Method selection, verification and validation (clause 7.2)

Clause 7.2 requires appropriate methods and procedures, kept up to date and available to personnel. The laboratory must verify that it can properly perform methods before introducing them; where a method is non-standard, laboratory-developed or used outside its intended scope, it must be validated.

Validation evidence for a calibration method

• Uncertainty evaluation demonstrating the claimed capability.
• Comparison with a reference method or interlaboratory comparison.
• Repeatability and reproducibility studies.
• Systematic assessment of influence factors (immersion, stability, environment).

Changes to a validated method require assessment of their influence and, where they affect results, re-validation.

Verification and validation are different obligations

The words are used loosely in ordinary speech and precisely in clause 7.2, and mixing them is the commonest source of trouble at assessment. Verification asks whether this laboratory, with these people, this equipment and this environment, can achieve what a published method claims. Validation asks whether a method achieves what the laboratory claims for it, where no published method covers the case. The first is confirmation of capability; the second is establishment of fitness for purpose, and it is a much larger piece of work.

What the evidence has to demonstrate

Clause 7.2.2.1 asks validation to be as extensive as is necessary to meet the needs of the given application or field of application — which is an invitation to proportion, not to minimise. For temperature calibration the practical content is an uncertainty evaluation that actually supports the claimed capability, a repeatability study under real conditions rather than best-day conditions, a reproducibility study across operators and days, and a systematic look at the influence factors that dominate this field: immersion depth, medium stability and uniformity, self-heating, ambient temperature and the loading of the bath or block.

The characteristics established by the validation, the procedure used, and a statement of fitness for the intended use all have to be recorded. A validation folder containing data but no conclusion is a common finding, because nobody can tell from it what the laboratory decided the method could do.

Changes to a validated method

• Assess every change for its influence on the result — including the ones that look administrative, such as a different readout or a relocated bath.
• Re-validate where the change affects results, and record the assessment even where it does not. An unrecorded assessment is indistinguishable from no assessment.
• Watch the quiet substitutions: a replacement reference thermometer of the same model has a different certificate, a different drift history, and a different contribution to the budget.
• Keep the method version tied to the results produced under it, so an amended method does not retrospectively describe old certificates.