Record the context of the decision
Use a stable case identifier and record the product, jurisdiction, application timestamp and decision timestamp. Distinguish an original application from a limit increase, renewal or manual reconsideration. Record the version of the policy and decisioning configuration used. This makes it possible to investigate why two cases were treated differently without relying on the latest configuration to explain an older outcome.
Connect inputs and transformations
Document the data categories used, their sources and relevant verification steps. Where a value is derived, record the transformation or configuration reference needed to reproduce it. Keep a correction history when an input changes. Access to personal data and retention need their own reviewed policies; the goal is not indiscriminate collection. Design the record so authorised reviewers can trace the decision while unnecessary data remains outside the evidence bundle.
Describe exceptions and review
Include a field for the actual decision reason, human review and override rationale. An override should not erase the original result. Link the revised outcome to the reviewer, authority and supporting evidence. Review how the record handles an interrupted application, duplicate submission or later correction. A traceable record is especially useful when several systems contribute information and none of them holds the full explanation.
Use the record in quality review
Sample cases against the intended process and identify missing records, unexplained overrides or inconsistent policy references. Classify a missing item as a gap to investigate rather than automatically concluding that a legal breach occurred. Resolve the question against the applicable requirement and facts. Set review triggers when the product, model, data source or policy changes. A good evidence design makes that investigation possible; it does not certify the outcome.
Primary sources
European Commission: consumer credit ↗Read the source in its current version. National implementing measures and product-specific applicability may require additional research.