A computer-vision defect scanner assigns every manufactured unit a defect score. Scrap or rework everything above a low threshold and true defects rarely escape — but so many good units get pulled that yield suffers. Raise the threshold, and yield improves while some real defects slip through to the customer.
The AI Quality Control Lab models a 3,000-unit production run. Sliding defect-flag sensitivity trades scrap volume against the number of true defects that make it out the door undetected.
The asymmetry in cost is what makes this trade-off hard in practice: an escaped defect usually costs far more downstream than a scrapped good unit, which pulls most real quality-control policies toward higher sensitivity than a pure yield-optimization view would suggest.
🧪 Try it yourself: the AI Quality Control Lab simulation lets you move the defect-flag sensitivity and watch the production-run outcome update live.
🧪 Try it yourself: the AI Quality Control Lab simulation lets you experiment with everything described above directly in your browser.