KYC QA / negative cases

Anti-fraud QA without real personal documents.

A reproducible test set checks rules, workflow order and error handling without mixing the result with changing real-world data.

Test each layer separately

An anti-fraud pipeline usually combines several systems. Each layer needs its own expected result for accurate diagnosis.

Format and completeness

Required fields, allowed characters, dates, number lengths and cross-field consistency.

Routing

Correct GEO, document type, OCR model and next verification step.

Decision and reason

An expected status should have a stable reason code so regressions can be grouped and traced.

Scenario matrix

Positive scenarios

  • all required fields are present;
  • formats match the selected GEO;
  • OCR and input data agree;
  • the route reaches the expected test status.

Negative scenarios

  • one required field is intentionally missing;
  • a date or number violates a known rule;
  • document type and route do not match;
  • a controlled cross-field contradiction is present.

Preserve reproducibility

Assign a stable identifier to each case. Store the expected result, rules version, OCR version and input-artifact checksum. After a pipeline change, run the same set and compare both the final status and reason code.

If a test uses random generation, save the seed or final input. Otherwise a difference between runs cannot be attributed confidently to a code change.

Safety boundary

Test materials must remain clearly marked and stay inside a controlled QA environment. Do not remove markings, insert real people's data or send the artifacts into legally significant workflows.