How to scope quality control for mixed SKUs: part 18

By admin
The short answer

Quality control for mixed SKUs fails when variety is averaged away. The scoping method that works is stratification: define homogeneous groups by product, material, or line, allocate sample and acceptance numbers to each group, verify and decide per group, and keep tallies that trace every carton to its stratum. Partial decisions then become possible, passing strata ship while failing ones are contained, instead of one blended verdict holding the whole container hostage. This eighteenth part of the series gives the full method, from drawing stratum lines to signing the per-stratum summary.

Key takeaways

  1. Mixed consignments must be divided into homogeneous strata before sampling, because a pooled sample across different products supports no clean verdict[1].
  2. Each stratum gets its own sample size and acceptance numbers, scaled to its quantity and risk rather than copied from single-SKU practice.
  3. Per-stratum verdicts allow partial decisions: ship the passing strata while the failing ones are sorted or re-checked.
  4. Documenting variety precisely, what is in the box and where, satisfies the same transparency that technical trade rules expect of conformity procedures[2].
  5. The tally discipline is doubled for mixed loads: per-carton counts during opening and per-stratum summaries against the packing list.

A single-SKU inspection asks one question of one population. A mixed-SKU consignment asks a dozen questions of a dozen populations that happen to share a container, and buyers who scope it as one big lot get verdicts that decide nothing. The discipline is stratification: divide the consignment into homogeneous groups, allocate sample and acceptance numbers to each, and let every stratum earn its own verdict. This article explains how to draw those lines before the inspector travels.

Sampling theory supports the approach directly: results from a sample generalise to the population the sample was drawn from, so a sample pooled across different products generalises to none of them cleanly[1]. Each SKU, material, or production line with its own defect risk deserves treatment as its own population. Trade practice reinforces the paperwork side: assorted consignments still need technical requirements met per product line without unnecessary friction, which starts with knowing what is actually in the box[2].

This eighteenth part of the scoping series covers stratum definition, sample allocation across strata, per-stratum verdicts and partial decisions, and the tally discipline that keeps assorted loads auditable. It assumes the sampling standard and AQL levels are already agreed; the subject here is how to spread them across variety without diluting them.

Draw the stratum lines before booking

A stratum is any subdivision whose defect risk differs from its neighbours: different model, different material batch, different production line, different colour where colour involves a regulated substance or a separate spray run. List them in the booking with quantities per stratum, and state explicitly which strata are in scope and which the buyer excludes in writing. Anything unlisted at booking becomes a loading-day argument, and loading day is the worst place to define populations.

Illustration: Draw the stratum lines before Decorative illustration for the section "Draw the stratum lines before"; visual only, carries no data.

Keep the strata count workable by grouping genuinely identical goods and splitting genuinely different ones. Colour-only variants of unregulated goods with shared construction can usually share a stratum, documented as such; the same model from two factories never can, since process risk differs by site. When the list grows past about eight strata, consider splitting the inspection rather than thinning every sample to a symbolic handful. A stratum too small to sample meaningfully should be 100 percent checked, excluded in writing, or given its own visit.

Allocate sample and verdicts per stratum

Acceptance numbers do not transfer between strata. A major-defect acceptance of fourteen for a three-hundred-piece sample means nothing applied to a fifty-piece draw from another model. Recomputing per stratum is mechanical work that the booking sheet should already contain, so the inspector verifies rather than improvises. Buyers who let strata share one pooled count have effectively inspected none of them.

Sample sizes that stay meaningful

Apply the sampling logic within each stratum: sample size follows the stratum quantity and the agreed level, not the consignment total[1]. Small strata therefore consume disproportionate inspection time, which is why the booking must show per-stratum samples and the resulting man-days. Surprises about duration on the day usually trace to a booking that quoted one lot and delivered ten.

Where a stratum is too small for its table sample, check all of it and record a 100 percent result rather than a sampling verdict. Full checks on tiny strata are cheap and decisive. What must never happen is treating ten checked pieces from a two-hundred-piece stratum as a verdict-grade sample when the plan required fifty.

