Pass, Fail, or Test Again? How Borderline Results Create Hidden Risk in Production Testing

The useful part

Results | Quality Magazine Measurement Measurement Pass, Fail, or Test Again? Retesting should follow a predefined decision rule rather than "test until it passes," which can bias acceptance toward marginal units. Manufacturers can improve automated testing by defining borderline-result policies in advance and monitoring trends, repeated PASS/FAIL switching, and test-system stability.

How it works

• What you can do is make the risk visible and decide how much of it you are willing to accept.
• How should manufacturers think about the difference between rejecting a good unit and accepting a bad one?
• If you reject a good unit, the manufacturer usually pays for it in yield, rework, test time, and investigation.
• From your experience in automated testing and metrology, what is a common mistake engineers make with production test data?
• The data around the decision can be more valuable than the decision itself.

What to take from it

Because the software gives you a binary answer, but the measurement behind that answer is not absolute. Over time, that can allow marginal units through simply because they were tested more often. If you only keep retesting, you can hide the symptom without solving the problem.

Example or evidence

• Greg Cenker and Henry Zumbrun Related Articles The Hidden Quality Risk Manufacturers Can't Inspect Away See More Beyond Pass/Fail:
• Tatiana Krasik is an independent researcher and measurement engineer with more than 15 years of experience in electronic measurement, RF and microwave engineering, automated testing, metrology, EMC...
• In practice, things get more complicated when a result lands very close to a specification limit or when a unit fails once and passes the next time.
• Let's say the upper specification limit is 10.0 and the tester reports 9.99.

Details worth keeping

Phuchit / iStock / Getty Images Plus (Creative #2118929800) September 24, 2026 Need to Know Pass/fail results near a specification limit carry more decision risk because measurement uncertainty can affect confidence in whether a product truly conforms. Why isn't a pass/fail result really as binary as it looks? Those two situations are not really the same.

Related coverage

• Qualitymag: Visual inspection is changing. For years, a typical machine vision application was relatively static: position a camera, capture an image, analyze it and make a pass/fail decision.
• Qualitymag: Conventional screening asks: does this unit violate a limit we set?
• Qualitymag: When test results are generated using inconsistent methods or non-standardized procedures, the results do not confidently reflect the true properties of your products.
• Digitalthoughtdisruption: <img data-recalc-dims="1" decoding="async" width="900" height="506" data-attachment-id="14979" data-permalink="https://digitalthoughtdisruption.com/2
• Qualitymag: Not every 95% confidence interval actually delivers 95% coverage.