AQL 2.5 in Practice: Reading the Sampling Table Without Getting It Wrong

Acceptance sampling is a statistical bargain: you inspect a fraction of a lot and accept a known risk of being wrong. The bargain only holds if the sample is genuinely random, the defect classification is agreed in advance, and everyone understands what the AQL number actually means.

What AQL means, precisely

The acceptable quality limit is the worst process average that would still be routinely accepted by the sampling plan. It is a property of the plan, not a promise about the lot in front of you. A lot at exactly the AQL has roughly a 95 per cent chance of acceptance. A lot at twice the AQL still has a meaningful chance of slipping through on a single inspection.

The corollary matters commercially: passing an AQL 2.5 inspection does not mean the shipment contains 2.5 per cent defects. It means you did not find enough evidence to reject it.

Reading the tables in the right order

1. Lot size gives you a code letter from the sample size code table, at General Inspection Level II unless the contract says otherwise.
2. Code letter gives you a sample size from the single sampling plan table.
3. Sample size plus AQL gives you accept (Ac) and reject (Re) numbers.

The standard apparel arrangement is AQL 2.5 for major defects and AQL 4.0 for minor, with zero tolerance for critical. Note that the two are evaluated against the same sample — you do not draw twice.

Defect classification must be agreed before, not during

• Critical — unsafe or unlawful: a broken needle fragment, a missing regulatory label, a choking hazard on childrenswear. Any occurrence rejects the lot.
• Major — likely to result in a return: a visible stain on the front panel, a broken seam, wrong measurement beyond tolerance.
• Minor — a departure from standard unlikely to affect usability: a slightly uneven stitch line in an unexposed area.

Ninety per cent of inspection disputes are classification disputes. Build a physical defect board with real garments showing the boundary case for each category, signed by both parties, and keep it in the inspection room.

The four mistakes

1. Non-random sampling. Drawing all cartons from the top layer of one pallet is the most common. Use the carton numbering and a random number list, and record which cartons were opened.
2. Inspecting before the lot is complete. Inspecting 60 per cent of a lot and extrapolating invalidates the plan entirely.
3. Ignoring switching rules. ISO 2859-1 defines normal, tightened and reduced inspection. Two rejections in five consecutive lots moves you to tightened inspection. Suppliers who consistently ignore this are running a weaker plan than the contract specifies.
4. Re-inspecting after sorting without disclosure. A rejected lot that is 100 per cent sorted and resubmitted must be identified as a resubmitted lot and inspected under tightened rules.

When sampling is the wrong tool entirely

If a defect is safety-critical, or if the process is not in statistical control, sampling gives false comfort. Metal contamination, for example, is handled by 100 per cent needle detection, not by AQL. Use acceptance sampling for what it is good at: economically verifying a process that is already stable.