Your inspection report comes back: AQL 2.5 / 4.0, 200 pieces sampled, 8 major defects found — PASS. Almost every first-time importer reads that as "the goods are fine". It does not say that. It says something much narrower, and knowing exactly what it says is the difference between accepting a shipment and arguing about one.
What AQL is not
The acronym is read as "acceptable quality level", which is where the misunderstanding starts. It sounds like a quality target the factory is promising to hit. It is not. AQL is a lot-acceptance rule: the defect rate at which a batch will still usually be accepted by the sampling plan.
The current standard is ISO 2859-1, whose third edition replaced the long-standing 1999 version. In North America the same tables and procedures appear as ANSI/ASQ Z1.4, so it makes no practical difference which one your inspection company names — the accept and reject numbers are the same.
Read literally, agreeing to AQL 2.5 for major defects means you have agreed in advance that a shipment running at roughly 2.5% major defects should normally pass. Not that it will be defect-free.
The three decisions that are actually yours
The table does the arithmetic. You make three choices before it can, and these are the ones worth spending time on.
- Inspection level. General Level II is the default for consumer goods and is what you will get unless you ask otherwise. Level I samples fewer units (cheaper, blunter), Level III samples more. The special levels S-1 to S-4 exist for checks that are slow or destructive — you do not cut open 200 units to verify a weld.
- AQL per defect class. The common convention for consumer goods is 0 for critical, 2.5 for major, 4.0 for minor. Zero for critical is not a formality: it means a single safety defect in the sample rejects the lot.
- What counts as major. This is the one that decides outcomes, and the one almost nobody writes down.
How the table actually gets used
Two lookups, in this order:
- Lot size + inspection level → a sample size code letter. A batch of 5,000 units at General Level II gives code letter L.
- Code letter + AQL → sample size and the accept/reject numbers. Code letter L means 200 units are inspected. At AQL 2.5, the plan accepts the lot on 10 or fewer defects found and rejects it on 11 or more.
We deliberately do not reproduce the full table here. While checking these figures we found a well-ranked inspection site with an entire column misaligned — it listed AQL 1.0 as accepting 10 defects in a sample of 200, which would be a 5% defect rate described as 1%. Take the numbers from the standard itself or from your inspection provider's calculator, not from an article.
The sample does not grow with your order
This surprises people. A 5,000-piece order and a 9,000-piece order both fall in the same lot-size band, so both are inspected at 200 units. Doubling your order does not double the scrutiny.
That is a property of sampling, not a shortcut by the inspector: the statistical confidence in a lot depends far more on the sample size than on the ratio of sample to lot. It also means the marginal cost of inspection falls as your orders grow, which is worth knowing when a factory tells you inspection is uneconomic at your volume.
What a pass actually means
Go back to the report at the top. Eight major defects in 200 units is a 4% major-defect rate in the sample. The lot passed because the plan allows up to 10. If the sample is representative, roughly 4% of everything in those cartons has a major defect.
On 5,000 units that is about 200 pieces. Whether that is acceptable is a commercial question, not a quality-control one, and it depends entirely on what happens next: 200 defective units in a wholesale pallet is an irritation, 200 defective units going out to Amazon customers is a listing suspension.
This is the single most useful thing to understand about AQL. It is a decision rule for accepting a shipment, not a measurement of how good the goods are. A pass is not a certificate; it is the absence of a reason to reject.
The definition of "major" is the real negotiation
Two inspectors, same standard, same product, different results — almost always because they classified defects differently. A loose thread: minor, or major? A shade variation against the approved sample: minor, or major? A missing warning label: major, or critical?
Nothing in the standard answers those questions. They are product-specific, and if you have not written them down, they are answered by whoever happens to be holding the clipboard.
The fix is unglamorous and works: agree a written defect classification before production, tied to the approved sample. List the defects that matter for your product and state which class each one falls into. It belongs in the purchase order, not in an email after the inspection failed. The sample approval process is where this list should be built, because that is when you and the factory are looking at the same physical object.
What the inspector needs on the table
An inspector arrives with the standard, a sample size and a checklist. What they do not arrive with is any knowledge of what you agreed with the factory. Everything in the previous section about defining "major" only works if the definition physically reaches the person doing the checking.
- The approved sample, sealed and labelled, on site. Not a photo of it. Colour, texture and fit disputes are settled by putting two objects next to each other, and by nothing else.
- The written defect classification, the one from your purchase order. Without it the inspector uses a generic list, and a generic list does not know that on your product a 2mm misalignment is cosmetic but a loose hinge is a return.
- Packing and labelling specification — barcodes, carton marks, inner and outer quantities, country-of-origin marking. A surprising share of rejected shipments fail on labelling rather than on the product.
- Anything that must be measured, with the tolerance written down. "Approximately 30cm" is not checkable; "300mm ±3mm" is.
The inspection is also only meaningful if there is something finished to inspect. Booking against your ship date rather than the date production and packing actually complete is the most common way to end up paying for a visit that inspects half an order and has to be repeated. The full list of what to specify, with the methods and tolerances that make each line judgeable, is in the pre-shipment inspection checklist.
The standard already has a rule for a factory that keeps failing
Most buyers treat each inspection as a standalone event: book it, read the verdict, ship or do not ship. The standard does not work that way. Both ISO 2859-1 and ANSI/ASQ Z1.4 contain switching rules that change the plan based on what happened on previous lots.
- Normal to tightened. If 2 lots out of any 5 consecutive lots are rejected, inspection switches to tightened — same sample size, lower accept numbers. The factory has to be visibly better to pass.
- Back to normal. Five consecutive lots accepted on original inspection returns you to the normal plan.
- Stop inspecting. If 10 consecutive lots are still on tightened inspection, the standard says sampling inspection should be discontinued pending action to improve the quality of what is being submitted.
That last rule is the one worth internalising, because it is the standard telling you something uncomfortable: when a factory cannot get off tightened inspection, the answer is not more inspection. Sampling cannot fix a supplier who is not capable of the specification. At that point the decision is a supplier decision, and continuing to inspect just spreads the cost of that decision over more shipments.
There is a movement in the other direction too — reduced inspection, with smaller samples, for a factory with a sustained clean record. The qualifying conditions differ slightly between the two standards, so ask your inspection provider which rule they apply rather than assuming.
In practice almost nobody uses any of this, because applying switching rules requires tracking lot history across orders and most inspections are booked one at a time as isolated jobs. If you reorder the same product from the same factory, keeping that history yourself costs nothing and tells you something no single report can: whether this factory is getting better, getting worse, or holding steady.
When to tighten, and when not to bother
Defaults are defaults because they suit the average case. Tighten when the case is not average:
- First order with a new factory. You have no history to lean on. Level III, or a tighter AQL on majors, for the first run only.
- Anything with a safety or certification dimension. Electricals, children's products, anything food-contact. Critical stays at zero and the definition of critical should be explicit.
- High cost of failure downstream. If a defect reaching a customer costs you a return, a refund and a review, the economics of a larger sample change completely.
Conversely, tightening everything on every order mostly buys you delays and argument. The useful move is usually not a stricter AQL but a clearer defect list.
How this fits into the wider process — when inspection happens relative to production and the shipping cut-off, and what to do when a lot fails — is covered on our quality inspection service page.
Which of the four inspection stages is worth buying for a given order is a separate decision, worked through in when to inspect a China order.
