Essay · The Contrapuntist
How to scope quality control for mixed SKUs: part 11
A purchase order with twelve SKUs is not one inspection job but twelve small ones sharing a single visit. Pooling them into one blended sample hides the weakest SKU behind the average of the group, which is exactly how a shipment passes overall while one variant fails at the customer. The correct scope treats each SKU as its own lot for acceptance purposes while sharing logistics, documentation review, and reporting. Quality management practice supports this structure: controlled identification and traceability of distinct product versions is a long-standing expectation of systems built around ISO 9000 ideas.
Key takeaways
- Treat each SKU as a separate lot for acceptance decisions, even when the SKUs share one inspection visit. [1]
- Allocate sample sizes per SKU by quantity and risk, not by equal splits that starve the largest lots.
- Require per-SKU defect counts in the report so one bad variant cannot hide inside a pooled average.
- Route material-level claims such as composition or restricted substances to accredited laboratory evidence. [2]
- Record the allocation logic in the booking so the factory prepares all variants and the inspector samples each one.
Mixed-SKU orders dominate small and mid-size importing: a dozen colorways, three sizes, two materials, one container. The commercial logic is impeccable and the quality logic is treacherous, because every sampling plan ever written assumes a homogeneous lot, and a mixed carton range is the opposite of homogeneous.
Part 11 of this series deals with the scoping decision that precedes all fieldwork: how to divide a fixed inspection budget across unequal SKUs. Get the division right and the report reads like twelve clean verdicts. Get it wrong and the report reads like one comfortable average with the problems averaged out.
The core ideas are borrowed from two mature systems. Quality management thinking insists that distinct product versions stay identifiable and traceable through production, and the accreditation system tells buyers where to turn when a claim needs laboratory-grade proof rather than a visual check.
Why pooled sampling flatters mixed orders
A single sample drawn across SKUs answers a question nobody asked: what is the average quality of the assortment. Buyers do not sell assortments; they sell individual variants, and a customer who ordered the blue lamp receives no comfort from the red lamp being flawless. Acceptance logic only works when the lot under verdict is the unit the customer will judge.
The distortion grows with imbalance. When one SKU fills half the order and nine others share the remainder, a proportional draw concentrates on the big variant while the small ones contribute a handful of pieces each. A defect rate of thirty percent in a fifty-piece SKU disappears inside a pooled result dominated by thousands of good pieces, yet that small SKU may be the entire stock of one retail listing.
With scoping help from a provider of third-party product inspection in China, buyers can split the booking into per-SKU lots with named sample sizes before the visit is confirmed. The sentence above describes ordinary good practice: the allocation is written down, the factory sees it, and nobody improvises the mathematics at the loading dock.
A practical allocation method
- List every SKU with quantitiesBreak the order into inspectable lots by model, color, size, and material, each with its finished and packed quantity.
- Rank by risk, not just volumeNew tooling, new materials, complex assembly, and high return rates earn larger samples than their quantities alone would suggest.
- Assign per-SKU sample sizesApply the chosen sampling plan to each SKU lot separately so every variant gets its own accept and reject numbers.
- Share what can be sharedDocumentation review, packaging checks, and container supervision stay common to the visit; only the acceptance verdicts stay separate.
- Write it into the bookingSend the factory the SKU split and the per-SKU sample sizes in advance so all variants are finished, packed, and accessible on the day.
Allocation patterns that work in the field
| Order pattern | Recommended split | Watch out for |
|---|---|---|
| One dominant SKU plus small ones | Full sample on the dominant SKU, reduced but standalone samples on each small SKU | Small SKUs skipped because their cartons were stored separately |
| Equal quantities, varied complexity | Equal base samples with extra pieces on the complex variants | Simple variants consuming time that complex ones needed |
| Same model, many colors | Color-by-color appearance checks with shared functional testing | Assuming shared tooling means shared color fastness |
| Two materials, one design | Separate material verification per variant with lab backup | Visual checks standing in for composition evidence |
| Accessories bundled with main goods | Standalone count and function checks on the accessory lot | Accessories treated as packaging and never sampled |
| Repeat order with one new SKU | Full attention on the new SKU, confirmation sampling on repeats | The new variant inheriting the old verdict without checks |
Traceability keeps the split honest
Per-SKU sampling only works if cartons, inner boxes, and pieces carry the SKU identity through packing. Mixed or mislabeled cartons force the inspector to sort before sampling, which burns visit time and undermines randomness. Ask the factory to segregate variants by carton marking before the inspection date, and make segregated storage a release condition for the visit itself.
The discipline mirrors a basic quality management expectation: product identification and traceability should survive every handling step [1]. When the factory's own system already separates versions cleanly, the inspector's per-SKU draw is quick and credible; when it does not, the booking has exposed a process weakness worth fixing before shipment.
Laboratory evidence follows the same split. A composition or safety test report belongs to the material and variant it names, and results from one SKU do not transfer to another that merely looks similar. Accreditation frameworks exist so buyers can check that a laboratory was competent for the specific test claimed [2], which is why per-SKU lab references beat a single generic certificate every time.
Write the SKU split into the booking before the visit. Sampling mathematics improvised at the factory always favors whoever is in a hurry.
Sources
- Wikipedia: ISO 9000 —— Overview of the ISO 9000 family of quality management standards.
- ILAC: International Laboratory Accreditation Cooperation —— ILAC homepage describing the international cooperation on laboratory and inspection body accreditation.
Frequently asked questions
Can small SKUs share one sample to save inspection time?
They can share visit logistics but should not share an acceptance verdict. Give each small SKU a reduced but standalone sample with its own accept numbers, and say so in the report. The cost of a few extra pieces per variant is trivial next to a full-listing recall of the one SKU that went unchecked.
How do I handle SKUs finished on different dates?
Scope them as separate inspection lots with separate visits or separate report sections. Sampling a half-finished SKU alongside finished ones mixes two production states into one verdict that describes neither. Book the second visit only for the outstanding variants rather than repeating the whole order.
What if the factory mixed variants in the same cartons?
Stop and re-scope before sampling. Either the factory re-sorts and segregates by SKU, or the inspection converts to a sorting supervision with new acceptance rules. Drawing a statistical sample from mixed cartons produces numbers without a defined lot, which cannot support any verdict.
Do I need lab tests for every SKU?
No, but you need lab evidence for every distinct material claim. Variants sharing one certified material batch can reference the same report; variants with different suppliers, dyes, or coatings need their own. Keep a matrix mapping each SKU to its material evidence so gaps are visible before booking.
Counter-arguments worth your Tuesday morning.
One free essay in your inbox each week — and a long-form piece for paying members every Friday. No ads, no nonsense.
Subscribe to the Newsletter →