By DoraPony editorial team
A shipment inspection ends with a familiar sentence: the lot passed AQL. For a baby feeding brand, that statement can help authorize a commercial decision, but it leaves important questions unanswered. Which products were sampled? Which defects were counted? What happens to the units that nobody inspected?
Acceptance sampling gives buyers and factories a shared rule for deciding what to do with a submitted lot. Its value depends on defining that rule before inspection, selecting samples fairly, and responding consistently to the result. A passing sample does not establish that every bottle, collar, cup, or silicone component is free of defects.
This guide explains the purchasing decisions behind an attribute sampling plan. The examples are proposed controls for an OEM/ODM inspection brief, not universal acceptance limits for baby products.
What does an AQL pass actually mean?
NIST describes acceptance sampling as using a randomly selected sample to decide the disposition of a lot. Its main purpose is an acceptance decision, rather than an estimate of the lot's actual quality. That distinction matters when a buyer reads a report containing no observed defects.
AQL means acceptance quality limit in the ISO sampling framework. It is an input to a sampling scheme, not a promise about every delivered unit. A purchase specification containing only an AQL number leaves out the sample size, decision thresholds, defect definitions, inspection level, and rules governing the inspection state.
For current projects, identify the exact standard and edition. ISO 2859-1:2026 is the third edition for AQL-indexed attribute sampling schemes and replaces the 1999 edition and its amendments. The official overview describes single, double, and multiple schemes, together with switching and skip-lot provisions. Use the agreed full standard to implement its requirements; a copied table or supplier shorthand is not the complete scheme.
Keep the report's conclusion narrow: the submitted lot met the stated acceptance rule for the inspected characteristics. Avoid translating that sentence into a safety certification, a lifetime performance claim, or a measured percentage of defects in the entire shipment.
Why can a weak lot still pass?
Sampling involves uncertainty because most units remain unobserved. NIST's explanation of sampling plans distinguishes the producer's risk of rejecting quality at the chosen AQL from the consumer's risk of accepting a lot at a specified poor quality level. Both risks belong in the buyer's discussion.
An operating characteristic curve shows how the probability of acceptance changes as the underlying defect proportion changes. It helps explain why the same plan may accept some poor lots and reject some better lots. A zero acceptance number changes the decision threshold; it does not make the chance of an undetected defect disappear.
For illustration only, assume independent observations from a very large population with a fixed one-percent defect probability. The probability of finding zero defects in 50 observations is 0.99 raised to the power 50, approximately 60.5 percent. This calculation is not an ISO table recommendation or a proposed baby-product release plan. Finite lots and clustered defects require appropriate treatment.
The lesson is to discuss the protection a plan provides, not simply ask for a stricter-looking number. NIST explains choosing a plan through its acceptance probabilities. A qualified quality specialist can evaluate those probabilities against the project's commercial and product risks.
Which units belong in the sample?
Before an inspector opens cartons, define the submitted lot and make it physically accessible. Record its product configuration, quantity, production scope, and packing status. If the inspection covers one bottle size, it should not silently authorize another size packed nearby. Mixed revisions or materials need an explicit grouping decision.
A useful purchase brief describes how the inspector selects cartons and units without supplier substitution. Sampling only the easiest cartons to reach can miss variation elsewhere in the shipment. Plan access across the submitted stock and preserve the random selection procedure. Operational convenience should not decide which units have a chance of inspection.
For example, a brand ordering several lid colors could ask whether each color is a separate lot or belongs to an agreed combined lot. The answer affects sample allocation and the scope of the decision. Document it before results arrive, especially when a new pigment, decoration route, or assembly configuration introduces a distinct inspection concern.
Keep extra diagnostic samples separate from the formal acceptance sample. An engineer may deliberately collect unusual parts to investigate a molding issue. Those targeted parts answer a different question from a randomly selected shipment sample, and the report should make that distinction visible.
Conceptual illustration: sample selection should cover the defined lot rather than convenient cartons alone.
What counts as a defect?
A numerical plan is only useful when inspectors agree on what they are counting. Prepare a defect catalogue with the characteristic, examination method, acceptance boundary, and proposed severity classification. Agree whether the plan counts nonconforming units or individual nonconformities; these are different counting systems.
A collar with both a surface mark and a dimensional issue illustrates the problem. One inspector might record one affected unit; another might record two observations. Both descriptions can be useful, but they must feed the correct decision rule. Preserve individual findings even when the acceptance count uses a unit-based method.
| Inspection question | Detail to agree before sampling |
|---|---|
| Is a surface mark acceptable? | Viewing conditions, location, approved limit sample, and classification |
| Does the closure function correctly? | Assembly configuration, test method, conditioning, and result limit |
| Is the pack complete? | SKU-specific component list and approved packaging revision |
Use labelled reference photographs in the controlled inspection document and resolve borderline examples before shipment day. A mark on a decorative outer wall and damage to a functional interface may require different treatment. Classification should follow the product risk assessment and agreed specification rather than a generic internet defect list.
Illustrative component conditions support discussion; actual defect boundaries require approved product-specific references.
Which checks need a separate release gate?
Keep commercial appearance sampling separate from mandatory compliance evidence and controls for serious product hazards. AQL inspection does not establish chemical compliance or replace a required laboratory test. It also does not validate the design simply because the selected parts assembled successfully during a short inspection.
For the United States, CPSC's testing and certification guidance explains requirements for applicable products, including third-party testing at CPSC-accepted laboratories for relevant children's product safety requirements. Confirm the obligations and any applicable exceptions for the actual product with the responsible compliance team.
The factory release checklist can therefore have separate gates for material approval, required reports, process controls, functional checks, and shipment sampling. A pass in one column should not automatically clear another. If a serious defect is found, hold the affected stock for investigation under the agreed escalation procedure, regardless of whether an unrelated cosmetic count passes.
Specify destructive testing separately too. Define which specimens it consumes, what population they represent, and how tested units are segregated from saleable stock. A routine visual inspection sample is not automatically the right sample for every durability or laboratory evaluation.
What happens after a failed inspection?
Agree the rejection workflow before production starts. Identify who places stock on hold, who investigates the cause, who approves sorting or rework, and who authorizes a new submission. Retain the original result. Repeatedly drawing fresh samples until one passes undermines the meaning of the original decision.
After corrective work, document what changed and which stock was covered. A second inspection should follow the agreed resubmission procedure. If a screening method misses the original failure mode, sorting every unit may still leave an unresolved problem. Verify the correction and inspection method rather than relying on the size of the sorting effort.
Conceptual containment and investigation zones keep a rejected lot separate from unresolved rework.
Over successive deliveries, review inspection history and apply the chosen scheme's switching rules. Avoid reducing inspection just because a launch deadline is close. Use recurring findings to improve the manufacturing process and defect catalogue, so the shipment check becomes one part of a working quality system.
For your next DoraPony OEM/ODM project, bring the product configuration, destination market, proposed lot definition, appearance expectations, and functional requirements into the quality discussion early. Contact DoraPony to align the inspection brief, including sampling responsibilities, separate release gates, and the response to a failed lot before production begins.
