When full inspection beats sampling: part 8

Key takeaways

1. Clustered failures, safety findings, far-above-limit major rates, and new-supplier instability are the signals that justify unit-level checks over further sampling.
2. Once a sample fails, the question changes from whether the lot is probably good to which units are good, and only defined acceptance work can answer it. ^[1]
3. Destination-market technical requirements attach per model under a multilateral framework that expects legitimate objectives pursued transparently. ^[2]
4. Compare the sorting fee against the priced rejected container and the return wave, including the hybrid option of sorting suspect ranges plus tightened sampling elsewhere.
5. Write the defect list, acceptance numbers, and handling path before the team arrives so the day runs on measurements.

This is the eighth part of a series on the boundary between sampling and full inspection, and it handles the decision under pressure: the sample has failed, the vessel date is approaching, and the buyer must choose between sorting, reworking, rejecting, or resampling. Earlier parts built the theory of clustered versus evenly spread defects; this part stays with the triage table and the cost arithmetic that follows. The goal is a calm sequence that protects the shipment without pretending the problem away.

Two wider frames keep the decision honest. Mandatory technical requirements in the destination market attach per model and pursue legitimate objectives under a multilateral framework that disciplines their design. ^[2] Sampling statistics, meanwhile, define what a sample can and cannot claim once a failure has appeared, because the question changes from whether the lot is probably good to which units are good. ^[1] Holding both frames in view prevents the two classic errors: shipping against the rules to meet a date, and resampling until a pass hides the signal.

The sections below set out the trigger signals, the triage sequence, the cost comparison, and the written rule that lets the inspection day run on measurements. Read it as a procedure to follow when the report comes back red rather than as an essay to consider at leisure.

Signals that sampling has done all it can

The first signal is geography inside the lot: defects map to specific carton ranges, production dates, or material batches while the rest of the lot looks clean. A random sample drawn across the whole lot dilutes such pockets and can easily return a passing verdict over a failing region, which is why clustering is the strongest argument for unit-level work. The defect map in the failed report is the document to trust here, not the overall defect rate.

The second signal is severity rather than quantity. Any safety failure, regulatory marking failure, or customer-facing functional failure removes the statistical discussion entirely, because no acceptance number above zero applies to properties that must simply hold. Sampling theory is explicit that its risk balances concern borderline quality, not known hazards, so continuing to sample after a safety finding is not thoroughness but evasion. ^[1]

The third signal is magnitude: a major-defect rate far above the acceptance number points at a shifted process rather than an unlucky draw, and redrawing is unlikely to reverse the story. The fourth is history, or its absence: a new supplier, a restarted line, or a changed material source gives sampling no stability to rest on. Any one strong signal justifies at least a targeted sort of the suspect region; two together make the whole-lot case difficult to refuse.

Triage sequence for the failed lot

1. Quarantine the suspect region firstMark and separate the suspect carton ranges or production codes from the rest of the stock so sorting, rework, and cleared goods cannot remix during the discussion.
2. Confirm scope from the evidenceUse the failed report defect map, counts, and photos to confirm which region and which defect types drive the failure before choosing the response.
3. Choose sort, rework, or reject explicitlySelect sorting of the suspect region or the whole lot, rework with defined acceptance, or rejection of the affected portion, and record the choice with its reasons.
4. Demand cause and prevention in writingAsk the factory for the cause behind the cluster and the process change that prevents recurrence, with a deadline tied to the next order rather than open-ended promises.
5. Verify before loadingVerify reworked or sorted goods under a recorded plan before authorizing loading, because unverified rework converts a controlled problem into a customer-discovered one.

Cost arithmetic without self-deception

Keeping technical duties in view under pressure

Date pressure tempts buyers to treat a vessel deadline as permission to ship marginal goods, but mandatory requirements in the destination market do not move with the sailing schedule. The multilateral framework for technical measures expects requirements to pursue legitimate objectives such as safety and environmental protection through transparent procedures. ^[2] A buyer who ships against those duties to meet a date converts a delay problem into a compliance problem, which is the more expensive of the two in every case the record shows.

The practical safeguard is to separate the loading decision from the compliance decision in the triage meeting. The loading decision asks which units meet the agreed acceptance criteria after sorting or rework; the compliance decision asks whether the per-model duties for safety, labeling, and restricted properties are evidenced for the shipped configuration. A full sort answers the first thoroughly and contributes to the second, but properties needing laboratory evidence still need their reports, and missing certificates still block compliant shipment regardless of how clean the sort looked.

Write both decisions into the order file with the evidence attached, so a later question from a customer or authority meets a documented answer rather than a reconstructed memory. The discipline feels slow during a crisis week and fast during every week that follows, because the next order inherits criteria, maps, and contacts instead of starting from a blank page.

The written rule that carries the day

A sorting day without a written rule dissolves into judgment calls, so the defect list with photos, the severity split, the acceptance numbers, and the handling path for failed units must reach the team before arrival. State who segregates failures, who labels cleared cartons, and who signs the final count, and the provider crew from an independent factory audit service background will run the day as a measurement exercise rather than a negotiation. The ten minutes spent writing the rule repay themselves by noon.

Include the stopping conditions explicitly: what happens if the suspect-region failure rate exceeds rework capacity, who authorizes escalation from sorting to rejection, and how cleared stock is marked against unchecked stock. These sentences prevent the mid-day phone calls that otherwise decide the shipment under time pressure with incomplete information.

Close the loop back to the system by recording what changed in the factory process and how the next order will be checked differently. Without that note the same clustering returns on schedule, and the buyer funds the same lesson twice. With it, the expensive week becomes the documented turning point after which the supplier history starts improving instead of repeating.

Frequently asked questions