Reliable Proposition 65 results begin with sampling, not chemistry: a written plan fixing sample size and selection, random draws by the inspector across the whole lot, sealed custody to the laboratory, component-level coverage of every distinct material, and interpretation that respects what the sample can claim about untested units. This guide walks through that test flow link by link so buyers commission results that actually support shipment decisions.
Key takeawaysThe sampling plan must precede selection, fixing sample size, selection method, and acceptance logic in writing. · Only random inspector-drawn samples represent the lot; factory-prepared units represent the factory's intentions. · Component-level coverage prevents minority materials from hiding listed substances. · Acceptance sampling theory defines what a sample result can claim about untested units in the lot. [1] · The finished-lot check plays the acceptance role familiar from equipment proving trials before handover. [2]
Proposition 65 testing fails most often before the laboratory receives anything: the wrong units were picked, too few of them, by the wrong person, without records. A perfect analysis of an unrepresentative sample produces a precise answer to the wrong question. Sampling discipline decides what the test flow can prove.
The complete flow runs from plan to report in six links: sampling plan, random selection, documented custody, component coverage, competent analysis, and lot-level interpretation. Every link constrains the next, and a weakness anywhere caps the strength of the conclusion. Buyers who manage the flow end to end get results they can act on; those who outsource it blindly get paper.
This guide details each link with the practical arrangements that make sampling representative, custody credible, and results interpretable against the shipped lot.
Writing the sampling plan first
A plan written after selection is rationalization, not planning. Before anyone opens a carton, fix the lot definition with quantities and carton count, the sample size with its rationale, the selection method across cartons, the components to be tested separately, and the acceptance logic for each possible outcome. Circulate the plan to buyer, inspector, and laboratory so all three work from the same definition of what the result will mean.
Sample size balances confidence against cost. Larger samples narrow the uncertainty about the lot but consume product and laboratory budget; smaller samples economize but leave wider doubt. Risk should drive the tradeoff: children's products, food-contact goods, and high-volume lots justify larger samples, while stable low-risk accessories tolerate leaner ones. Whatever the size, record why it was chosen.
Acceptance logic deserves the same advance agreement. Define which results release the lot, which trigger component-level follow-up, and which hold the shipment pending investigation. Agreeing these branches before results arrive prevents pressured reinterpretation after a borderline reading. The plan is a decision procedure, not just a headcount.
Random selection that represents the lot
Representativeness comes from method, not intent. The inspector selects cartons from across the full stacking layout, opens them, and draws units without letting the factory steer toward favorable stock. Every carton in the lot should have a known chance of contributing samples. Selection from the top layer of the nearest pallets is convenience sampling, and its results describe those pallets only.
Statistical sampling practice formalizes this intuition: defined procedures for drawing units let results speak for the lot within quantified risk bounds, while judgmental picks cannot support lot-level claims. [1] The inspector need not lecture the factory on theory, but the selection record should show carton numbers, positions, and quantities that demonstrate coverage. Coverage documented is coverage defensible.
Seal and label samples immediately at the draw point. Tamper-evident seals, sample tags with lot and date, and photographs of sealed units convert a claim of randomness into auditable fact. Factories that watch a rigorous draw also learn that future lots face the same discipline, which steadily improves the honesty of what gets presented.
Test flow from carton to conclusion
- Fix the planDefine the lot, sample size, selection method, component coverage, and acceptance branches in a written plan.
- Select randomlyDraw units randomly from cartons spread across the full lot layout, recording carton and position details.
- Secure custodySeal, tag, and photograph samples at the draw point, then log custody through handover to the laboratory.
- Cover componentsTest each distinct material separately so minority components cannot hide behind clean substrates.
- Check analysis qualityReview methods, limits, and scope fit before accepting numbers as decision evidence.
- Interpret for the lotRead results against the plan's acceptance branches and record the lot decision with its evidence.
Custody and component coverage
Chain of custody sounds bureaucratic until a result is challenged. The custody log should show who drew each sample, when, from which carton, sealed with what number, handed to whom, and received by the laboratory when. Gaps in this chain give skeptics room to argue the tested units differ from the lot, and that single doubt can neutralize an otherwise solid report.
Component coverage is where most test flows underperform. Finished-product testing of one composite sample routinely misses substances concentrated in coatings, prints, adhesives, or small metal parts. Disassemble conceptually before testing physically: list every distinct material the consumer could contact, and test each one. The extra analyses cost less than one held shipment.
Coordinate the laboratory handover explicitly. Confirm the lab received sealed units intact, verify the requested analytes match the screening flags, and agree reporting units and limits before analysis begins. A laboratory that understands the threshold question designs its reporting around it; one handed units without context reports numbers that the buyer must then translate alone.
Interpreting results for the lot decision
Interpretation starts by replaying the plan: given this sample, drawn this way, with these results, which acceptance branch applies. Results within limits across all components support release; detections below thresholds with sound exposure reasoning support release with documentation; exceedances or unsuitable scopes trigger hold and investigation. The plan's branches remove improvisation from a pressured moment.
Respect the limits of what sampling can claim. A passing sample bounds the lot's risk but never proves every unit clean, which is why acceptance practice pairs sampling with process controls and periodic verification rather than treating one pass as permanent clearance. [2] File the result with its plan and selection record so the next reviewer sees the full basis.
Feed every result back into the program. Clean rounds on a stable input justify maintaining cadence; detections, even below thresholds, flag inputs for closer watching; failures trigger the containment and root cause workflow. Commissioning Prop 65 compliance testing lot by lot without this feedback loop pays for data without buying learning. Prop 65 compliance testing
Sources
- Wikipedia: Acceptance sampling —— Overview of acceptance sampling including AQL concepts used in product inspection.
- Wikipedia: Factory acceptance test —— Overview of factory acceptance testing performed before shipment or installation.
Frequently asked questions
How many units should a Prop 65 sample include?
Size follows lot quantity, risk level, and the decision the result must support. Larger lots and higher-exposure products justify larger samples; screening questions tolerate smaller ones. Agree the plan in writing before selection, referencing the sampling standard or scheme used. Post-hoc justification of an arbitrary sample convinces nobody.
Who should select the test samples?
Random selection by the inspector from across all cartons, never factory handover of prepared units. Prepared samples reflect what the factory wishes to show; random draws reflect the lot. Sealed, labeled, and documented selection converts sampling from a ritual into evidence.
Should components be tested separately or as a finished unit?
Yes. Test each distinct material separately: substrates, coatings, inks, adhesives, elastic parts, and metal findings. Pooling dilutes a high reading from one component across clean ones and can mask a genuine exceedance. Component-level results also direct corrective action to the right input instead of condemning the whole design.
What follows a failed sample result?
Quarantine the lot, investigate whether the cause is bounded or systemic, and resample randomly from corrected production rather than reusing the same specimens. A failed sample has done its job by revealing the problem; reusing it for confirmation adds no information. The resampling plan should be at least as rigorous as the original.