The unit does not define the result
Two values expressed in the same mass-per-mass unit are not necessarily measurements of the same thing.
One value may describe a dry powder as sold; another may describe the same product reconstituted according to label directions. One may represent the drained edible portion; another the entire container. One may be corrected to dry matter; another may retain native moisture. The numerals and units can look compatible while the measured quantities are not.
Metrology starts with the measurand—the quantity intended to be measured. For food contaminants, the measurand cannot be adequately specified by naming an element and a unit. Product matrix, state, portion, and analytical definition are part of the quantity.
Product state is a scientific variable
Food moves through states that change concentration without necessarily changing the absolute amount of an element in the source material. Drying removes water. Reconstitution adds it. Draining removes liquid that may contain a different analyte concentration than the retained solids. Cooking can add water, remove water, or redistribute material between food and cooking medium.
Regulatory instruments therefore attach limits to defined states. EU contaminant rules, for example, distinguish products as placed on the market from products ready to use after reconstitution and require processing, dilution, or concentration factors to be justified when a specific limit is absent for a processed form. The basis is part of the rule, not an after-the-fact conversion preference.
A conversion can aid interpretation. It cannot silently replace the native product state that gave a value its legal or scientific meaning.
Edible portion and preparation define the sample
Before an instrument measures anything, someone decides what counts as the sample. Rind may be removed or retained. A canned product may be drained or analyzed with its liquid. A powder may be prepared with a specified volume of water. Those decisions define the population to which the result speaks.
FDA’s Elemental Analysis Manual gives product-specific edible-portion instructions and separate procedures for homogenization. The emphasis is consequential: a representative analytical portion can be obtained only after the analytical sample is made sufficiently homogeneous for the portion size used. Otherwise, instrument precision may conceal sample-level non-representativeness.
Name the exact food, formulation class, and physical form represented.
State what was removed, drained, retained, combined, or prepared.
Record how the analytical sample became suitable for representative subsampling.
Retain the mass or volume actually taken into the method.
Normalization creates assumptions
Converting from dry to wet basis, from concentrate to prepared product, or from serving to mass can be mathematically straightforward. The evidentiary question is whether the necessary factor was measured, reported, imputed from a label, or borrowed from another source.
A measured moisture fraction and an assumed category-average moisture fraction do not carry the same uncertainty. Neither should overwrite the source value. A rigorous index preserves the native record, stores the transformation as a separate operation, names the factor, and marks whether that factor was source-reported or inferred.
- 1Preserve
Keep the source-reported value, unit, basis, and product state unchanged.
- 2Declare
State the target basis and why comparison requires it.
- 3Transform
Record the formula, factor, provenance, and direction of conversion.
- 4Qualify
Carry forward the uncertainty and any limitation created by the transformation.
The minimum comparability record
Evidence should not enter a pooled distribution merely because a parser can align the units. Before aggregation, the record should resolve at least the analyte, matrix, product form, edible portion, preparation state, reporting basis, sampling unit, analytical portion, method, and treatment of non-detects.
A standard needs the same discipline. The conformity cell must state the product identity and native basis to which the requirement applies. If a laboratory reports on a different basis, the conversion rule and responsibility for the supporting factor must be known before the result can be judged.
Conclusion
Comparability is not a visual property of a spreadsheet. It is a claim that two results describe sufficiently aligned measurands.
Preserving native basis may make a dataset less tidy. That is a sign of scientific honesty, not a defect. A useful evidence system makes differences visible, permits justified transformations, and refuses to manufacture comparability by deleting the context that gives each number meaning.