When testing a report's accuracy, which approach is appropriate?

When testing a report's accuracy, which approach is appropriate?

• A. Rerun with random prompts and hope results match.
• B. Export the report to CSV and inspect only the header.
• C. Run with representative prompts, compare results to source data, and verify totals and subtotals.
• D. Rely on automated checks that never compare to source data.

Answer: (C)

The key idea is to validate a report’s accuracy by testing it with representative prompts and checking the results against the source data, including totals and subtotals. Use inputs that reflect how users will actually run the report, then compare the returned values to the underlying records and to the calculated aggregations. Verifying totals and subtotals ensures the aggregation logic and the way the report groups data are correct, not just that individual rows look right. This approach catches issues in filters, joins, or calculations that can cause mismatches when data changes. Rerunning with random prompts is unreliable because it may miss typical scenarios and edge cases. Exporting to CSV and inspecting only the header misses the actual data you need to verify. Relying on automated checks that never compare to source data has no ground truth to validate against and can overlook real data integrity problems.

The key idea is to validate a report’s accuracy by testing it with representative prompts and checking the results against the source data, including totals and subtotals. Use inputs that reflect how users will actually run the report, then compare the returned values to the underlying records and to the calculated aggregations. Verifying totals and subtotals ensures the aggregation logic and the way the report groups data are correct, not just that individual rows look right. This approach catches issues in filters, joins, or calculations that can cause mismatches when data changes.

Rerunning with random prompts is unreliable because it may miss typical scenarios and edge cases. Exporting to CSV and inspecting only the header misses the actual data you need to verify. Relying on automated checks that never compare to source data has no ground truth to validate against and can overlook real data integrity problems.