The AQL sampling plan is what you actually book when you order a final random inspection: an agreed rule for how many units get checked and how many defects the lot may carry before it is held. Get the lot definition, the sample size, the defect classes, and the acceptance numbers in writing before inspection day. Everything else on site then becomes execution rather than negotiation.
Key takeaways
- A sampling plan is a commercial agreement on sample size, defect classes, and acceptance numbers, not an internal detail of the inspection team.
- Sample sizes grow with lot size but never proportionally, which is why the code-letter tables exist and why guessing a sample count usually gets it wrong.
- Defect classes carry different AQL levels because a safety fault and a cosmetic thread end do not carry the same commercial weight.
- Acceptance numbers are read per classification, so one failed class holds the lot even when the other classes pass comfortably.
- The plan works only when carton selection is genuinely random and the lot is fully packed, since sampling theory assumes every unit had a chance of selection [1].
Importers often treat the sampling plan as the inspection company's internal detail. In practice the plan is a commercial agreement: it states how many units will be checked, how defects will be classified, and how many defects the lot can carry before it is held. Booking an inspection without agreeing on these points leaves the release decision to improvisation.
The underlying method is acceptance sampling, a mature statistical approach for judging lots by examining a representative part of them. Its concepts of sample size, defect classification, and acceptance numbers give buyers and factories a shared language, so inspection day produces a decision both sides can read from the same table.
This article walks through what importers actually book: the lot definition, the sample code letter logic, the standard AQL levels by defect class, and the acceptance arithmetic that turns counts into pass, fail, or hold outcomes.
Start with the lot definition
A sampling plan starts with the lot: the defined population the verdict applies to. For final random inspection the lot is normally one purchase order, or one clearly identified part of it, fully produced and packed at the time of inspection. Mixing two orders into one verdict, or inspecting a half-packed lot, breaks the logic because the sample no longer represents a defined population.
Lot size drives the sample size through code-letter tables. Small lots need proportionally larger samples to give any confidence, while very large lots gain little from checking thousands more units. The widely used single-sampling tables formalize this relationship, and their structure and operating logic are documented in the literature on acceptance sampling [1]. Buyers do not need to memorize the tables, but they should confirm which inspection level and which tables their provider applies.
The general inspection levels most buyers encounter run from reduced checking for stable suppliers through the normal level used by default, up to tightened checking after repeated failures. Special levels exist for checks that are destructive or very time-consuming, where smaller samples are accepted with eyes open. Agree the level at booking; changing it on inspection day changes the meaning of the verdict.
A standard booking confirmation for a UTS pre-shipment inspection therefore names the lot, the lot size, the inspection level, the resulting sample size, and the AQL per defect class: UTS pre-shipment inspection
AQL levels and what the numbers mean
AQL stands for acceptance quality limit: the worst average defect rate the sampling plan tolerates for a given defect class. Lower numbers mean stricter control. Critical defects normally carry no allowance at all, major defects typically sit in the low single digits, and minor defects carry higher allowances that reflect their limited effect on use or sale.
Common commercial practice clusters around a few familiar pairs, often a stricter limit for majors and a looser one for minors, with criticals judged at zero tolerance. The exact pair should be written into the purchase order or quality agreement rather than left to the inspector's habit, because factories plan their internal checks against the numbers they were given.
Read the number together with the sample size. The same AQL allows different absolute defect counts depending on how many units were examined, which is why the sampling table pairs each code letter and AQL with a concrete acceptance number. Two inspections that quote the same AQL but examined different sample sizes are not applying the same rule.
Quality management thinking supports this discipline: consistent criteria, documented evidence, and corrective follow-through are core themes of the ISO 9000 family of standards [2], and writing the AQL into the order is how a small buying team applies those themes without building a bureaucracy.
Defect classes and acceptance logic
| Defect class | Typical AQL direction | How the verdict works |
|---|---|---|
| Critical defects | Zero tolerance in normal practice | Any confirmed critical normally holds the shipment regardless of counts |
| Major defects | Strict limit, low single digits | Sample count above the acceptance number holds the lot for rework |
| Minor defects | Higher allowance than majors | Sample count above the acceptance number holds the lot, often with sorting |
| Quantity and assortment | Agreed tolerance in the order | Short, over, or wrong-ratio findings hold shipment until reconciled |
| Labels and markings | Agreed specification, no sampling allowance on wording | Wrong content or unscannable codes hold the affected population |
Read each row across: the class decides the strictness, the AQL sets the tolerated rate, and the acceptance number gives the concrete hold trigger for the examined sample.
Booking checklist before shipment
- Define the lot in writingState the order number, the product, the total lot quantity, and whether the verdict covers the whole order or a defined part of it. Ambiguous lot definitions produce verdicts nobody can apply.
- Fix level, size, and limitsConfirm the inspection level, the sample size, and the AQL per defect class, and write all three into the booking confirmation. Verbal agreement on strictness evaporates under deadline pressure.
- Share samples and documents earlySend the approved sample reference, the specification sheets, the packing list, and the assortment breakdown before inspection day. Inspectors cannot verify ratios or artwork they never received.
- Demand a genuinely inspectable lotRequire that the lot is fully packed and accessible, with cartons stacked so any pallet can be reached. Half-packed lots and blocked aisles force clustered selection that undermines random draws.
- Leave room for correctionReserve days between inspection and container loading for rework and re-inspection. A plan without a correction window turns every failure into an argument for shipping anyway.
One caution buyers learn the hard way: never negotiate the sampling plan after seeing the findings. Loosening an AQL or redefining the lot to make a failed lot pass does not change the goods; it only transfers the discovery to the retailer intake or the end customer, where the cost is higher and the evidence trail is weaker. Hold the line, require correction, and re-inspect.
Sources
- Wikipedia: Acceptance sampling —— Overview of acceptance sampling including AQL concepts used in product inspection.
- Wikipedia: ISO 9000 —— Overview of the ISO 9000 family of quality management standards.
Frequently asked questions
What information does the inspection team need at booking?
Give the inspector the purchase order quantity, the packed quantity, the product description, the assortment breakdown, the destination market, and the approved sample or specification reference. Missing assortment detail is the most common booking gap, because the inspector cannot verify a size ratio nobody shared.
Should I use different AQL levels for different defect classes?
Tighten the AQL for the classification that drives your returns: often major defects for functional goods or workmanship for visible goods. Loosening an AQL to make a lot pass does not change what the end customer receives; it only moves the discovery point to the returns desk.
Do I need the same sampling plan for repeat orders?
Yes, because procedures drift between orders: operators change, substitute materials arrive, and packing teams improvise near deadlines. Repeating the same sampling plan creates comparable records, so trends in defect rates become visible instead of anecdotal.
What happens when findings exceed the acceptance number?
Ask for correction of the systematic cause across the affected population plus a re-inspection of the corrected scope, rather than replacement of the sampled pieces alone. A defect found by random sampling usually runs wider than the sample, so fixing only the pieces the inspector touched leaves the lot risk unchanged.