All sourcing guides

Acceptance sampling guide

AQL inspection in China: sampling levels and defect limits

AQL inspection uses an agreed acceptance-sampling plan to inspect a random sample from a production lot and decide whether that lot meets the plan's acceptance criteria. The buyer must define the lot, referenced standard, inspection level, defect classifications and AQL values before inspection.

The quick answer: AQL is a sampling decision system, not a quality guarantee

A buyer usually cannot inspect every unit in a commercial order. Acceptance sampling addresses that constraint by selecting units from a defined lot, checking them against agreed requirements and using a sampling plan to support a lot-level decision. The plan links lot size, inspection level, sample-size code, selected AQL and acceptance and rejection numbers.

AQL does not mean that an agreed percentage of defective goods is automatically acceptable. It does not guarantee every unit, certify the product or replace laboratory testing, legal compliance review, process control or 100 percent inspection where the risk requires them.

The current ISO reference is ISO 2859-1:2026, the third edition for AQL-indexed sampling schemes for lot-by-lot inspection by attributes. It replaces the withdrawn 1999 edition. Buyers should confirm which standard and edition their provider or agreement uses instead of assuming every online calculator follows the same source.

What acceptance sampling means for a China order

Acceptance sampling is inspection of a sample from a lot to decide whether to accept that lot under a defined plan. In inspection by attributes, each checked characteristic is recorded as conforming or nonconforming, and defects or defective units are counted under agreed rules. This differs from monitoring a process average during production or testing every unit.

The lot might be one completed SKU, one colour and size combination, or a defined group produced under similar conditions. Buyers should decide whether different SKUs, factories, production dates or packaging configurations belong in one lot. Combining unlike goods merely to reduce inspection effort can weaken the sample's meaning.

Random selection matters. An inspector should not check only cartons placed nearest the door or units prepared by the supplier. If production is incomplete, cartons are inaccessible or selection is restricted, the report should record that limitation.

AQL is only one part of a wider pre-shipment inspection checklist. The inspection still needs specifications, approved samples, artwork, labels, packaging instructions, measurements, functions, quantity and shipment-readiness checks.

Inputs buyers must define before selecting a sample size

First define the lot: product, SKU or variant, total quantity, production status and carton distribution. If an order contains materially different products, ask whether each requires separate representation.

Then state the sampling standard and plan. The instruction should identify the standard or agreed method, plan type and inspection state. These choices should not be inferred after defects are found.

The inspection level links lot size to a sample-size reference under the chosen standard. It is not an AQL value. A smaller sample saves time but provides less evidence; a larger sample adds effort but still does not guarantee the whole lot.

Defect classification comes before counting. Critical, major and minor findings should reflect safety, legal, functional, commercial and customer consequences for the actual product and destination market.

Finally, the buyer must agree the AQL for each defect class and whether the plan counts defects, defective units or another measure. One unit can contain multiple defects, so inconsistent counting can change the apparent result.

Lot size, inspection level and sample-size code in buyer language

Lot size is the number of units covered by the acceptance decision. It should be a real, identifiable group available for selection. If a supplier says 10,000 units are complete but only 6,000 are packed and accessible, the report should not blur that difference.

The inspection level expresses how much inspection effort the plan applies relative to the lot. Under a referenced standard, lot-size range and inspection level lead to a code letter or equivalent reference, which then leads to a sample size under the selected plan. A sample size copied from an old report is not automatically correct for a new quantity, product mix or inspection state.

General and special inspection levels are plan inputs, not rankings from bad to good. A special level may fit a costly, slow or destructive test, while broader visual checks may use a general level. The choice should be made before inspection and documented.

Do not rely on a calculator that gives only a sample number without identifying the standard, edition and assumptions. Use the authorised standard or a qualified provider for sample sizes and acceptance/rejection numbers. This guide intentionally does not reproduce proprietary sampling tables.

Critical, major and minor defects: classify before counting

Critical defects are findings that create an unacceptable safety, legal or severe usability risk. The classification must be product-specific. A hazardous sharp edge, missing mandatory warning or dangerous failure may be critical, but applicable requirements should be confirmed with qualified specialists.

Major defects materially affect function, saleability, appearance or the agreed specification. Examples may include a failed primary function, wrong material, incorrect dimension, missing accessory or retail packaging that prevents sale.

Minor defects depart from the agreed workmanship standard without materially reducing normal use or saleability. A small mark may be minor on one product and major on a premium-facing surface. Price point, brand promise and the approved sample affect the decision.

A clear sample approval record helps because it shows which materials, colours, functions, dimensions and packaging details were accepted before production. Laboratory failures and legal requirements cannot be downgraded through an ordinary commercial inspection discussion.

Acceptance and rejection numbers: how the lot decision works

After the inputs are agreed, the authorised sampling plan provides acceptance and rejection criteria for each selected AQL and sample size. The inspector examines randomly selected units, records findings by defect class and compares totals with those criteria. Staying within the acceptance number produces a sampled pass under the plan; reaching the rejection number produces a sampled fail.

These are lot decisions under a sampling method, not automatic commands to pay or ship. A buyer may still hold a sampled pass if there is a critical issue, wrong product, incomplete quantity, failed test, unauthorised material change or missing compliance evidence. After a sampled fail, sorting, rework, replacement, 100 percent checking or reinspection may be more appropriate than immediate cancellation.

