AQL Is a Sampling Rule, and a Sampling Rule Needs a Clause

What AQL Specifies, and What It Does Not

An acceptable quality limit is a statistical statement about a lot. It says: given this lot size and this sampling plan, this is the defect rate at which the lot will usually be accepted. It is a producer's risk statement, not a promise about the goods in the container.

The distinction matters because buyers use the number as a quality target. A supplier told that the AQL is 2.5 reads it as permission to ship 2.5 percent defective, and in a sense that reading is correct: a lot at that level will pass the plan more often than it fails. What the buyer meant was usually something closer to zero, with a sampling plan that catches the bad runs.

The second thing AQL does not do is define a defect. The standard supplies the statistics; the parties supply the definitions. Two buyers can run inspections at the same level on the same product and reach opposite conclusions, because one counted a scuff as minor and the other counted it as major. The standard never adjudicates that. The clause does.

Critical, Major, Minor: Three Buckets

Defects are grouped by consequence, and the grouping drives the acceptance number. A critical defect is one that makes the product unsafe or illegal to sell: a live electrical fault, a banned substance, a choking hazard on a children's item, a missing regulatory mark where the mark is mandatory. Critical is normally set at zero, meaning one occurrence rejects the lot.

A major defect makes the product unusable or unsaleable as intended: a function that does not work, a component missing, a size outside tolerance, a print so far off register that the product cannot be displayed. A minor defect is a deviation that does not affect function or saleability: a scuff on a hidden surface, a colour variation within a stated range, a label slightly off centre.

Common splits are 0 critical, 2.5 major, 4.0 minor. That is a convention taken from general consumer goods inspection, and it is a starting point rather than a recommendation. A promotional item at a low price point and a medical accessory do not belong on the same numbers, and neither belongs on the same defect list.

The useful work is in the list, not the numbers. Write down, for your product, what counts as a major defect. Ten or fifteen lines is enough, and those lines are what an inspector applies on the floor.

Inspection Levels and the Default

The standard offers general inspection levels I, II and III, and special levels S1 to S4. The level sets how many units are drawn for a given lot size. Level II is the default in nearly all consumer goods inspection and the level a buyer gets when nothing is specified.

Level I samples less and accepts more risk. Level III samples more and is used where the consequence of a bad lot is high, or where the process is known to be unstable. The special levels sample very small quantities and exist for destructive or expensive tests rather than for judging a production run.

Where the level is not written down, the inspector chooses, and the inspector will choose Level II because it is the convention. That is usually fine. It becomes a problem when the buyer later argues that the sample was too small, because the sample size was determined by a level nobody agreed to.

Write the level into the purchase order next to the AQL values. One line: general inspection level II, critical 0, major 2.5, minor 4.0, with the defect list attached. That line converts a conversation into an instruction.

Reading the Table Without Memorising It

The table looks intimidating and works in three steps. First, the lot size gives you a code letter: a range of quantities maps to a letter of the alphabet. Second, the code letter plus the inspection level gives you a sample size, in units. Third, the sample size plus the AQL value gives you two numbers, accept and reject.

The logic is worth understanding rather than memorising. For a given lot, the sample size is fixed by the code letter, so a larger lot does not mean a proportionally larger sample. One thousand units and ten thousand units in the same code band draw the same number of pieces. That is why the plan detects a bad lot rather than measuring the defect rate precisely.

The accept and reject numbers are usually adjacent, such as accept 3 and reject 4. There is no grey zone and no negotiation inside the sample: if the count of defects reaches the reject number, the lot is rejected. That bluntness is a feature. It gives both parties one number to point at.

Ask the inspection company to state, in the report, the lot size, the code letter, the sample size, the defect counts by class, and the acceptance numbers used. A report showing only a result cannot be checked afterwards, by you or by anyone else.

The Clause That Survives a Dispute

A quality clause will be tested in exactly one situation: when the goods are not what you expected and the supplier disagrees. Six elements decide whether it holds.

The standard and the level. Name the inspection standard and the version, the general inspection level, and the AQL value for each defect class. Vagueness here is fatal, because the inspector will otherwise apply a default you did not choose.

The defect list. Attached to the order, specific to the product, with photographs where a description is ambiguous. This is the part buyers skip and the part that decides most disputes.

The measurement method. For dimensional checks, the instrument, the tolerance and the measurement point. For colour, the reference standard and the lighting condition. For function, the test procedure and the number of cycles.

Who inspects. A named third party, or the buyer's representative. Self-inspection by the factory is a production control, not an acceptance test.

When. Before packing, or after packing but before loading, with the goods available for a re-check. An inspection after the container is sealed has no remedy attached.

What happens on failure. Rework and re-inspection at whose cost, a replacement run, a price adjustment, or cancellation with the deposit returned. Name the remedy and the window for the supplier to respond. A clause that states a limit without a consequence is a suggestion.

Enforcing It Before the Container Sails

The value of a failed inspection is entirely a function of timing. A failure found while the goods are still on the factory floor is a rework instruction. A failure found after loading is a claim, and a claim against a supplier who has already been paid is a long conversation.

Book the inspection so that it finishes at least a week before the loading date. That gap is what allows a rework, a re-inspection and a rebooking without missing the sailing. Inspections booked for the day before loading are common and convert every finding into a delay.

When the report fails, respond in writing within the window the order allows, reference the report number, and state the remedy you are choosing. Then agree a re-inspection date before the rework starts, not after. Suppliers prioritise work with a date attached to it.

Keep three things from every inspection: the report, the photograph set, and the correspondence that followed. The report proves the finding. The photographs prove the condition. The correspondence proves you raised it in time, which is the element most often missing when a claim is refused.

What a Sampling Plan Cannot Tell You

A sampling plan is a filter on visual and dimensional defects. It is a poor instrument for four other things, and buyers routinely expect it to cover them.

Function over time. A sample tested once does not predict a product that fails after fifty cycles or three months of use. That is a reliability question and it needs a different test, run longer, often at a laboratory.

Regulatory compliance. An inspector checks against the specification you wrote. If the specification does not name the applicable regulation, the inspection will not test it, and a compliant-looking shipment can still be refused at the border. Compliance testing belongs to a laboratory, not to a pre-shipment inspection.

Packaging and labelling in the destination language. Inspectors check what the order says. If the order does not specify the barcode format, the warning text, the country of origin mark and the importer address, those are not checked, and each of them can stop a container.

The lot that was not sampled. A plan inspects one lot. If the factory ships three lots from three lines, one report covers one of them. Where a run is split across sites or weeks, the inspection plan needs to split with it, and that is a scheduling decision rather than a statistical one.

FAQ