Test Reports and Evidence Standards: What Makes Proof Count

A test report is evidence when four things are true of it: the sample is identified with the specification it represents (versioned, dated, traceable to production); the method is the current recognised method for the property, named precisely; the laboratory is competent for that method (accreditation for the scope, or demonstrable competence where accreditation does not reach); and the report states results with the units, limits and uncertainty the requirement is judged against - not a bare 'pass'. Reports failing any leg are not worthless, but they are weaker evidence, and the file should know its own weak legs before an authority finds them.

This guide is the judging standard for the file's whole evidence population, including the reports suppliers send you.

Accreditation to ISO/IEC 17025 for the specific method is the gold standard because it audits competence continuously rather than episodically. Its limits deserve equal clarity: accreditation attaches to methods in scope, not to institutions generally - a laboratory accredited for metals analysis is not thereby accredited for PFAS panels, and the file should record the scope relied on, verified against the accreditation body's listing at the time of the test. Where no accredited method exists (novel properties, new criteria), the file compensates: method validation data, interlaboratory comparisons, the laboratory's reasoned competence statement. Evidence law in this system is a hierarchy with documented substitutions, not a cliff edge.

More conformity files die of sample identity than of bad results. The golden sample - handmade, pre-production, from a different line or a different continent's plant - proves a product that was never placed. The controls are mundane: samples drawn from production or from delivered goods by a stated sampling plan; the sample's batch and specification recorded on the report request (so the laboratory prints them on the report); retention samples kept for the file's life for the high-risk properties. Importers apply the mirror discipline from the substances station: their own sampling of delivered consignments, because the upstream report's sample was chosen by the upstream's optimism.

Supplier test reports and certificates enter the file under the same four legs, plus the scope checks the certification guide prescribed: the certificate's named material, process and site against what you actually buy; validity across the supply period; the chain's continuity. A quarterly evidence review - the file's reports and certificates listed with their expiry and currency status - converts evidence management from incident response into calendar. Expired evidence supporting live placings is among the commonest audit findings across all product law; it is also the most preventable, being visible in a list sorted by date.

Manufacturers' own laboratories generate much of the file's performance evidence (distribution testing, seal and barrier work). In-house results count when their competence is documented as a system: calibrated equipment with records, written methods tied to the recognised standards they implement, trained operators, and periodic cross-checks against external laboratories on reference samples. The file should hold the in-house laboratory's competence dossier once, centrally, referenced by every report it issues - the same architecture as the accredited laboratory's scope, self-administered with external spot-verification. What in-house testing cannot credibly do is grade its own homework on the contested properties; PFAS, metals and content figures with market consequences belong with independent laboratories on the risk rhythm the substances station set.

A word on measurement uncertainty, the number reports carry and files ignore. A metals sum reported at 92 mg/kg with an uncertainty of plus or minus 15 is not the same evidence as 92 with an uncertainty of 3: the first straddles the limit, and the honest file treats it as a borderline case - retest on a larger sample, tighten the input, record the reasoning. Surveillance laboratories work with the same physics and reach the same borderline; the file that has already reasoned about it owns the conversation. Teach report readers to find the uncertainty line before the result line; in borderline territory it is the result line.