Why protein label claims are often wrong, and how we test them

A number on a tub is a claim, not a measurement. Here is where those numbers drift, and the checks we run before any of them can be ranked.

Last updated: 18 May 2026.

The label is a claim, not a measurement

The protein figure printed on a tub is a declared value. It comes from a recipe, a supplier specification, or a periodic lab test on a reference batch, and then it is rounded and printed once. The powder you actually scoop is a different physical object from the one that produced that number. Most of the time the gap is small. Sometimes it is not, and nothing on the package tells you which case you are looking at.

Where the numbers drift

Several ordinary things move the real value away from the printed one. Recipes are specified on dry ingredients, but flavour systems, sweeteners, thickeners and anti-caking agents dilute the finished blend, so the protein share of a flavoured powder is usually lower than the unflavoured reference. Rounding rules differ by region, and a value can be converted between bases (per serving, per 100 g, as-sold versus as-prepared) more than once before it reaches a catalogue. Then there is the most common source of all: plain data-entry error in product feeds and retailer catalogues, where a serving figure is filed as a per-100 g figure, a decimal point moves, or a unit is dropped.

Nitrogen is not protein

The standard way to measure protein is to measure nitrogen and multiply by a fixed factor. That works because protein is the main source of nitrogen in real food. It stops working when nitrogen is added that is not whole protein. Free amino acids, peptides and other nitrogen-bearing compounds all raise the measured number without delivering the complete protein a buyer is paying for. This is a well-documented failure mode in the supplement category, and it is exactly the kind of inflated figure a buyer cannot detect from the package. It is one of the reasons we treat a single declared number as an input to be checked, never as a result to be trusted.

What we actually do

Every nutrition figure passes a validation engine before it can enter a ranking. A dry protein powder has hard physical limits: it cannot contain more protein than its own mass, and real concentrates, isolates and hydrolysates sit inside known, narrow bands. A value outside what is physically possible for that kind of product is rejected and quarantined rather than shown. A value we cannot verify is excluded, not guessed, because a quarantined product is safer than a confidently wrong number. Where the same product appears several times under slightly different names or sizes, entity resolution merges those listings into one canonical variant, so it is judged once on its true figures instead of repeatedly on conflicting ones.

What this means for you

You do not have to audit a label yourself. The point of this project is that the audit happens before anything is ranked, and that the result is reproducible rather than a matter of trust. The full method is on the how we rank page, and you can watch the validation and de-duplication work on synthetic data on the demo. If you want the companion piece on price, see what “€ per 100 g of protein” actually means.