Does AQL alone determine the sample size?

No. A specified scheme normally requires other inputs, and different plans can have different operating characteristics.

Is an accepted lot certified to contain fewer defects than the AQL?

No. Acceptance is a decision under a sampling rule, not a complete measurement of the lot's true quality.

Are "acceptance quality limit" and "acceptable quality level" always interchangeable?

They are closely related AQL terminology, but wording and interpretation should follow the cited standard or reference. State the convention used in an actual agreement.

AI Glossary / Statistics and probability / Intermediate

Acceptable quality limit

Definition

Acceptable quality limit is a term commonly associated with AQL, a quality level used to select or describe acceptance-sampling schemes. ISO terminology uses "acceptance quality limit", while older references often use "acceptable quality level". For a continuing series of lots, it concerns a limiting satisfactory process average, not a guarantee about every accepted lot.

What it means

AQL helps specify how an inspection scheme should behave when the underlying quality is at an agreed reference level. A well-chosen scheme usually has a high probability of accepting lots at that level, while also considering protection against substantially worse quality.

The AQL is not simply the maximum number of defective items permitted in an inspected sample. That decision uses an acceptance number determined by the sampling plan. Nor does acceptance prove that the whole lot's nonconforming proportion is below the AQL. Sampling decisions remain uncertain because only part of the lot is examined.

How it works

The parties first define what counts as nonconforming and how quality is expressed, for example percentage of nonconforming units. They select an appropriate sampling scheme and the required inputs, which can include AQL, lot size and inspection level. The scheme specifies sample sizes, decision criteria and any switching rules.

For a simple single-sample plan, inspect n randomly selected units and accept if the number of nonconforming units does not exceed c. Its operating-characteristic curve shows the probability of acceptance at different underlying quality levels.

Producer's risk describes rejection at the designated acceptable quality point. Consumer's risk concerns acceptance at a separately designated poor-quality point. AQL alone therefore does not fully describe the plan's protection. The intended series-of-lots setting and the applicable scheme must also be considered.

Three practical examples

1. A purchasing team uses an agreed AQL-indexed scheme to define inspection decisions for a continuing supply of manufactured components.

2. A data-processing team adapts an explicitly specified sampling procedure to inspect batches of records, defining the error categories before drawing the sample.

3. A quality analyst compares operating-characteristic curves to explain why two plans associated with similar quality targets can provide different protection against poor lots.

Think of it this way

Think of AQL as one reference point used to design a screening procedure. It is like specifying how a detector should behave under a known condition, rather than claiming that every item passing the detector has been individually verified. The reference level and the actual decision rule are related but different.

Worked example

Consider an original illustrative plan that samples 50 units and accepts a lot if at most one sampled unit is nonconforming. Assume independent Bernoulli outcomes with nonconforming probability p, as a binomial approximation appropriate to the sampling setting. This is not a prescribed ISO table plan.

At p = 0.01, the acceptance probability is P(X=0)+P(X=1) = 0.99^50 + 50(0.01)(0.99^49), approximately 0.9106. If 1% is the designated AQL for this example, the producer's rejection risk there is approximately 1 - 0.9106 = 0.0894, or 8.94%.

At p = 0.10, the same plan accepts with probability 0.9^50 + 50(0.1)(0.9^49), approximately 0.0338. Even this much poorer quality can occasionally pass. The calculation demonstrates why an AQL label is not a deterministic boundary separating every accepted and rejected lot.

Illustrative case study

Fictional case study: Cedar Components agrees an inspection procedure with a supplier. A manager assumes that an AQL of 1% means every accepted delivery contains no more than 1% nonconforming units.

The quality analyst explains the sampling uncertainty and presents the plan's acceptance probabilities at several quality levels. The team records the applicable scheme, inspection conditions, error classifications and response to rejected lots. It also continues monitoring the production process instead of treating inspection acceptance as evidence that no improvement is needed.

This fictional case illustrates communication about a quality agreement. It does not recommend a particular AQL or claim compliance with a standard; those choices require the actual product requirements, risks and complete sampling procedure.

Common misconceptions and mistakes

AQL is not an allowance to knowingly produce defective items, and it is not the observed sample defect percentage. Another mistake is confusing AQL with the acceptance number c: one is a quality reference, while the other is a count used in a particular sample decision.

Do not assume that a zero-acceptance-number plan guarantees a defect-free lot. Undetected nonconforming items can remain outside the sample. Also, counting nonconforming units differs from counting nonconformities: one unit can contain several nonconformities, so the measurement basis must be explicit.

Business use

For businesses, AQL can support consistent agreements about inspection rather than ad hoc acceptance decisions. Its value depends on using the full sampling scheme and understanding both sides of the risk tradeoff.

Specify the lot definition, selection method, inspection criteria and disposition rules. Review whether ordinary acceptance sampling is suitable for the consequences of missed errors, rather than borrowing a percentage from an unrelated product. Keep process improvement separate from lot acceptance: a scheme helps make disposition decisions but does not identify or eliminate the causes of poor quality.

Check your understanding

A plan samples 80 units and accepts at most two nonconforming units. The sample contains two. Does accepting the lot prove its true nonconforming percentage is 2/80?

Answer: No. The lot passes that sample rule, but 2/80 is only the observed sample proportion. The unsampled units and sampling uncertainty remain relevant.

Related terms

Acceptance sampling
acceptance number
operating-characteristic curve
producer's risk
consumer's risk
limiting quality
nonconforming unit and process average
its decisions and its statistical protection

Learn next

plot acceptance probability against underlying nonconforming proportion for two simple plans. Mark the acceptable and poor-quality reference points and explain both risks. Then compare the illustration with the complete procedure required for the intended application.

Acceptance probability

Sources

Primary sources: ISO, ISO 2859-1:2026 overview and AQL terminology: https://www.iso.org/standard/2859-1

NIST/SEMATECH, Lot Acceptance Sampling Plans and definitions: https://www.itl.nist.gov/div898/handbook/pmc/section2/pmc22.htm

NIST/SEMATECH, Choosing a Sampling Plan with a given OC Curve: https://www.itl.nist.gov/div898/handbook/pmc/section2/pmc222.htm

Previous: Accented syllables Next: Acceptance probability