AQL inspection in China: a practical guide

If you have ever received an inspection report from a factory in China, it almost certainly referenced AQL. It is the framework most pre-shipment inspections are built on, and understanding it is the difference between reading a report and being reassured by one.

What AQL is

AQL stands for Acceptable Quality Limit. It comes from a family of statistical sampling standards, the best known being ISO 2859 and its predecessors, that answer a narrow question: given a production run of a certain size, how many units should be checked, and how many defects in that sample mean the run should be rejected?

It is a sampling method, not a quality standard. AQL does not define what good looks like for your product. It defines how many units to look at and what number of problems is too many. What counts as a problem is something you have to specify.

Why brands use sampling rather than checking everything

Checking every unit in a run of several thousand is slow and expensive, and for many products it is not meaningfully more informative. Sampling accepts a defined, quantified risk of missing something in exchange for an inspection that can actually be completed before the goods ship.

The important word is quantified. A sampling plan tells you what risk you are taking. Someone glancing at a few cartons tells you nothing.

Critical, major and minor defects

AQL plans separate defects into classes, because a sharp edge and a crooked label are not the same problem. Classification is set before inspection, and it is where most of the useful thinking happens.

Tolerances are commercial decisions, agreed between buyer and supplier per product. A cosmetic scuff on a phone case and the same scuff on a gift box are not equivalent, and no standard can tell you which matters to your customer.

Inspection levels and what they change

Sampling standards define general inspection levels, usually labelled I, II and III. Level II is the common default. Level I looks at fewer units, level III at more.

Raising the level means a larger sample, more inspection time and a better chance of catching a low-rate defect. Lowering it means the opposite. Brands typically raise the level when a supplier is new, when a previous run had problems, or when the cost of a defect reaching customers is high.

How sample size is decided

You do not pick a sample size directly. It falls out of two inputs: the size of the production lot and the inspection level. Those give a sample-size code, and the code plus your chosen AQL figures give two numbers, the acceptance number and the rejection number.

The mechanism is simple once you see it. Inspect the sample. Count defects by class. If the count for a class is at or below its acceptance number, the lot passes on that class. If it reaches the rejection number, it fails. Sample size grows with lot size, but far more slowly, which is the whole point of sampling.

Neofulfill does not publish a fixed AQL level or defect tolerance as a universal policy, because the right figures depend on the product, the market and what the buyer has agreed with the supplier. They are set per project rather than inherited from a template.

Pre-shipment inspection, and when to do it

A pre-shipment inspection happens when production is substantially complete and the goods are packed or ready to pack. It is the last point at which a problem can be caught before the run becomes your inventory in another country.

Timing matters more than most brands expect. Inspect too early and much of the run does not exist yet. Inspect after the goods have shipped and you have swapped a production problem for a logistics one. A useful inspection also happens while there is still commercial leverage to get something corrected.

Inspection is normally more than a defect count: verifying quantity, checking that packaging and labelling match what the destination market requires, and confirming the product matches the approved sample. See quality control and production follow-up for how that runs alongside production.

What AQL cannot do

This is the part that gets left out of inspection reports.

• A pass is not a guarantee. Sampling means some units were not examined. A passed lot can still contain defects, and the plan quantifies that risk rather than removing it.
• It cannot judge a specification you never wrote. If nobody defined the acceptable colour range, an inspector cannot fail the run for being the wrong colour.
• It does not test durability. Visual and functional checks at inspection say nothing about how a product behaves after three months of use.
• It cannot fix a supplier problem. A repeatedly failing run is a sourcing decision, not an inspection one.
• It says nothing about consistency between runs. A good batch does not promise the next one.

Where QC sits in sourcing

Inspection is the checkpoint, not the system. What makes it useful is everything that happens before: a specification precise enough to inspect against, an approved sample that defines what correct looks like, and production follow-up so problems surface while they can still be corrected rather than at the final check.

That is the argument for keeping supplier selection, production follow-up and inspection connected rather than buying an inspection as a standalone service. When they run together, an inspection finding goes back to the supplier who caused it, with the leverage to get it addressed. See China sourcing for how the stages connect.