I have a reel of 1000 LEDs with a label on them showing that 10 samples were measured to meet my wavelength specification. The vendor guarantees that all 1000 meet my specification. My DA(Design Assurance) engineer says a sample of 30 would be needed to come close to zero defect. Inspection Level S1

Perhaps semiconductor manufacturers have such tight controls on the process steps that this allows them to use smaller sample sizes. Historical data analysis and confidence interval statistics come to mind. ISO16269 (procedures for establishing statistical tolerance intervals) perhaps?

Thoughts?

Thanks!