01Identify the two sets of boundaries

JCGM 106 describes guarded acceptance in which an acceptance boundary lies inside a tolerance boundary. This reduces the risk of accepting a nonconforming item, with a corresponding trade-off in rejecting conforming items.JCGM 106:2012 — The role of measurement uncertainty in conformity assessment (opens in a new tab)

Consider an original hypothetical assay specification from 9.0 to 11.0 mg, inclusive. Suppose the agreed rule requires a guard band of 0.2 mg at each end. The acceptance interval then runs from 9.2 to 10.8 mg, inclusive.

The example does not recommend a peptide specification or a universal guard-band size. It demonstrates the consequences of one explicitly chosen rule.

02Apply the rule to specific measurements

A result of 10.7 mg lies inside both intervals and is accepted under the example’s rule. A result of 10.9 mg lies inside the specification interval but outside the acceptance interval, so it is not accepted under that rule.

A result of 11.1 mg lies above both upper boundaries. These cases should not be compressed into the same explanation: the central estimate occupies a different position in each.

If a system uses only pass and fail labels, its procedure still needs to explain what those labels mean. The word fail can otherwise conceal whether the measured value exceeds a specification or falls within a guarded decision region.

03Read how the band was selected

JCGM 106 discusses guard bands expressed as multiples of measurement uncertainty. Their selection depends on the decision framework and risk assumptions; the presence of an uncertainty statement does not by itself specify the multiplier.JCGM 106:2012 — The role of measurement uncertainty in conformity assessment (opens in a new tab)

If the example’s 0.2 mg band were defined as the expanded uncertainty, that definition would need the coverage factor and uncertainty basis. It should not be silently replaced with a standard uncertainty that has a smaller numerical value.

The uncertainty may also vary with the result or measurement conditions. A fixed-width illustration does not establish that one fixed width is appropriate throughout an actual assay’s range.

Reducing the chance of false acceptance can make acceptance harder for material close to a boundary. That trade-off is a reason to define and agree the rule in advance, not to choose it after seeing a preferred result.

04Look for enough information to reconstruct the decision

A useful conformity statement identifies the measured attribute, result, specification, applicable decision rule and how uncertainty enters it. If the report states only within specification, it may not reveal whether any guard band was used.

Boundary conventions matter too. This example includes 10.8 mg in the acceptance interval, but a different documented rule could use a strict inequality. Rounding a value onto the boundary should not silently decide its status.

Do not add another guard band when reviewing a result already assessed under one. First understand the laboratory’s rule and distinguish your separate evaluation from the statement actually issued.

When two laboratories give different conformity labels for similar values, compare their decision rules before assuming a factual disagreement about composition. Different accepted risks or uncertainty estimates can explain different decisions.