Samples that fail twice usually fail for a reason the first report already named: the factory fixed the piece but not the process. Reworked samples can look convincing while the underlying fixture, setting, or material control stays exactly as it was, ready to reproduce the defect across the production lot. This second part of the defect-patterns series explains why findings repeat, how documented process control breaks the cycle, and what evidence to demand before approving a re-submitted sample.
Key takeaways
- Repeat defects across sample rounds signal unchanged processes, not bad luck, and demand system corrections rather than reworked pieces.
- Quality management practice holds that controlled, documented processes are what make good results repeatable from samples into production. [1]
- Laboratory competence underpins every material or safety claim, and accreditation cooperation is how that competence is recognised across borders. [2]
- Each sample round should close with written causes, process changes, and re-verification evidence, not verbal assurances.
- Buyers who track defect history across rounds build leverage that polite pressure can never match.
The first part of this series catalogued the five defect patterns that inspectors meet most often in sample evaluation. This second part asks the harder question: why do the same patterns repeat across sample rounds, even after the factory has promised corrections? The answer rarely lies in bad intentions. It lies in the gap between correcting a single defective piece and correcting the system that produced it.
A factory that reworks the sample without changing the process will send you a better-looking second sample that fails again in production. The defect was treated as an event; its cause was left running. Buyers who understand this distinction stop asking whether the new sample looks good and start asking what changed in the process since the last one. That single shift in questioning transforms sample evaluation from a beauty contest into a development tool.
The ideas behind this shift come from two established sources. Quality management thinking holds that consistent output requires controlled processes, documented responsibilities, and records that allow verification. And the international accreditation system exists because test results are only as trustworthy as the competence of the laboratory behind them. Both ideas have direct, practical consequences for how importers run sample rounds.
The difference between fixing a sample and fixing a process
Consider a housing whose mounting holes drift between sample rounds. The event-level fix is to hand-finish the holes on the sample pieces and ship them for approval. The sample passes, the buyer celebrates, and production begins with the same wandering fixture that caused the drift. Within weeks the assembly line abroad fills with housings that do not mate, and the investigation traces the failure to a cause the first sample round had already revealed. Nothing was learned because nothing was changed.
The process-level fix looks slower but costs less. The factory investigates why the fixture moves, clamps or rebuilds it, documents the setup, measures a run of consecutive pieces, and submits the data alongside the corrected sample. The second sample takes longer to arrive, but it arrives with proof that the cause is gone rather than hidden. Buyers should explicitly demand this form of correction: state in the evaluation feedback that re-submission must include the cause analysis and the process evidence, not merely improved pieces.
This discipline is the heart of quality management thinking applied to sourcing. The well-known quality management standards treat control of processes, clarity of responsibility, and retention of records as the mechanisms by which good outcomes become repeatable. [1] A supplier that cannot show which process step it changed, who owns that step, and what the new records say has not corrected anything, whatever the new sample looks like. Importers who internalise this test find their sample rounds getting shorter over time, because factories learn that cosmetic rework no longer earns approval.
What repeat offenders reveal, pattern by pattern
When dimensional findings survive into the second round, suspect the measurement system before blaming the operators. Factories sometimes inspect samples with worn gauges, uncalibrated instruments, or methods that differ from the buyer's, so both sides measure honestly and disagree sincerely. Ask for calibration status and the exact measurement procedure, and reconcile the methods before demanding tighter work. Half of all persistent dimensional disputes dissolve once both sides measure the same way.
When cosmetic findings repeat, the cause is usually inspection discipline rather than production skill. The line can produce clean pieces; the factory's own final check does not separate them from flawed ones. The effective demand is a defined cosmetic standard with viewing conditions and boundary samples, plus evidence that the factory's inspectors apply it. Without that, each round becomes a lottery in which the buyer hopes the submitted pieces happen to be the clean ones.
Material and functional repeats are the most serious because they survive into production silently. A substituted material looks identical and assembles normally; it fails months later in the field or under a laboratory test. This is where independent verification earns its place. Reports from UTS Inspection UTS Inspection on evaluated samples give buyers a neutral record of what was tested, against which criteria, and with what result, so that repeat findings accumulate into a documented history instead of dissolving into verbal assurances. That history is what makes the third-round conversation factual rather than emotional.
Scoring a factory across sample rounds
| Round criterion | What progress looks like | Warning sign of a stalled supplier |
|---|---|---|
| Finding count trend | Fewer and less severe findings each round | Same defects return with new ones added |
| Cause analysis quality | Named process step plus mechanism for each major finding | Generic replies about attention and effort |
| Process evidence | Setup records, measurement runs, revised instructions | Only the improved pieces, no supporting data |
| Response speed | Faster correction as rounds progress and knowledge builds | Each round slower as goodwill erodes |
| Record discipline | Complete file linking samples, data, and decisions | Scattered messages with no consolidated record |
| Stability of fixes | Corrected findings stay fixed in later rounds | Old findings reappear once attention shifts |
Score each row honestly after every sample round: a factory can be strong on correction speed yet weak on evidence, and the table keeps the two judgments separate.
Demanding evidence that deserves trust
- Require a written cause for every major findingFor each major finding, require the factory to name the process step, the mechanism, and the adjustment, with an owner and a date. Corrections without documented causes are decoration.
- Align measurement before re-verdictReconcile gauges, methods, and viewing conditions with the factory before judging the re-submitted sample, so agreement on the verdict rests on shared measurement rather than parallel monologues.
- Put critical claims through competent laboratoriesChemical, safety, and durability claims need results from competent laboratories, and the international accreditation system is the mechanism by which that competence is mutually recognised across economies. Confirm the test sits inside the laboratory's accredited scope. [2]
- Close every round in writingEach round ends with a consolidated record: findings, causes, changes, evidence, and the next re-submission requirement, confirmed by both sides and filed against the sample identity.
The buyers who get the most from sample evaluation share an unglamorous habit: they keep the file. Every report, every cause analysis, every measurement run, and every approval with its conditions goes into a folder keyed to the supplier and the product. Across seasons that folder becomes institutional memory that no staff change can erase. New team members inherit the factory's real history instead of its sales presentation, and negotiations start from documented behaviour. Begin the file with the current round, however imperfect, and let completeness compound from there.
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
The factory says the defect will not happen again. What answer should I accept?
Ask for the control change, not the apology. A credible answer names the process step that produced the defect, the adjustment made to that step, and the evidence that the adjustment worked, such as new measurement data or a revised work instruction. Vague answers like closer attention or stronger inspection signal that nothing in the process changed, and the defect will return with the next order.
The sample passed but the factory looks disorganised. Should I approve?
Score them separately. The sample verdict judges the object against the specification; the system verdict judges whether the factory can repeat the result. A passing sample from a factory with no records, no calibration, and no change control is a fragile pass. Approve the sample conditionally and require the system gaps to close before mass production, or the next lot will teach you what the sample did not.
When is accredited laboratory testing worth the cost at sample stage?
Accredited testing matters most for properties you cannot verify by eye: chemical composition, electrical safety, and durability claims. For those properties the factory's internal statement is weak evidence in any later dispute. Commission the critical tests at a competent laboratory during the sample stage, when quantities are small and the cost is lowest, and keep the reports tied to the sample batch.
Is it normal for first samples to fail?
Yes, and this is normal. New tooling, provisional fixtures, and early process settings produce a learning curve, and first samples often carry its marks. What matters is the direction of travel: findings should shrink between rounds and stay fixed once corrected. A factory whose second sample repeats the first sample's defects is not learning, and that is the signal to pause rather than to approve.