AQL 2.5 / 4.0 explained for buyers: what a pre-shipment inspection does and doesn't guarantee
AQL is the worst process average a sampling plan will accept most of the time, not a defect-rate promise. How ANSI/ASQ Z1.4 turns lot size into a sample and an accept number, worked for 125- and 200-piece samples, and the three things a passed inspection does not tell you.
An AQL of 2.5 does not mean your shipment is 2.5% defective, and it does not mean it is less than 2.5% defective. It means the inspector drew a fixed sample from the packed goods, counted defects against a checklist, and the count came in at or under a number the standard sets for that lot size. That is a decision about whether to release the lot, made on a sample. It is a good decision rule, and it is worth understanding exactly what it is and what it is not before you read your first inspection report.
This post covers the common default in pre-shipment inspection, and the one we use: ANSI/ASQ Z1.4, single sampling, normal inspection, general inspection level II, with AQL 2.5 for major defects and 4.0 for minor. It is the fourth of six checkpoints in the quality system described on our quality and inspection page, and the pre-shipment inspection is where the system is tested, not where it starts. For how the checkpoints fit together, the supplier quality guide is the full picture.
What the standard is and what the default plan means
ANSI/ASQ Z1.4 is a table-driven sampling standard. You enter it with two facts, the size of the lot and the inspection level, and it hands you a sample size code letter. You then take that letter and the AQL you have chosen to a second table, which gives you the sample size, an accept number (Ac) and a reject number (Re). Count the defects in the sample. At or under Ac, the lot is accepted. At or over Re, it is rejected.
“Single sampling” means one sample and one decision; the standard also has double and multiple plans, which are rarely used in pre-shipment work. “Normal inspection” is the starting severity; the standard has switching rules to tightened inspection after a run of rejections and to reduced inspection after a run of acceptances. “General inspection level II” is the default discrimination level and the one assumed whenever a quote says “AQL 2.5/4.0” without saying more. Level I draws a smaller sample and is less discriminating; level III draws a larger one.
The AQL itself is defined by the standard as the worst tolerable process average when a continuing series of lots is submitted for acceptance sampling. In plain terms: if the factory’s process is running at the AQL, this plan will pass its lots most of the time. If the process is running well above the AQL, the plan will fail its lots most of the time. In between, it is a probability, not a line.
Critical, major, minor, and the 0 / 2.5 / 4.0 pairing
Defects are classed before they are counted, and the classes are agreed with the factory before the first PO, not argued over at the inspection.
A critical defect is one that makes the product unsafe or illegal to sell, or that would stop it working entirely. A sharp edge on a pet toy, a missing ground on a corded product, a wrong label on a food-contact item. A major defect is one that would likely cause the product to be returned or rejected by your customer: a non-functional feature, a dimension out of tolerance on a mating surface, a visible crack. A minor defect is a departure from spec that most customers would not return the product for: a light scuff on a non-visible face, a slightly off-shade carton print.
The common pairing is 0 for critical, 2.5 for major and 4.0 for minor. The “0” is a convention rather than an AQL from the table: it means the accept number for criticals is zero, so a single critical defect in the sample rejects the lot regardless of what else the sample shows. The 2.5 and 4.0 are looked up separately, each with its own accept and reject number, and the lot has to pass all three counts to be released.
Tighter pairings exist. A medical housing or an automotive sub-component might run 0 / 1.0 / 2.5, with the larger sample and smaller accept numbers that implies. The right AQL is a commercial decision about what a defect costs you downstream, and it is one of the first things we settle when we write the inspection plan for a part.
How lot size sets the sample and the accept number
Here is the general inspection level II path through the standard for four common lot sizes, single sampling, normal inspection. These are the published table values.
| Lot size | Code letter | Sample size | AQL 2.5 (major) Ac / Re | AQL 4.0 (minor) Ac / Re |
|---|---|---|---|---|
| 501–1,200 | J | 80 | 5 / 6 | 7 / 8 |
| 1,201–3,200 | K | 125 | 7 / 8 | 10 / 11 |
| 3,201–10,000 | L | 200 | 10 / 11 | 14 / 15 |
| 10,001–35,000 | M | 315 | 14 / 15 | 21 / 22 |
Read the K row as a worked example. Suppose a PO of 2,400 molded enclosures is packed and ready. The lot falls in 1,201–3,200, so the code letter is K and the sample is 125 pieces, drawn from cartons across the lot rather than from the top layer of the first pallet. The inspector checks all 125 against the checklist and the golden sample. If the count of major defects is 7 or fewer, the lot passes on majors; at 8 it fails. If the count of minors is 10 or fewer, it passes on minors; at 11 it fails. Any critical defect fails it outright.
