AQL numbers appear on every professional inspection report, yet they are among the most misread numbers in sourcing. This article explains what acceptance quality limit sampling actually decides, works through a real example with accept and reject numbers, and shows how to choose AQL values you can defend. It is written for ecommerce buyers who commission inspections and need to read the results correctly.
What AQL actually says — and what it does not
AQL stands for acceptance quality limit. It is the statistical framework — standardized in ISO 2859-1 and its US counterpart ANSI/ASQ Z1.4 — that answers one question: based on a random sample from this shipment, should I accept or reject the whole lot? The AQL value you set is the worst defect rate you are willing to tolerate as a process average, expressed as an index that maps to specific sample sizes and accept/reject numbers.
What AQL is not: a guarantee. A shipment that passes an AQL 2.5 inspection is not defect-free, and sampling can miss defects that cluster in unpacked cartons or uneven production runs. An inspection report is a statistically grounded decision about a lot, produced from a sample of it. Anyone who reads a pass as "no defects" has misunderstood the tool.
The three inputs that determine your sample
Every AQL calculation starts from three inputs.
- Lot size. The number of units in the shipment being inspected. Bigger lots get bigger samples, but proportionally smaller — doubling your order does not double the inspection workload.
- Inspection level. General Inspection Level II is the default for consumer goods. Levels I and III shift sample sizes down and up; special levels exist for destructive or costly tests.
- AQL values per defect class. Typically one value for critical defects, one for major, one for minor. Consumer-goods programs commonly run critical at zero, major at 2.5 and minor at 4.0.
A worked example: 5,000 units
Take a 5,000-unit order inspected at General Inspection Level II with critical 0, major 2.5, minor 4.0 — the most common consumer-goods setup. The master tables assign this lot a sample of 125 units and produce the accept and reject numbers below.
| Defect class | AQL | Sample size | Accept if | Reject if |
|---|---|---|---|---|
| Critical | 0 | 125 | 0 found | 1 or more found |
| Major | 2.5 | 125 | 5 or fewer found | 6 or more found |
| Minor | 4.0 | 125 | 7 or fewer found | 8 or more found |
Read it concretely: if the inspector finds six major defects in the 125 sampled units, the lot fails on majors — regardless of how clean the rest of the sample looked. One critical defect (a sharp edge, a missing required warning, a safety failure) fails the lot on its own. Some cells in the published tables resolve to a different sample size via arrows rather than fixed numbers; inspection software and professional inspectors handle this, but it is one more reason the standard should be applied by people who use it daily.
Choosing your numbers
The 0 / 2.5 / 4.0 setup is a reasonable default for most consumer goods, and it is the baseline we commonly reference on quality programs — but defaults deserve challenge in both directions.
- Tighten when defects are safety-relevant, when your category punishes quality failures (children's products, anything touching food or skin), or when a single defect type has burned you before. Moving majors from 2.5 to 1.0 raises inspection cost and rejection risk — deliberately, because the cost of misses is higher.
- Loosen for low-value promotional goods where a minor blemish has no commercial consequence and an over-strict standard just produces manufactured disputes.
- Stay consistent once you choose. The number's real power is as a fixed reference the factory plans against; renegotiating it per shipment destroys its value.
If you are unsure where your category sits, that is a solvable question — send us your product and price point and we will help you land on a standard that is defensible rather than decorative.
How a pre-shipment inspection actually runs
The AQL machinery is applied at the gate described in our guide to inspection types — the pre-shipment inspection. The sequence looks like this:
- Trigger. Production is complete and most units are packed — the standard trigger is 100% produced and typically 80% or more packed.
- Random drawing. The inspector selects sample units across cartons and across the production run — not from the pallets the factory helpfully points at.
- Checking against the standard. Each unit is compared with the golden sample and specification; functional tests and measurements run as the standard requires.
- Classification. Every defect found is recorded and classified as critical, major or minor, with photos.
- Decision. Counts are compared with the accept/reject numbers, and the report states pass, fail or a conditional result.
Pass, fail, and the gray zone
A clean pass still deserves a read: the defect photos tell you what your customers will eventually see. A fail is leverage — the goods have not shipped, the factory is motivated, and your options (sort and rework, discount, reject) are all still on the table. The gray zone is the conditional result: minors over the accept number with majors clean, or a borderline classification dispute. This is where the quality of your defect classification list — agreed before production, not invented during the inspection — decides whether the conversation is short or ugly. Our defect classification guide covers how to write that list.
One more reading habit: track results across shipments. Acceptable defect counts drifting steadily upward within passing inspections are an early-warning signal that a process is degrading — visible long before a lot finally fails. A quality program that only reacts to failures is always one shipment behind.
Four misreadings worth unlearning
- "AQL 2.5 means 2.5% of my shipment can be defective." The value is an index tied to sampling probabilities, not an entitlement. Treat it as a budget and you will be systematically disappointed.
- "It passed, so there are no defects." There are almost always some. The question a pass answers is whether they are within the tolerance you chose.
- "Sampling replaces process control." It measures the outcome. If defect counts matter to your margin, in-process checks prevent the outcome instead of pricing it.
- "One AQL setting fits everything." Critical, major and minor are counted separately against separate values — and a critical failure ignores the accept numbers entirely.
None of this makes AQL weak. It makes AQL a decision rule — and decision rules work when you understand what they decide. Used well, one page of an inspection report tells you whether five thousand units should board a vessel, and gives you the evidence to defend that call to a factory, a marketplace or an insurer.
Frequently asked questions
Who chooses the AQL values — me or the inspection company?+
You do. Inspection companies apply your standard; good ones will advise a defensible default for your category and price point. Decide before the first inspection and write it into the QC standard, so results are comparable across shipments.
Does AQL apply to small orders?+
Yes, the tables scale down — small lots get small samples, which means wider uncertainty: one defect in a 13-unit sample moves the result more than one defect in 125. For very small lots, 100% checking can be more informative than sampling.
What happens after a failed inspection?+
The shipment holds. You request a root cause and corrective action plan from the factory, agree on sorting or rework, and schedule re-inspection. Treating a fail as the start of a corrective loop — not the end of a transaction — is what separates a quality program from a series of arguments.
Is AQL used during production too?+
The same sampling logic can be applied mid-run in a DUPRO, with smaller lot sizes. The benefit is catching drift early, when rework is cheap and the production calendar can still absorb a correction.
Where do the accept and reject numbers come from?+
From published master tables in ISO 2859-1 and ANSI/ASQ Z1.4. You do not calculate them; you look them up from lot size, inspection level and AQL value. Inspection reports should state all three so you can verify the arithmetic.
