How to Run an AQL Acceptance-Sampling Inspection
Set an apparel final inspection on statistical footing: choose the inspection level and AQL, read the ISO 2859-1 / ANSI-ASQ Z1.4 tables for a sample size and accept-reject number, and record the result defensibly.
Reviewed acceptance-sampling references for ISO 2859-1 and ANSI/ASQ Z1.4 inspection levels, AQL selection, sample-size code letters, single-sampling accept-reject numbers, and switching rules; no universal AQL, defect classification, inspection level, or acceptance decision is prescribed, and the buyer-supplier agreement and the full standard govern.
AQL (Acceptable Quality Limit) inspection decides whether to accept or reject a production lot from a random sample instead of checking every unit. It is defined by ISO 2859-1, published in the US as ANSI/ASQ Z1.4, which share the same tables. To run one: agree the AQLs before production (a common apparel set is 0 critical, 2.5 major, 4.0 minor, but this is an agreement, not a rule); pick an inspection level (General Level II is the usual default); take the lot size to the sample-size code-letter table to get a letter; take that letter and each AQL to the single-sampling table to get the sample size and the accept and reject numbers; then inspect that many random units and accept the lot only if defects at each class stay at or below the accept number. Record the lot, plan, and evidence, because the AQL is a lot-acceptance criterion, not a promise every shipped unit is perfect.
An AQL inspection replaces "we checked some and it looked fine" with a statistical sampling plan that both sides agreed in advance. It does not measure the exact defect rate of the lot; it gives a defensible accept-or-reject decision at an agreed quality level, with a known sample size and a fixed pass mark.
Agree the inputs before production, not at the inspection
Four decisions drive the whole plan, and all should be in the purchase order or quality agreement before goods exist, so the inspection is a check, not a negotiation.
| Defect classification | Agree a defect list sorted into critical, major, and minor, with photographed limits and approved standards. The classification is product-specific and is the most disputed part of any inspection, so fix it in writing first. |
| AQL per class | The maximum percent defective a lot may contain and still be acceptable, set separately per class. A common apparel agreement is 0 for critical, 2.5 for major, 4.0 for minor, but tighter or looser values are chosen by risk, price, and market. This is an agreement between buyer and supplier. |
| Inspection level | ISO 2859-1 defines General Levels I, II, and III and Special Levels S-1 to S-4. General Level II is the usual default; Level III takes a larger sample for more discrimination; special levels take very small samples for costly or destructive tests. |
| Lot and sampling basis | Define what one lot is (often one style, one PO, or one shipment), how units are drawn at random across cartons, and whether single, double, or multiple sampling is used. This guide uses single sampling. |
Read the tables in two steps
| Step 1 — code letter | Take the lot size and the chosen inspection level to the sample-size code-letter table. It returns a single letter (A to R). Larger lots and higher levels give a later letter and therefore a larger sample. |
| Step 2 — sample size and Ac / Re | Take the code letter and each AQL to the single-sampling table for normal inspection. It returns the sample size n and, for each AQL, an acceptance number Ac and a rejection number Re (Re is always Ac plus one). |
| Decision | Inspect n random units. For each defect class, accept the lot if the count of defects is at or below Ac and reject it if the count reaches Re. A lot is accepted only if it passes at every class inspected. |
| Arrows on the table | An arrow means the sample size does not exist at that code letter and AQL; follow the arrow and use the sampling plan (and sample size) it points to, not the original letter's sample size. |
Because the three classes usually carry different AQLs, one inspection of the same n units is judged against three different accept numbers at once: a unit can pass on minors but the lot still fails on majors.
Switching and the limits of the method
| Switching rules | ISO 2859-1 is a scheme, not a single table: it starts at normal inspection and switches to tightened inspection after a run of rejected lots, and can move to reduced inspection after a stable run of accepted lots. Using the tables without the switching rules loses much of the protection the standard provides. |
| AQL is not a target | The AQL is the quality level the plan is designed to accept most of the time; it is not permission to produce that many defects, and it is not the measured defect rate of the lot. |
| Sampling risk is real | Any sampling plan carries producer's and consumer's risk: a good lot can be rejected and a bad lot accepted. The plan controls, but does not eliminate, those probabilities. |
| It does not replace tests | Acceptance sampling by attributes checks conforming versus non-conforming units. It does not replace dimensional (points-of-measure), colour, or physical-performance testing, which have their own methods and plans. |
For the measurement side of acceptance, pair this with points of measure and tolerances in apparel manufacturing; for defect definitions tied to construction, see stitch and seam classes.
Inspection setup and result record
AQL inspection plan and result record
Fix the plan before goods exist, then record the draw, the counts, and the decision so the accept-or-reject call is defensible.
Saved in this browser only, never sent to a server.
Agreement and plan (set before production)
These are agreed inputs, not inspection-day choices.
Table lookup
Show the two-step lookup so anyone can reproduce it.
Inspection result
Record counts against each class and the decision.
Sources and decision boundaries
The inspection-level, AQL, code-letter, and accept-reject procedure here follows acceptance-sampling references for ISO 2859-1 and its US equivalent ANSI/ASQ Z1.4, including the Tetra Inspection AQL guide and tables, QualityInspection.org on inspection levels, Sofeast on inspection levels in ISO 2859-1 and ANSI/ASQ Z1.4, and the GQC ANSI/ASQ Z1.4 sampling tables. The AQL values, defect classification, inspection level, and acceptance decision are a buyer-supplier agreement; the full ISO 2859-1 or ANSI/ASQ Z1.4 standard, including the switching rules, governs the actual plan. This guide does not set an AQL, classify defects, or make an acceptance decision for any specific product.
Work with a technical designer.
If two parties keep disagreeing at inspection, the fault is almost always upstream: the defect classification and AQLs were never fixed in writing with boundary samples. Settle those before the next production run and most inspection disputes disappear.
FAQ
What does an AQL of 2.5 actually mean?
It is the Acceptable Quality Limit for that defect class: the quality level, expressed as a percent defective, that the sampling plan is designed to accept most of the time. It is not a target to produce up to, and it is not the measured defect rate of the lot. It combines with the inspection level and lot size to set the sample size and the accept-reject numbers.
Which inspection level should I use?
General Inspection Level II is the usual default for consumer apparel. Level III takes a larger sample and discriminates more strongly, useful for higher-risk or troubled suppliers, while Level I takes a smaller sample. Special Levels S-1 to S-4 take very small samples for costly or destructive checks. The level is agreed with the supplier, not chosen at the inspection.
How many pieces do I need to inspect?
You do not pick the number; the tables do. Take the lot size and inspection level to the sample-size code-letter table for a letter, then take that letter and each AQL to the single-sampling table for the sample size n and the accept and reject numbers. Larger lots and higher levels give larger samples.
Does passing an AQL inspection mean every unit is good?
No. Acceptance sampling makes a lot-level decision from a random sample and carries a known risk of accepting some non-conforming units and rejecting some good lots. It does not guarantee that every shipped unit is defect-free, and it does not replace measurement, colour, or performance testing.
Are ISO 2859-1 and ANSI/ASQ Z1.4 the same?
They share the same sample-size code letters, sampling tables, and accept-reject numbers, so a plan read from one matches the other. They are maintained separately and can differ in administrative detail and switching wording, so cite the specific standard your agreement names and follow its switching rules.