How to validate a factory's automated final test
Challenge the production tester with independently checked good and faulty units. Confirm it detects the intended defects, investigate inconsistent results, and retain the exact programme, limits and settings used. Agree what stops batch release and when validation must be repeated. Arrange factory witnessing through Cambridge China Bridge.
Define what the tester must detect
Ask the factory for a map connecting each relevant product defect to a test step, measured result and rejection limit. Include missing functions and incorrect assembly, rather than checking only whether the product powers up. Record which defects the station cannot detect and how those will be checked elsewhere.
Agree a written validation plan before the demonstration: sample identities, expected outcomes, operating conditions, evidence to retain and who approves release. This checks the tester's ability to distinguish acceptable products from defective ones. For deciding which production units to inspect, see full inspection or sampling. For accepting the testing equipment before dispatch, see our factory acceptance test guide. For testing equipment after installation, see our site acceptance test guide.
Prepare identifiable good and faulty units
Have a competent engineer independently confirm each sample's condition using a suitable reference method. Include verified good units, representative faulty units and samples close to the agreed acceptance limits. Record the actual defect and expected failing step. An obviously dead product alone gives little evidence that subtle defects will be detected.
Label and segregate challenge samples so they cannot enter saleable stock. Ask the engineer to select safe ways to represent faults; do not improvise hazardous damage. Retain each sample's identity, condition record and storage arrangements, and recheck its condition before reuse. A sample that has deteriorated is no longer a reliable reference.
Challenge the actual production station
Use the station, fixture, programme and product variant intended for your order. Mix the sample sequence without telling the operator the expected result, where practical. Repeat loading and testing across relevant fixtures and operators. Watch the physical reject route and confirm that the saved result belongs to the unit actually tested. See remote witnessing of factory tests for arranging observation.
Challenge missing connections, interrupted tests, incorrect product selection and permitted retests. An incomplete or skipped test should not produce a releasable pass. Check that earlier results cannot be reused for another unit and that a failed unit cannot return to stock merely because its label was replaced.
Investigate false passes and false fails
A false pass means a verified defective unit is accepted; a false fail means a verified acceptable unit is rejected. Keep the individual readings and compare them with the independent reference results. Investigate poor contacts, fixture wear, unstable supplies, measurement variation and incorrect limits. For the measurement investigation, use checking factory measurements.
For samples near a limit, ask the test engineer to explain measurement uncertainty and the agreed decision method. Do not widen limits simply to clear rejects. Keep the original failure and subsequent attempts linked. Agree that unexpected results suspend approval until the cause is understood, corrected and challenged again.
Working out your own order? Send us the product on WhatsApp and we reply the same working day: WhatsApp us
Retain settings and control release
Retain the approved executable programme or a controlled recoverable copy, its version and file fingerprint, test sequence, limits, units, timing, instrument configuration and fixture identity. Include product firmware, calibration status, access permissions and retest rules. Keep the settings export with the validation results; a pass-screen photograph does not show which limits were active. Coordinate product versions through factory firmware and software approval.
Make each production result traceable to its unit, station and approved programme version. Agree reference-sample checks at start-up and after relevant changes, with a recorded schedule during production. If a check fails, hold potentially affected stock back to the last satisfactory check and investigate. Revalidate affected functions after programme, limit, fixture, instrument or product changes. Cambridge China Bridge's staff in China can witness the agreed challenge and collect the settings and results for buyer review. For hi-pot testing, see our factory hi-pot test record guide.
Frequently asked questions
Is a known good unit enough to validate a tester?
No. It shows the tester can accept that unit. Add independently verified faulty samples to check that the relevant defects are actually rejected.
What if a faulty unit passes the factory test?
Suspend approval, preserve its readings and active settings, and investigate the missed defect. Correct the cause and repeat the relevant challenges before releasing affected stock.
Can the factory retest failed units until they pass?
Agree controlled retest rules. Preserve every attempt and require a recorded explanation for release. Repeated testing alone does not demonstrate that a defect has been removed.
Which test-programme records should I keep?
Keep the approved programme copy, version, limits, configuration, fixture identity, access and retest rules, plus challenge results. Link production records to the approved version.