Prepare three different expected outcomes
Choose a clearly compatible case, one that must be excluded and one with a missing specification. Record product ID, variant, rule source and what has actually been approved. Buyer-friendly wording alone is not technical evidence.
Fictional example: a sign and stand share the right format, but the specification says nothing about the chosen mounting surface. Format compatibility may be supported while the overall outcome remains unknown. Write that conclusion before viewing the tool's answer.
Change variants and units in controlled tests
Test two articles in the same family with different dimensions or connections. Ensure generic family text does not make both compatible. Include boundaries where the product owner has decided whether equality is permitted.
Try equivalent values in the units the tool actually supports. A wrong unit must not silently become a different dimension. Where conversion is unavailable, instructions and correction messages need to explain what users should enter.
| Reference case | What is tested | Outcome to compare |
|---|---|---|
| Supported combination | All required properties are known | Correct candidate within the stated scope |
| Documented incompatibility | A required property fails | Candidate is not wrongly recommended |
| Missing property | Evidence is insufficient | Unknown with an assessment route |
| Different variant | Similar name, different properties | Rule uses the correct article |
| Changed need | Previous candidate no longer applies | Updated result with an explanation |
Test zero, one and several candidates
A tool that always displays one product may hide a lack of matching rules. Create a case with no known match within the sample and explain it without claiming the entire market has no solution.
Then test several approved candidates. Buyers should be able to compare relevant facts. Display order need not be a suitability ranking; avoid labels such as best when the evidence does not support that assessment.
Follow a source change through to the result
Change a relevant specification in the controlled test source and rerun affected reference cases. Establish which evidence has actually been published and whether an old answer remains unintentionally. Identify who reviews the change.
Exercise questions and results on mobile and with a keyboard. Note whether instructions, group names and contact routes are understandable. Report sample, rule version, source and unresolved cases. Passing is not general technical approval for the application.