Where nutrition numbers come from, and why they disagree

Take a dry roasted, unsalted almond. Look up its carbohydrate in the official French table, then in the official American one. You will not find the same number.

The gap reaches 87% depending on which entries you compare. On something as simple as an almond, with no recipe, no brand, no processing. Two public bodies, two laboratories, two answers.

The temptation is to conclude that one of them is wrong. It is not, and the real explanation is more interesting: they are not counting the same thing, and both say so very clearly. You just have to read their documentation, which nobody does.

This article explains what is really happening when you read a nutrition figure, whatever its source.

First surprise: there is no international reference table. Each country compiles its own, with its own samples, its own laboratories and its own priorities.

One network, EuroFIR, brings them together at a single point of entry so that researchers do not have to query forty of them separately. It catalogs 32 European tables, and a dozen others: United States, Canada, Japan, Australia, New Zealand, Morocco, Tunisia.

You often read that European countries follow the same rules. That is true, but not at the level people assume.

What is shared is the label, not the table. EU regulation 1169/2011 imposes a single calculation method on manufacturers, across the twenty-seven member states and in the United Kingdom, which kept it after Brexit. A package sold in Berlin, Madrid or Paris therefore shows its figures under exactly the same conventions.

The national tables, however, do as they please. And some depart from it deliberately.

Britain's CoFID is the best example. Its documentation states in black and white that labeling requires nitrogen to be multiplied by 6.25 for every food, whereas more specific factors have been used in CoFID. In other words: the British scientific table does not calculate the way British packages do, and it owns it.

Three disagreements account for most of the gap between Europe and the United States. None is a mistake, all are documented.

Is fiber a carbohydrate?

In Europe, no. The carbohydrate figure leaves fiber out, and fiber is declared separately. In the United States, yes: the line is called Carbohydrate, by difference and fiber is inside it.

That difference alone creates a huge gap on a fibrous food. Cooked lentils go from 14.5 g of carbohydrate to 22.4 g, without a single molecule having changed.

And this is not a split between two camps of equal weight. Fiber kept separate is the international standard: the Codex Alimentarius guidelines require the declaration of "available carbohydrate, that is, dietary carbohydrate excluding dietary fiber". Most countries adopt it by default, for want of writing a rule of their own.

Brazil offers the best illustration of the trap. Its label shows carboidratos totais, "total" carbohydrate, and yet its regulation explicitly excludes fiber from that total. The word says one thing, the definition says the opposite.

This convention is where the American term net carbs comes from: the total minus the fiber. It exists only in the United States, for a simple reason. The label there shows the total, so anyone watching their carbohydrate has to do the subtraction themselves. The habit is so widespread that it appears as a selling point on packaging. A European reader never had to learn it: their label already gives them the net figure.

Nobody measures protein

This is the least known point, and the most disconcerting. No laboratory weighs the protein in a food. What is measured is total nitrogen, then multiplied by a factor to work back to protein.

The historical factor is 6.25, which assumes that all protein contains 16% nitrogen. It is an average. In 1931 Jones published food-specific factors to correct that approximation, nuts for instance holding proportionally more non-protein nitrogen.

The USDA applies these factors. The European regulation imposes 6.25 for everything. On an almond the gap therefore reaches 21%, from the choice of multiplier alone.

Three quarters of foods are indifferent to that choice, because their factor happens to be 6.25. It is grains, nuts and dairy that move.

A calorie, but which one?

Europe applies uniform coefficients: 4 kcal per gram of carbohydrate and protein, 9 for fat, 2 for fiber, 3 for organic acids. The USDA uses coefficients specific to each food, which account for what is not digested.

For the almond those coefficients are 3.47 for protein, 8.37 for fat and 4.07 for carbohydrate. They reproduce the American figure exactly.

Back to our dry roasted, unsalted almond, per 100 g.

On carbohydrate the two tables are within 1% of each other. 10 against 10.1. There is no disagreement between them on that point, once fiber is taken out on both sides.

So where did the 87% come from? From three causes stacked on top of each other, usually mistaken for one.

The first is the product. Comparing a branded almond with a generic one is not comparing two tables. A branded almond can show half the carbohydrate of the generic, and it is not wrong for that: it is a different batch, a different grade, a different analysis.

The second is the convention. American carbohydrate includes fiber. On an almond that adds nearly 11 g to a figure worth 10. A doubling, for nothing.

The third is the denominator. The carbohydrate in an almond is a small number. A 4.7 g gap weighs 87% on a base of 5.4, but the same gap would be only 10% on white bread. And since American carbohydrate is worked out by difference, from water, protein, fat and ash, any imprecision in those lands in the smallest number on the row.

A nutrition figure means nothing without its calculation rule. Comparing two numbers without knowing where they come from is comparing two distances without knowing whether they are in kilometres or miles.

A gap is not a mistake. Before concluding that a source is wrong, check three things: is it the same product, the same convention, the same denominator.

Carbohydrate and fat are robust, once fiber is put back on the same side. Protein and calories are far less so, because they depend on conversion choices rather than direct measurement.

And for you, in practice: look at which side the country where you shop falls on. In the United States or Canada, your carbohydrate includes fiber. Almost everywhere else, it does not. That is why this site lets you choose your reference in preferences rather than imposing one.

On crowdsourced databases, fed by shoppers themselves rather than public laboratories, see our article on Open Food Facts. And on what the word "saturated" really covers, Saturated fat: why the number on the label says almost nothing.