Quality Mistakes

Quality and Metrology Mistakes That Wreck Your Yield Numbers

The most common and expensive errors in quality and metrology, from confusing First Pass Yield with final yield to under-sizing inspection samples, each with its symptom, root cause, and a numeric fix.

The first mistake is reporting final Yield when the customer wants First Pass Yield. Symptom: your line shows 98 percent yield but scrap and rework hours keep climbing. Root cause: final yield counts reworked parts as good, so it hides the parts that failed inspection and got fixed. If 1000 parts enter, 60 fail first inspection, and 45 are reworked to spec, final yield reads 985 over 1000 (98.5 percent) while First Pass Yield is only 940 over 1000 (94 percent). Fix: track both in the First Pass Yield calculator and Yield calculator side by side. The gap between them is your hidden rework load.

The second mistake is multiplying step yields wrong across a multi-stage process. Symptom: each station reports 97 to 99 percent, so managers assume the line runs near 98 percent, yet finished-goods yield sits at 88 percent. Root cause: Rolled Throughput Yield is the product of every step, not the average. Five stations at 0.97 each give 0.97 to the fifth power, which is 0.859, not 0.97. Averaging inflates the number by roughly 11 points here. Fix: run every station through the Rolled Throughput Yield calculator and chase the lowest step first, since one station at 0.90 drags the whole chain down more than three at 0.99 help it.

The third mistake is treating defect rate and defects per unit as the same thing. Symptom: a supplier quotes 2 percent defective, but your incoming inspection finds far more rejects than expected. Root cause: a single unit can carry several defects, so 2 percent defective units can mean 4 or 5 percent of opportunities failing when parts are complex. If 100 units have 8 total defects spread across 5 bad units, defective rate is 5 percent but defects per unit is 0.08. Fix: define the opportunity count before you measure, feed consistent units into the Defect Rate calculator, and never convert between the two without the opportunity denominator.

The fourth mistake is a unit or PPM slip that shifts your Sigma Level by a full point. Symptom: your Sigma Level jumps from 4.0 to 5.0 overnight with no process change. Root cause: someone mixed defects per million opportunities with defective parts per million, or dropped the 1.5 sigma shift convention halfway through a report. A DPMO of 6210 is about 4.0 sigma with the shift; 233 DPMO is about 5.0. A factor-of-ten error in the denominator moves you a whole level. Fix: lock the opportunity definition and the shift assumption before running the Sigma Level calculator, and document which convention each report uses.

The fifth mistake is under-sizing the inspection sample and calling the lot good. Symptom: parts pass a 20-piece check, ship, then trigger field returns at 3 percent. Root cause: a sample of 20 with zero defects only proves, at 95 percent confidence, that the defect rate is likely below about 14 percent, nowhere near the 1 percent you assumed. Small samples have wide confidence intervals. Fix: size the sample to the AQL and lot size using the Inspection Sampling calculator. To be 95 percent confident the rate is under 1 percent on zero rejects, you need roughly 300 pieces, not 20, and that gap is where escapes hide.

The sixth mistake is ignoring measurement system error, so gage variation gets counted as real defects. Symptom: parts fail at a 0.05 mm tolerance, but re-measuring on a better instrument passes them. Root cause: the gage consumes too much of the tolerance band. If your measurement repeatability and reproducibility eats 30 percent of the tolerance, a part sitting 0.01 mm inside the limit can read as out. Fix: keep gage R and R under 10 percent of tolerance for critical features and under 30 percent for the rest. Above that, your Defect Rate and First Pass Yield are partly measuring the gage, not the part.

The seventh mistake is stale or double-counted scrap data feeding your quality picture. Symptom: Scrap Cost and Cost of Poor Quality totals swing 20 percent month to month with steady production. Root cause: parts scrapped at final are also counted at an earlier operation, or setup scrap is lumped with process scrap. If 40 setup pieces and 40 process rejects both hit the same bucket, your scrap rate doubles on paper. Fix: tag every scrap event by operation and reason code, reconcile it against material issued, and only then push clean numbers into the Scrap Cost and Cost of Poor Quality calculators.

The eighth mistake is letting warranty and return data lag so the field failure signal arrives too late. Symptom: your Warranty Return Rate looks like 0.4 percent for a launch that is actually failing at 2 percent. Root cause: returns trickle in over 6 to 12 months, so early months divide a few returns by full shipment volume and understate the true rate. Fix: age the return rate by cohort, dividing returns from units shipped in a given month by that month's volume, and track it in the Warranty Return Rate calculator. A 0.5 percent reading at month two on a slow-return product can mean 2 to 3 percent at maturity.

Published 2026-07-01.