufficient. Two errors can offset each other. Add checks for missing identifiers, duplicate cases, unsupported statuses and stale dates. Reconcile important subpopulations, such as entities using a legacy platform or customers with several open reviews.

Track exceptions by consequence and age. A minor formatting defect differs from a missing feed that excludes an entire business line. Where a known gap affects the reported conclusion, show it beside the metric. Do not bury it in a technical appendix that the decision-maker will never see.

Work through a misleading improvement

In a hypothetical bank, review completion rises from 91% to 98% after a reporting change. Investigation shows that cases with an unrecognised status were dropped from the denominator. The operational backlog has not improved; its visibility has deteriorated.

A sound repair reinstates the missing population, clarifies the status mapping and recalculates the affected periods where feasible. It also assesses decisions made using the distorted figure. The bank should test whether similar filters exist in other dashboards rather than treating this as one isolated reporting mistake.

Test reporting under stress

Ask the reporting team to answer an unfamiliar but relevant question, such as how many affected customers belong to subsidiaries relying on a disrupted provider. Observe the time required, manual dependencies and assumptions. A dashboard can be reliable for a fixed monthly template yet fail when management urgently needs a different view.

Use the exercise to prioritise data work. Critical identifiers, controlled definitions and reconciled feeds usually deserve attention before a more attractive visual interface. Document what remains uncertain and the compensating checks available while the underlying problem is repaired.

Make ownership operational

Assign the business owner responsibility for meaning and use, with technology owners responsible for the relevant processing controls. Compliance should challenge whether the metric supports its conclusion; internal audit can assess the wider framework independently. Ownership should include authority and resources to correct problems.

The decisive question is simple: could a qualified reviewer reproduce the figure and explain its limitations? If the answer is yes, the board can use the report with informed confidence. If not, the bank has a measurement risk that belongs alongside the compliance risk it is trying to report.