Two things follow from the table that surprise people the first time. The sample does not scale with the lot: a lot of 10,000 is checked with 200 pieces, a lot of 2,400 with 125. And the accept numbers are not small. Seven major defects in 125 pieces is 5.6% of the sample, and the lot still passes. That is not a flaw in the plan. It is the plan doing what it was designed to do, which is to accept lots from a process running at or near 2.5% and to reject lots from a process running well above it, with a sample small enough to be affordable on every shipment.
What the accept number actually buys you
The honest way to read a sampling plan is as a curve, not a threshold. Take the K plan at AQL 2.5: 125 pieces, accept on 7. Straight binomial arithmetic on that plan, which is illustrative and not a property of any particular factory, gives roughly these odds:
A lot that is actually 2.5% defective passes about 99 times in 100. A lot that is actually 5% defective still passes about 7 times in 10. A lot that is 8% defective passes about 1 time in 5. A lot that is 10% defective passes about 6 times in 100.
So a passed inspection at AQL 2.5 tells you the lot is very unlikely to be badly out of control, and it tells you nothing like “this lot is under 2.5% defective.” A factory running at 5% majors will clear this plan most months and fail it some months, and the failures will look random to everyone involved. If you see that pattern on the monthly scorecard, the answer is not to argue about the inspector. It is to go back to the process, which is where the in-process check and the first article live.
What a pre-shipment inspection does not guarantee
It is a lot-acceptance decision, not 100% screening. Nobody looked at the other 2,275 pieces. The defects that exist in the unsampled cartons are still in the container when it is loaded. If your downstream cost of a single defect is high enough that you need every piece checked, that is a different service, priced per piece, and the right place to do it is at the factory before packing, not at your dock.
Defects outside the checklist are not counted. The inspector checks what the inspection plan says to check: the dimensions called out, the functions listed, the cosmetic standard agreed, the packing and marking spec. A problem that nobody wrote down, such as a material substitution that passes visual and dimensional checks but fails in the field, will not show up as a defect because the plan did not ask the question. The checklist is only as good as the spec it came from, which is why the golden sample and first-article inspection matter more than any single pre-shipment report. They are what the checklist is written against.
It happens too late to prevent anything. A pre-shipment inspection is done when 80–100% of the order is finished and packed. By then the material has been bought, the run is complete and the ship date is days away. A failed lot means sorting, rework, re-inspection or replacement, and a delay. The remediation policy agreed before the first PO says who pays and in what sequence, so the conversation is short, but the schedule still slips. That is the structural reason the quality system puts the factory audit, the first article and the in-process check at 20–30% of the run ahead of the pre-shipment inspection. By the time the inspector opens cartons, the question should be “did the process hold,” not “can this factory make the part.”
Reading the report
A good inspection report shows the lot size, the code letter and sample size actually drawn, the AQLs applied, the count in each defect class with a photo of each defect found, the result against Ac/Re for each class, and the carton numbers sampled. It also records on-site checks that are not sampled defects: quantity against the PO, carton markings, packing method, and a few function or dimension checks on a smaller sub-sample.
What you do with it is decide. The shipment does not leave until you release it, and a report that passes on the numbers but shows a pattern you do not like, say six majors all of the same kind, is a reason to call before releasing. A single kind of defect at six-in-125 is usually a process issue that will be worse next time, and catching it on the report is cheaper than catching it on the reorder.
If you want the plan tightened, say so before the next PO. Moving a part from 2.5 to 1.0 on majors shrinks the accept number and may change the sample size; the cost is a longer inspection day. The fee covers the inspection either way, as the pricing page sets out, so the decision is about what a defect costs you, not what the inspection costs.
What to do with this
- Find the AQL and inspection level on your current inspection reports. If the report does not state lot size, code letter and sample size, you cannot tell what the pass means. Ask for them.
- Write the defect classes down per part, with examples. What counts as major on a visible housing is minor on an internal bracket. If this is not agreed before the PO, the inspection will be argued after it.
- Treat a run of near-miss passes as a process signal. Five or six majors in a 125-piece sample, month after month, is a factory running above the AQL and getting lucky. Move the attention upstream, to the in-process check and the first article.
- Decide what a defect costs you before you decide the AQL. If a field failure costs a customer, 0 / 1.0 / 2.5 is worth the longer inspection. If a scuffed carton costs nothing, 4.0 on minors is fine. The FAQ covers the defaults and how to change them.
If the current factory has never been audited, that is the first thing a re-quote fixes. Send us the SKUs — ten business days, no fee, and the audit report is yours either way.