AQL: what an acceptance quality limit tells you about a shipment

What an AQL does not promise

AQL 2.5 does not mean that no more than that share of your goods is defective. A sampling plan built on an AQL is designed to give a high probability of accepting a lot that has a defect level less than or equal to the AQL^[1]. That protects the factory from having good lots rejected. It says less about how bad a lot can be and still slip through.

The NIST handbook is plain about the limits: the main purpose of acceptance sampling is to decide whether or not the lot is likely to be acceptable, not to estimate the quality of the lot^[2]. The buyer's side of the plan is the lot tolerance percent defective (LTPD), a designated high defect level that would be unacceptable to the consumer^[1], which a good plan should rarely accept. If your inspector reports only the AQL, ask what defect rate the same plan would still pass some of the time.

How a sampling plan is chosen

1. Set an AQL for each class of defect: critical, major and minor.
2. Choose an inspection level, which ties the sample size to the lot size. The widely used sampling standards offer three general and four special levels^[3].
3. Look up the sample size for your lot size and level.
4. Read the accept and reject numbers for each AQL at that sample size.
5. Inspect the sample, count defects by class, and pass or fail the lot on the counts.

Why a first order needs extra care

ANSI/ASQ Z1.4, the common US standard, is meant to be used with switching rules on a continuing stream of lots^[4]. The switching rules tighten inspection when a supplier's recent lots fail and relax it when they keep passing. A first order from a new factory is one lot with no history, so those rules never act. For a first order, set a stricter AQL for major defects or a higher inspection level, and check the pre-production sample closely, because the inspection judges the goods against it.

Defects or defective units

ASQ describes the ANSI/ASQ Z1.4 plans as covering percent nonconforming or nonconformities per 100 units^[4]. Nonconforming means defective units; nonconformities means defects, and the two are not the same. One unit with a loose thread and a scuff is one defective unit but two defects. Say in the order which one your AQLs count, and define each defect class with examples, ideally photos, so the inspector and the factory classify them the same way.

Example: a lot that passes

An inspection result on an order of 3,000 units

Line	Value
Sample size under the chosen plan	125 units
AQL for major defects	2.5
Major defects allowed in the sample	7
Major defects found	7
Sampled units with a major defect	7 of 125
Result	Lot passes

Illustrative figures, not a quote. Real prices depend on the spec, quantity and factory.

The lot passes, yet the sample's major defect rate is more than twice the AQL. That is the plan working as designed, not an error. If that rate is too high for your customers, the fix is a stricter AQL or level in the order, agreed before production, not a dispute after the inspection. Put the AQL for each defect class, the inspection level and the defect definitions in the purchase order, so the quality inspection tests the terms both sides signed. A factory audit answers a different question: whether the supplier's systems can hold that quality from one order to the next.