The report should show the lot, sample size, selection constraints, defect descriptions, photographs, criteria and result. A single word such as "pass" gives the buyer too little evidence.

A practical example without a universal AQL recommendation

Consider a custom household accessory produced in several retail colours. Before inspection, the buyer and provider decide whether the colours can form one lot or need separate representation, then confirm production is complete enough and cartons are accessible.

The buyer states the referenced standard, inspection level and plan. The authorised standard or qualified provider supplies the sample size and acceptance/rejection numbers. The buyer does not copy numbers from an unrelated apparel inspection or select them after seeing defects.

The checklist classifies a broken primary function and incorrect retail artwork as major, a small hidden cosmetic mark as minor, and escalates safety findings separately. Units are selected across colours and cartons, and findings are recorded by class.

If the major count reaches the rejection point, the buyer decides whether the supplier should sort the lot, rework packaging, replace goods or prepare for reinspection. A pass supports the shipment review but still does not prove every uninspected unit conforms.

Why AQL is not an allowed defect percentage

The phrase "AQL 2.5" is often misunderstood as permission for exactly 2.5 percent defective goods. AQL is an index used in an acceptance-sampling scheme for a continuing series of lots. The decision depends on the plan, sample size and acceptance/rejection criteria, not simply multiplying shipment quantity by a percentage.

Calling it an allowed percentage invites a supplier to treat the number as permission to ship known defects and encourages a buyer to assume a sampled pass predicts the exact full-lot defect rate. Neither conclusion follows from the sampled result.

The production target should be consistently better than the acceptance limit, supported by specifications, sample approval and corrective action. If material changes, delays or rework appear, production follow-up should carry those risks into inspection.

Normal, tightened and reduced inspection are not casual labels

ISO 2859-1 includes switching rules for a continuing series of lots. Normal inspection is generally the starting state. Tightened inspection can apply when results indicate deteriorating quality, while reduced inspection may become available when performance remains consistently satisfactory and the scheme's conditions are met.

These states should not change merely because the buyer wants lower cost or the supplier wants an easier pass. They belong to a defined scheme. A buyer handling isolated orders, changing products or irregular suppliers should ask whether continuing-lot switching rules fit the relationship.

Supplier capability and current-order evidence remain separate. Our guide to factory audits versus pre-shipment inspection explains that distinction. A larger sample is still sampling, and high-consequence defects may require a different control.

When AQL sampling is not enough

AQL sampling is useful for visible or testable attributes when a sample can support a commercial decision. It is weaker when the risk is hidden, intermittent, destructive to test or safety-critical.

Regulated products may require accredited laboratory testing, certification, technical documentation, traceability and legal review. An inspector can check visible markings or collect samples, but ordinary visual inspection does not prove compliance.

A product may look correct while using a different alloy, chemical formulation, fibre content, coating or internal component. Material identity may need supplier records and qualified testing. Brief function checks also may not predict durability, battery life, fatigue, leakage or colourfastness.

If one hazardous unit creates unacceptable harm, conventional sampling may not control the risk. Product design, process control, mistake-proofing, testing, traceability and possibly 100 percent checks may all be relevant. AQL also cannot repair a weak specification; start with a clear China quality-control plan.

What buyers should prepare before an AQL inspection

  • Define the production lot by SKU, variant, factory, batch and accessible quantity.
  • State the referenced sampling standard and edition, plan type and inspection state.
  • Agree the inspection level and AQL for each defect class before inspection.
  • Provide the approved sample, specifications, tolerances, artwork and packaging files.
  • Write product-specific critical, major and minor defect definitions.
  • Define the response to a pass, fail, incomplete sample or unexpected finding.

Also list checks that sit outside the AQL count, such as quantity, carton configuration, shipment readiness, documents, laboratory reports or destructive tests. Define who can approve the report and what evidence is required after rework.

What to do after a pass, fail or unclear result

After a pass, review whether the lot was complete, selection was reasonably random, the approved reference was available and all required checks were completed. Confirm packaging, quantity, documents and shipment readiness before release.

After a fail, request a written correction plan covering affected products, likely cause, sorting or rework, timing and evidence. The response may include 100 percent sorting, replacement, packaging correction, delayed shipment or reinspection.

An unclear result may come from inaccessible cartons, conflicting specifications, incomplete tests or an unexpected defect. Do not force an artificial pass or fail. Record the limitation and decide whether clarification, targeted evidence or another inspection is needed.

For teams without China-based staff, a China sourcing agent can keep the purchase order, approved sample, supplier updates, inspection scope and correction evidence connected. The value is applying the chosen method to the real order, not choosing a universal AQL number.

Final buyer decision

AQL inspection is useful when the buyer needs a repeatable way to sample a defined lot and compare observed defects with agreed criteria. Its value comes from disciplined inputs: the right lot, current standard, documented inspection level, product-specific defect classes, authorised plan values and random selection.

It should not be permission to ship known defects, proof that every unit is correct or a substitute for testing and compliance. The buyer still needs to consider critical findings, production changes, packaging, quantity, documents, payment leverage and the cost of discovering a problem after arrival.