Decide stratum by stratum

Each stratum earns its own pass, conditional, or fail against its acceptance numbers, and the consignment decision is assembled from those parts. Passing strata can ship while failing strata are sorted, reworked, or re-checked, provided the packing and loading plan can segregate them physically. This partial-decision power is the entire payoff of stratification: without it, one failing colour holds nine passing ones hostage.

Record cross-stratum patterns separately from verdicts. The same minor defect across five strata suggests a shared material or process worth a corrective action even where every stratum passed. Per-stratum verdicts decide the shipment; cross-stratum reading improves the next one.

Tally discipline for assorted loads

Tally stageRecorded howFailure it prevents
Pre-visit stratum listSKU, quantity, carton range per stratum in the bookingMystery cartons nobody planned to check
Carton opening countPer-carton tick sheet with stratum code as openedSample drawn from the wrong population
Per-stratum summaryChecked count and findings per stratum vs packing listBlended verdict hiding a failing variant
Quantity reconciliationLoaded cartons per stratum signed by both sidesDestination shortage disputes
Documented exclusionsWritten buyer exclusion for out-of-scope strataArguments about goods never meant to be checked

Multi-stratum bookings need coordination between buyer, factory, and inspector before the visit day. The team at UTS handles stratified sampling plans routinely, matching sample sizes and man-days to the actual variety in the consignment.

Illustration: Tally discipline for assorted Decorative illustration for the section "Tally discipline for assorted"; visual only, carries no data.

Run the mixed inspection without chaos

  1. Stage by stratum in advance.Stage cartons by stratum at the factory before the inspector arrives, with zone labels visible. An inspector who hunts cartons across the warehouse burns the sampling budget on walking.
  2. Check one stratum at a time.Open and verify stratum by stratum, finishing each tally before starting the next. Interleaved checking produces interleaved records that cannot be audited.
  3. Photograph per stratum.Transparent per-product records keep technical checks smooth for every line in the box, the same clarity trade rules expect when varied goods cross borders[2].
  4. Close the tallies on site.Fill the per-stratum summary before leaving the site, with both sides signing the tallies. Discrepancies resolved on the floor cost minutes; discrepancies discovered from the office cost a revisit.

The habit to keep: never let variety blur the verdict. Define strata at booking, sample within each, decide per stratum, and tally every carton to its group. Assorted loads handled this way ship faster than single-SKU loads handled carelessly.

Sources

  1. Wikipedia: Acceptance sampling —— Overview of acceptance sampling including AQL concepts used in product inspection.
  2. WTO: Agreement on Technical Barriers to Trade —— WTO portal for the TBT Agreement covering technical regulations, standards and conformity assessment.

Frequently asked questions

My supplier added six more colours last minute. What now?

Cap the strata at booking and allocate the remaining sample proportionally, or split the order into two inspections. Beyond about eight strata the per-stratum samples become too thin for meaningful verdicts, and the visit turns into a tour. Better to inspect the main strata properly this time and schedule the tail separately than to pretend thin coverage decided anything.

Does mixed cargo cost more inspection time?

Sample size follows the lot size and the agreed levels per stratum, so ten small strata need more total pieces than one large lot of the same total quantity. State the per-stratum sample and acceptance numbers in the booking so the visit length and cost are predictable. Skipping small strata to save a man-day is legitimate only when the buyer formally excludes them in writing.

Can I ever treat variants as one stratum?

Only when the variation cannot affect conformity: identical construction and materials differing solely in a non-regulated attribute, documented in the booking. Colour-only differences in unregulated goods are the usual case. Where safety, function, or regulated substances could vary, each variant is its own stratum regardless of how similar it looks.

How do I keep quantity control sane across many SKUs?

Two complete counts: the per-carton check tally recorded during opening, and the per-stratum summary compared against the packing list. Discrepancies are resolved before sealing by opening the relevant cartons under record, never by adjusting the spreadsheet. The signed tally sheets travel with the report and settle destination quantity disputes in minutes.