Common Challenges in Setting sample sizes and acceptance criteria
A sample size cannot be chosen sensibly without defining the lot, characteristic, inspection method, defect severity, acceptance rule, and consequences of a wrong decision. Larger samples may increase the chance of detecting some problems but add time and cost; they still do not inspect every unit. Buyers should agree the plan before production is ready.
Key takeaways
- Define the lot and the exact characteristic to evaluate.
- Separate critical, major, and minor findings with objective criteria.
- Choose a sampling scheme suited to the decision and contract.
- State acceptance numbers, methods, and treatment of borderline results.
- Document sample limits and actions for accepted, rejected, or inconclusive lots.
Sampling is a decision procedure with assumptions and risks; it is not a guarantee that every item meets requirements.
1. Specify the lot before calculating a sample
Define which units form one inspection lot: same product, revision, production period, site, and process status. State lot size, carton or pallet structure, mixed variants, and whether goods are complete and accessible. If products from different lines or batches are combined, decide whether to inspect them separately.
A sample plan applied to an unclear lot can produce a misleading result. Identify where units are located, how they can be selected, and whether selected cartons will represent the entire quantity. Record excluded goods and the reason for exclusion.
2. Identify the decision and the risk of error
Ask what the inspection result must support: accept, reject, hold for investigation, sort, rework, or request another sample. Consider the impact of accepting defective product and the impact of rejecting a conforming lot. Higher-consequence characteristics may require different control than minor appearance preferences.
Define who accepts the decision risk and who may approve a deviation. A sampling plan should reflect buyer requirements and supplier agreement, not be selected solely because a table or software makes it convenient.
3. Write objective acceptance criteria
For each characteristic, define requirement, measurement method, units, instrument, locations, sample condition, and tolerance or pass condition. For visual checks, specify defect examples, viewing setup, severity categories, and how borderline cases are escalated.
Separate product specification from defect classification. Define critical, major, and minor categories only when their effect is understood and the buyer can explain why a condition belongs in that class. Avoid vague criteria such as acceptable quality or workmanship is good without examples or thresholds.
4. Choose attribute or variable measurement carefully
Attribute inspection classifies observations as conforming or nonconforming according to set rules. Variables inspection measures a quantity such as length, thickness, or weight and compares it with limits. The appropriate method depends on characteristic, measurement reliability, distribution assumptions, lot structure, and the governing specification.
Do not treat a numeric measurement as inherently more reliable. Check instrument resolution, calibration, operator technique, and measurement location. For destructive or expensive tests, define specimen selection, sample custody, and how test results apply to the lot.
5. Select a sampling standard and parameters
ISO 2859-1:2026 describes AQL-indexed sampling procedures for inspection by attributes. If adopted, specify the edition, inspection level, lot size basis, AQL values, switching rules, and normal, tightened, or reduced status as applicable. Confirm contract fit and applicability before using its tables. [1]
AQL is a parameter in a sampling system; it is not a promise that a lot contains exactly that defect rate or that all accepted goods are defect-free. The buyer should understand the plan’s acceptance and rejection behavior and obtain qualified statistical advice for consequential decisions.
6. Select sample size and sample locations
Use the chosen scheme and lot definition to determine sample size, then state how cartons, pallets, production dates, and variants are selected. Random or otherwise justified selection reduces the chance that only easy-to-reach or best-looking items enter the sample.
If products vary across line, shift, material lot, or subcontractor, consider stratifying selection so the sample covers relevant sources. More observations from one convenient carton do not necessarily represent a diverse lot. Record the selection method and any access limitation.
7. Handle defect counts and borderline results
Specify acceptance number and rejection number, or the measurement decision rule, before inspection. Define how repeated defects on one unit are counted, how mixed severity is handled, and whether a failed critical characteristic overrides the lot-level plan. Avoid deciding after seeing results that a different threshold would be more convenient.
Set a process for borderline or inconsistent results: repeat measurement, second reviewer, reference sample, additional sample, or hold for technical decision. Document initial and repeat results. Do not silently average away a nonconforming measurement unless the approved method explicitly requires that calculation.
8. Distinguish sampling from 100 percent sorting
Sampling estimates or classifies a lot under a defined procedure; it does not mean every unit has been checked. If a critical issue requires sorting, define competent personnel, inspection method, marking of checked goods, segregation of failures, rework, and verification that uninspected units are not released.
Sorting itself can introduce errors or damage. Set production controls and records for accepted, rejected, repaired, and unverified quantities. Identify who pays for sorting and how the remaining lot is reinspected or dispositioned.
9. Tie result to report and release action
The inspection report should identify lot, product revision, sample basis, plan and edition, criteria, methods, observed results, defect counts, instruments, photos, limitations, and recommendation. Provide the report before the shipment cutoff so the buyer can investigate or correct issues.
Define what happens after accept, reject, or inconclusive outcomes. Shipment release should be made by the authorized buyer-side role after required documents and critical findings are reviewed. A passing sample does not remove the need for release controls stated in the contract.
10. Revisit the plan as evidence changes
Use receiving defects, complaints, supplier performance, process changes, and inspection history to review sample design. A stable process may justify a different documented plan; repeated failures or new risks may require stronger checks. Make changes prospectively and tell suppliers the effective date.
Buyers setting plans with Learn more can align sampling rules, criteria, reporting, and shipment actions before the final inspection is booked. Retain the rationale and approvals so later teams understand the plan’s assumptions and limits.
Sources
Frequently asked questions
Does a larger sample guarantee a good lot?
No. It changes the sampling decision behavior but does not inspect every unit or eliminate all acceptance risk.
Is AQL the same as an allowed defect percentage in every shipment?
No. It is a parameter in a sampling system, not a direct guarantee or exact lot defect allowance.
Can sample size be chosen after seeing defects?
Set the plan before inspection to avoid changing decision rules in response to observed results.
When is 100 percent sorting appropriate?
Consider it when risk requires checking every unit and a reliable method, competent staff, and traceable segregation are available.