What an AQL Inspection Actually Tells You — and What It Cannot

Quick Answer

An AQL inspection answers one question: is this batch good enough to accept, given that nobody inspects every piece? It is a sampling rule, not a guarantee. AQL 2.5 does not mean “2.5% defects allowed”; it means a lot at that defect level has about a 95% chance of being accepted, and a lot noticeably worse than that has a shrinking chance of passing.

What it reliably catches: a batch that has genuinely gone wrong — a bad print run, a mould flashing, a colour lot off spec, a packing error. What it cannot catch: anything that only appears with use, over time, in transit, or on every thousandth piece. For that you need performance tests, not sampling.

Below: the actual sample sizes behind the numbers, which AQL to write into your order, what belongs on the defect list for drinkware, how to read the report, and what to do when a lot fails.

What Does AQL Actually Mean?

AQL stands for Acceptable Quality Limit. The standard behind it — ISO 2859-1, and the equivalent ANSI/ASQ Z1.4 — sets out how many pieces to inspect from a lot and how many defects are too many. The number you choose, the AQL, is the defect level the plan is designed to accept most of the time.

How Many Pieces Actually Get Checked?

Sample size is set by lot size, not by percentage. This is where most expectations go wrong: a 10,000-piece order is not checked by opening 250 cartons. Under General Inspection Level II — the default for consumer goods — the numbers look like this.

Read the third column carefully: at 3,201–10,000 pieces, 200 pieces are checked and up to 10 majors are allowed. Ten defects pass. If ten defective pieces in 200 is not acceptable to you, AQL 2.5 is the wrong number for your product — the fix is a lower AQL, not a longer argument at the end of production.

Why Does “Passed at 2.5” Not Mean “Under 2.5% Are Bad”?

Because the plan is built to accept lots at that level most of the time. A lot with exactly 2.5% defective has roughly a 95% chance of passing. A lot with 5% defective still passes fairly often. The protection is real but it is statistical, and it loosens as the sample gets smaller.

Worked the other way: with 80 pieces inspected at Ac 5, a lot with 2% defective has a good chance of showing 1–2 defects and passing quietly. A lot with 8% defective will usually show 6 or more and fail. Inspection is good at catching a batch that has drifted badly and weak at catching a mild drift — which is precisely why process control inside the factory matters more than the final report.

Which AQL Should Be Written Into Your Order?

There is no universal right answer, but there is a defensible default for drinkware: critical 0, major 2.5, minor 4.0, tightened to 1.5 / 2.5 when the logo or the gift presentation is the product.

When Should You Tighten the Level?

Tighten one step — for example from 2.5 to 1.5 major — when any of these is true: the product carries your brand promise on the outside, the order is going straight to a retail or marketplace channel that charges back, the previous shipment had a complaint, or the product is part of a gift set where two imperfect pieces make the whole set unsellable.

Do the opposite when the goods are promotional, low-value, or going to a market where a small cosmetic variation is normal. Tightening everything is not rigour — it produces inspections that fail for cosmetic reasons, rework that damages good stock, and a team that stops taking the standard seriously.

Who Pays for the Inspection?

Inside the factory price, in-house inspection is our cost. A third-party inspection is normally the buyer's cost, and it should be quoted that way rather than absorbed invisibly. What must be agreed in advance is who pays for sorting and rework after a fail — normally the party whose process caused the defect, and never a surprise at the end.

One practical note: the number has to match the defect definition. “Scratch” at AQL 2.5 means nothing until you say how long, how deep, and whether it is visible from 60 cm in normal light. A defect list without measurement rules gets renegotiated on the inspection floor.

What Belongs on the Defect List?

For drinkware, these are the items that actually show up. Use the list as a starting point and delete what does not apply to your product.

What Can an Inspection Not See?

Six things, and every one of them has caused a return that a passing report could not have prevented.

• Performance over time. A seal that holds today and fails after twenty dishwasher cycles passes every inspection.
• Transit damage. Carton strength, pallet pattern and humidity are tested by shipping, not by opening boxes in the factory. Ask for a drop test on the packed carton instead.
• Batch drift after the inspection. The last 500 pieces produced after the inspector left are not covered by the report.
• Fit with parts from another supplier. Your lid from supplier A and body from supplier B are each in spec and still may not seal together.
• Low-frequency defects. A 1-in-2,000 fault has a good chance of never appearing in a 200-piece sample.
• Anything about the material. Whether the resin is the grade you paid for needs a lab test, not a visual check.

The answer to all six is the same: put the specific risk into a test, not into a bigger sample. If it matters, measure it deliberately — drop the carton, run the dishwasher cycles, pull the print, measure the wall.

How Is an Inspection Actually Run?

There are three points where checking happens, and they catch different things. Most buyers only pay for the third.

Practical rules that save money: book the pre-shipment inspection when production is finished and packed, not 80% through — inspecting unpacked goods misses carton and labelling faults. Give the inspector your defect list and your AQL, in writing, or you will get their default. And if you use a third-party inspector, tell the factory in advance: an unannounced inspection in a running plant usually means a wasted day for everyone.

How Do You Read the Report?

A usable report contains six things. If any of them is missing, the conclusion is not evidence.

1. Lot size and sample size — how many pieces were in the delivery and how many were opened.
2. The AQL level and the standard used — 2.5 or 4.0, Level II, normal severity. Without this the accept/reject decision is not checkable.
3. Defect counts split by class — critical, major, minor, with a description of each.
4. Photographs of the actual defects — not stock photos of the product.
5. Measurements where dimensions matter — with the instrument and the reading.
6. A clear conclusion — pass, fail, or pass with a remark, plus the inspector's name and the date.

What Happens When a Lot Fails?

A failed inspection is a decision point, not a disaster. Four options, in order of cost:

1. 100% sorting. Every piece is checked and the defects are pulled. Workable when the defect rate is low and the defect is visible. Agree who pays for the sorting time before it starts — normally the party responsible for the defect.
2. Rework. Only for defects that can genuinely be fixed without making the product worse: a reprint, a re-pack, a trimmed flash line. Cosmetically reworked goods should be re-inspected, not assumed fixed.
3. Re-inspect at tightened severity. If the lot failed narrowly, a second inspection at a stricter level decides whether it was a bad sample or a bad lot. Do not simply re-run the same inspection until it passes — that turns the standard into decoration.
4. Reject the lot. For critical defects, a wrong material, or a defect that cannot be sorted economically. In that case the conversation moves to the order terms, not the inspection report.

Whichever route you take, get the cause written down and make the next inspection target it. The most expensive quality failure is not the one that was caught — it is the one that shows up again on the following PO because nobody wrote down what caused it. If the dispute ends up being about who is responsible, the order process page sets out where approvals sit; if you are still choosing a supplier, the quality control page shows how we structure the three inspection stages.

Related reading

Does Your Custom Cup Need a New Mould?

The arithmetic that decides whether to pay for tooling, and what actually moves a mould price.

Why Your Cup Colour Doesn't Match the Chip

The four renderings of one Pantone number, what ΔE measures, and how to approve colour properly.

Quality Control and Testing

The three inspection stages, the sampling table we run, the six finished-goods tests, and the paperwork you receive.