Reading Nutrition Labels: What the Numbers Legally Mean

A nutrition label is a legal document, not a measurement. The calorie figure on the packet in your hand was not determined by testing that packet, and regulators do not require it to match. What they require is that the declared value be a defensible estimate of the average product, within tolerances that are considerably wider than most people assume.

Knowing where the slack sits changes how you read a label. It also explains why careful tracking can be systematically off by ten percent while every individual entry looks correct.

Where the calorie number comes from

Calorie values are almost never measured directly on finished food. They are calculated, using Atwater factors: 4 kcal per gram for protein, 4 for carbohydrate, 9 for fat, and 7 for alcohol. Wilbur Atwater derived these figures in the late nineteenth century by burning foods in a bomb calorimeter and subtracting estimated losses in urine and faeces.

They are averages across food categories, and they carry known biases. Whole almonds are the best-documented case: Novotny, Gebauer and Baer, in the American Journal of Clinical Nutrition (2012), measured metabolisable energy from almonds and found roughly a fifth fewer calories available than the Atwater calculation predicted, because the intact cell walls resist digestion. Similar overestimates have been reported for other whole nuts and for some high-fibre foods.

The direction runs the other way with heavily processed food. Cooking and mechanical processing increase energy availability by disrupting cell structure and gelatinising starch, and the label does not distinguish between raw and cooked forms of the same nominal ingredient. The general effect: labels tend to slightly overstate the calories you extract from whole, fibrous, minimally processed food and to understate them for the highly processed end of the range.

None of this makes labels useless. It means the numbers are accurate to a category, not to a mouthful.

The tolerances the law actually permits

This is the part that surprises people. Regulators allow a substantial margin between declared and actual values, and the margin is deliberately asymmetric.

In the United States, the FDA's compliance approach distinguishes nutrients you would want more of from nutrients you would want less of. For added nutrients such as vitamins, minerals, protein and fibre, a shortfall disadvantages the consumer, so the measured value must be at least eighty percent of the declared amount. For calories, fat, sugars and sodium, the measured value must not exceed one hundred and twenty percent of what the label claims.

So a bar declaring 200 kcal can legally test at 240. A product claiming 10 g of fibre can legally contain 8. Neither is a violation, and neither requires a correction.

The European Union sets its guidance tolerances in a similar spirit, published as guidance for competent authorities rather than binding limits, and expressed as bands that vary with the amount present. For energy the permitted deviation is broadly in the same territory. The rationale is the same in both jurisdictions: natural variation in agricultural inputs, manufacturing variation, and analytical uncertainty in the test methods themselves all have to be absorbed somewhere.

Compounding matters. Errors in a single food are unlikely to matter. Ten tracked items a day, each drawn from labels with an asymmetric tolerance skewed in the same direction, produce a total whose bias does not average out. This is one reason your tracked intake and your weight trend can disagree while both are behaving correctly, and it is a large part of why calibrating against your own weight data beats trusting the arithmetic. We work through that process in calorie tracking accuracy.

Serving size is the number that misleads most

The panel describes a serving, and the serving is defined by regulation rather than by how anyone eats.

US serving sizes are set from reference amounts customarily consumed, which were updated in 2016 to reflect actual eating patterns more closely, and packages between one and two servings must now declare the whole container. That closed the worst of the loophole. What remains is that a bag of crisps may still contain two and a half servings, and every number on the panel refers to one of them.

Two habits fix this. Read the per-container column when there is one, and check the gram weight of the declared serving against what you are actually eating. A serving defined as 30 g of granola is a modest amount, and the difference between that and a filled bowl is not a rounding error. It is often double.

Nothing about this is illegal or even unusual. It is simply a mismatch between a regulatory reference amount and a human portion, and the label is describing the former.

What each line permits

Several entries on the panel mean something narrower than they appear.

Calories. Calculated from Atwater factors, roundable to the nearest 5 below 50 kcal and to the nearest 10 above it, subject to the twenty percent upper tolerance.

