Missing, unreadable and zero are three different nutrition-data states
Set up explicit missing-value states so incomplete scans cannot silently produce misleading comparisons.
A blank nutrient field should not become zero merely because a calculator needs a number. Distinguish a printed zero, a field that is not listed, and a field you cannot read. Each state leads to a different next action.
Create a status beside the amount
Use an amount field that can remain empty and a separate status field. Useful statuses include transcribed, not listed in captured panel, unreadable, and not checked. Record the printed text as well: a label that says less than a quantity should keep that qualification instead of being converted into an exact value.
Label formats do not all expose the same categories. The European Commission nutrition guide distinguishes required nutrition information from additional voluntary information. That is a reason to inspect the actual panel rather than expecting every field in your template to be populated.
Follow three hypothetical records
Suppose product A visibly lists 0 g for a nutrient, product B has no row for that nutrient in the captured panel, and product C has a row obscured by glare. A can be recorded as a printed zero, without making a laboratory claim about absolute absence. B stays not listed in captured panel. C stays unreadable.
None of these records should share a bare numeric zero in a spreadsheet. Otherwise a later sort will rank all three as equal, even though only one supplies a readable declared number.
Define how calculations handle gaps
For a two-product comparison, calculate a difference only when both required inputs are available and use compatible units and bases. If one is missing, display comparison unavailable and identify the missing input. Do not quietly drop the missing product or substitute a category average.
If you summarize several products, state the coverage. An illustrative list with six readable values out of eight products supports a summary of those six observations, not a claim about all eight. Keep the two incomplete entries visible so a reader sees the gap.
To resolve an unreadable field, recapture the image or check a matching manufacturer source. To investigate an unlisted field, request product-specific information rather than reverse-engineering an exact quantity from the ingredient list. Once new evidence arrives, update the status and preserve the old reason for the gap. This makes completeness measurable without disguising uncertainty as a favorable result.