Why sampling works
Checking every piece of a large order is slow and expensive. Statistical sampling checks a random selection from across the lot and gives a reliable picture of the whole, as long as the sample is truly random and drawn from different cartons.
From lot size to sample size
The standard groups lot sizes into ranges. With the most common setting, General Inspection Level II, a lot of 1,201 to 3,200 pieces gives sample code K, which means 125 pieces are checked. A lot of 3,201 to 10,000 gives code L, which means 200 pieces.
Level I checks fewer pieces and Level III more; special levels are used for tests that destroy the product.
Reading the acceptance numbers
For each defect class you choose an AQL value, and the table gives an acceptance number (Ac) and a rejection number (Re). With 125 pieces, AQL 2.5 allows up to 7 major defects (8 fails), and AQL 4.0 allows up to 10 minor defects (11 fails). Critical defects are usually set to zero tolerance.
- Critical: unsafe or illegal; normally zero allowed.
- Major: the product fails or would likely be returned.
- Minor: small cosmetic issues that do not stop a sale.
What a pass really means
A pass means the sample met your limits; it does not mean the lot has zero defects. That is why the defect definitions in your checklist matter as much as the AQL numbers: a scratch can be minor on the bottom of a product and major on its front.
Writing good defect definitions
The AQL numbers only work if everyone agrees what counts as a defect. Write clear definitions with photos where possible: the size of a scratch that is acceptable, the colour tolerance, which surfaces are visible to the customer. Inspectors and factories then judge the same way.
Choosing your AQL levels
Common settings are 0 for critical, 2.5 for major and 4.0 for minor defects. Premium products or demanding retailers may use stricter values, such as 1.5 for major. Stricter levels mean fewer defects are accepted, and factories may price that in, so choose levels that match the market you sell to.
