Inspection policy

Defect-detection sensitivity: 70%

This case study models a computer-vision inspection system on a food packaging line processing 5000 items per shift. Higher sensitivity catches more true defects (contamination, packaging failure, wrong labeling) but also rejects more good product as false positives.

Per-shift outcome

True defects caught—

Good product wrongly rejected—

Defects that reach shipping—

Items by defect-likelihood score (orange = rejected, red ring = true defect)