AQL, Explained Without the Mystery
AQL is the most misunderstood set of letters in garment quality. Buyers specify it, factories submit to it, and both sides routinely misread what the tables actually promise. Understanding the mechanics changes how a factory prepares for inspection, and more importantly, how it runs quality before the inspector arrives.
What the tables do
Acceptable Quality Limit sampling answers a narrow question: from this lot, inspect this many pieces, and accept the lot if defects found do not exceed this number. The table trades off inspection effort against risk: sample sizes scale with lot size, and the accept/reject numbers derive from the AQL level, commonly 2.5 for major defects, 4.0 for minor, that the buyer contract names.
Two things the tables do not promise. They do not certify that an accepted lot contains at most 2.5 percent defects; they certify that lots at that quality level will usually pass. And they say nothing about the pieces not inspected: sampling is a statistical bet, not a search.
The factory's real lesson
Because sampling is probabilistic, a factory living near its AQL threshold is gambling every submission: the same lot that passed Tuesday can fail Thursday with different random pieces drawn. The stable strategy is not better luck at final inspection; it is production quality far enough inside the threshold that the dice stop mattering. That is why final AQL results belong on the same dashboard as inline and endline QC: final inspection is the exam, but the course was taught at the needle.
Run the math before the inspector does
Everything the buyer's inspector will do is computable in advance: lot size determines sample size determines accept/reject numbers. A factory that runs its own pre-final on the same basis, same sampling plan, same defect classifications, knows its pass probability before booking the inspection, and can re-screen a marginal lot instead of burning an inspection slot and a reputation on a coin flip.
Classification discipline underneath
The whole edifice rests on defects being classified consistently: what counts as major versus minor, applied the same way by every checker on every shift. That consistency is a taxonomy problem before it is a training problem, and it is where AQL programs quietly fail: a factory whose checkers disagree with the buyer's inspector about what a major defect is will lose lots it thought were safe, entirely on definitional ground.