Zero calories. Anything under 5 kcal per serving may be declared as zero. Cooking sprays are the standard illustration: a declared serving is a fraction of a second of spray, so a genuinely calorie-containing oil reads as nothing. Several seconds of spraying is real fat and real calories that no label will show you.

Total fat and trans fat. Trans fat under 0.5 g per serving may be declared as zero. Partially hydrogenated oils have largely been removed from the US food supply since 2018, which resolved most of this problem, but the rounding rule remains. The ingredient list is the reliable check.

Total carbohydrate. Calculated by difference in most cases, meaning it is what remains after water, protein, fat and ash are measured. It therefore inherits the error of everything else on the panel.

Fibre. Regulators tightened the definition, and in the US only fibres with demonstrated physiological benefit may be counted. This matters because manufacturers had been adding isolated fibres to raise the declared figure without the effect that fibre from whole foods produces.

Added sugars. A genuinely useful addition to the US panel, since total sugars cannot distinguish the lactose in milk from cane sugar poured in during manufacture. EU labels do not carry an equivalent line, which is a real gap.

Sodium. Declared in milligrams, and easy to underweight against a daily reference of 2,300 mg. A single restaurant-style ready meal can supply half of it.

Protein. Usually determined by measuring nitrogen and multiplying by 6.25, an approach that does not assess protein quality at all. Two foods declaring 20 g of protein can differ substantially in usable amino acids, which is the subject of protein requirements explained.

Percent daily value, and who it describes

The %DV column tells you how much of a reference intake one serving supplies. The reference is a 2,000 kcal diet, chosen as a round number for labelling rather than as a recommendation for any particular person.

That figure is above what many women need and below what many active men need. Using our TDEE calculator puts a personal number against it. A woman maintaining on around 1,800 kcal is reading percentages that understate her intake by roughly a tenth, while a man on 2,900 kcal is reading percentages that overstate his.

Where %DV genuinely helps is with micronutrients and sodium, since it converts an unfamiliar milligram figure into something interpretable. The convenient shorthand is that five percent or less is a little and twenty percent or more is a lot. For calories and macronutrients, use the gram figures and your own target instead.

Front-of-pack claims and the ingredient list

Claims on the front of a package are regulated but permissive, and several mean less than they sound.

Light or lite requires a reduction relative to a reference product, so a light version of something calorie-dense may still be calorie-dense. Reduced means at least twenty-five percent less of something than the comparator, which tells you nothing absolute. Natural has no meaningful federal definition in the US for most foods. Multigrain means more than one grain, not whole grain. The phrase to look for is whole grain as the first ingredient. Sugar-free permits under 0.5 g per serving and says nothing about sugar alcohols or refined starch.

The ingredient list is harder to dress up. Items appear in descending order by weight, which makes the first three or four the substance of the product. Sugars appear under many names (cane juice, dextrose, maltodextrin, rice syrup, fruit juice concentrate), and splitting them across several entries keeps each one lower in the order than the combined amount would sit. The added sugars line on the US panel now catches that, which is precisely why it was added.

Using labels well

Read the ingredient list before the panel. It tells you what the food is, and the panel only tells you what it contains.

Check the serving weight against your actual portion, every time, and weigh dense foods rather than estimating them. Oils, nut butters, cheese and anything fried generate the largest absolute errors when guessed.

Then hold the whole exercise at the right precision. Labels are accurate to roughly ten to twenty percent, portions add their own error, and the difference between two reasonable macro splits is smaller than either. Being consistently close over months beats being precisely wrong on any given day, which is why the macro calculator output is best treated as a target range rather than a specification.

One note on scope. If you have coeliac disease, a diagnosed food allergy, phenylketonuria or another condition where a specific ingredient carries real clinical risk, allergen declarations and precautionary labelling follow separate rules from the nutrition panel, and the management of that belongs with your clinician rather than with general guidance.