AQL Sampling in Practice: Garment Quality Control Done Right

Start with the lot, not the table

An AQL decision is only valid for the lot it was drawn from. If half the shipment is still in finishing when the inspector arrives, the inspection covers a different population than the one that ships. Define the lot on paper before the first carton is opened: purchase order, style, colour, destination and the exact carton count offered for inspection.

Partial presentation is the single most common reason a passed inspection is followed by a customer claim. When less than 80 percent of the lot is packed, either reschedule or record the inspection explicitly as a partial check with limited coverage.

Randomness is a procedure, not an intention

Cartons stacked nearest the door are not a random sample. Number the cartons, use a random generator or a fixed skip interval, and pull pieces from the top, middle and bottom of each selected carton. Record which carton each defect came from — clustering tells you whether the problem is systemic or line-specific.

• Draw from at least the square root of the total carton count
• Never let the factory pre-select the cartons or pre-open them
• Log carton IDs on the report so the sample can be reconstructed

Switching rules exist — use them

Normal, tightened and reduced inspection are not decorative. A supplier that fails two of five consecutive lots should move to tightened inspection, and a supplier with ten clean consecutive lots has earned reduced sampling. Applying a static plan to every supplier wastes inspection hours on the good ones and under-samples the risky ones.

Classify defects before you count them

Critical, major and minor need written definitions with photographs, agreed with the buyer before production. Without an approved defect catalogue, every inspection ends in negotiation. With one, the report is arithmetic.