AQL sampling that actually protects the shipment
Most factories run AQL because the buyer asks for it. Here is how to run it so it genuinely catches risk before the container closes.
AQL is a sampling plan, not a quality system. It tells you how many units to pull and how many defects you can accept — nothing more. Yet in most factories it quietly becomes the only line of defence, and that is where shipments start to fail.
## Start from the defect history, not the table
Before choosing a level, look at the last twelve months of final-inspection results for the same product family. If measurement and stitching defects dominate, a general Level II plan on a single carton pull will miss them. Sampling has to follow where the risk lives: across sizes, across lines, across shifts.
## Pull the sample the way the buyer would
A sample that comes from cartons stacked nearest the door is not random. Number the cartons, draw them across the pallet, and record which cartons were opened. When an inspection is challenged, the traceability of the pull matters as much as the result.
## Treat the accept number as a warning, not a target
A lot that passes at exactly the accept number is a lot that will fail next week. Track the ratio of found defects to the accept number over time; when that ratio climbs above roughly 0.6, the process upstream is drifting even though every report still says PASS.
## Close the loop within 24 hours
The value of an inspection is the corrective action it triggers. Every major defect should reach the line supervisor, the mechanic and the quality manager the same day, with the specific operation named. An AQL result that only reaches a spreadsheet has protected nobody.
## The practical checklist
- Choose the level from defect history, not habit. - Randomise the carton pull and record it. - Trend defects against the accept number, not just pass or fail. - Route every major defect to the operation that caused it, same day. - Re-inspect after corrective action, not after the next order.