Separating rating inputs, rule versions and outputs

A rating-system test is easier to review when the input data, rule version and resulting output can be examined separately. This supports traceability without assuming that a particular rating method or result is appropriate.

Identify the test inputs

Record the approved test case, its source and any transformations applied before processing. Distinguish provided values from defaults and derived values. Use controlled test data and avoid including unnecessary personal information in test evidence.

Pin the rule and software version

Capture the rule-set identifier, application build and relevant configuration used for the test. If a version changes during testing, make a new evidence entry rather than silently replacing the old result.

Compare outputs to approved expectations

• Link each expected output to its approved requirement.
• Record actual output and any difference.
• Include boundary, missing-input and regression cases.
• Track review and resolution of unexpected results.

Before release, accountable product and underwriting owners must decide whether the rules and expected results are correct for the relevant insurance product. This article concerns generic test governance; it does not specify pricing rules, premium levels or VYPR capabilities.