20% Tolerance (FDA)

The FDA 20 per cent tolerance, codified in 21 CFR 101.9(g), is the regulatory tolerance band that defines when a Nutrition Facts label is compliant. It is the single most important number a dietary-assessment researcher should know about U.S. food labelling, because it defines the floor of label precision beneath which every downstream calculation sits.

Class I vs Class II

The regulation distinguishes two categories of nutrients:

• Class I nutrients are those added to a food by fortification or enrichment — the calcium added to orange juice, the B vitamins added to refined flour, the vitamin D added to milk. For Class I nutrients, the analytically measured value must be at least 100 per cent of the declared label value. There is no upper bound — a product may legally contain more than it claims.
• Class II nutrients are those naturally occurring in the food — the protein in yogurt, the fat in beef, the calorie content of nearly all foods whose calories come from intrinsic macronutrients. For Class II nutrients, the measured value must be at least 80 per cent of the declared value. Again, no upper bound.

The calorie case

Calories are, in nearly all foods, Class II — they derive from naturally-occurring macronutrients. This means the ±20 per cent tolerance applies. A product labelled at 200 kcal per serving is compliant if the measured value is 160 kcal or more. A product labelled at 100 kcal per serving is compliant if the measured value is 80 kcal or more. In both cases there is no upper limit: a 200-kcal label on a product that actually contains 280 kcal is within the regulation.

The direction of asymmetry is important. The tolerance protects consumers against under-delivery of nutrients they want (Class I, fortified) and against over-declaration of undesirable nutrients they want less of (the upper bound on sodium, saturated fat, added sugars, under 21 CFR 101.9(g)(5)). It does not bound over-delivery of calories.

Quantitative implications

The tolerance-plus-rounding structure means a label-derived calorie figure carries roughly the following precision floor: rounding increment on the order of ±5 kcal, Class II tolerance on the order of ±20 per cent downward and unbounded upward. Across a day of meals, the compounded effect is that a consumer who logs, say, 2,000 kcal by scanning labels could in principle have consumed anywhere from roughly 1,600 to 2,400+ kcal. This is not a defect in the tracker — the tracker is reading labels accurately — it is the inherent precision of the underlying system.

Why the tolerance exists

The 20 per cent figure was derived from studies of biological variation in food composition. Natural products vary: the fat content of a given cut of beef depends on breed, feed, age, and cut anatomy; the carbohydrate content of a potato varies with variety and growing conditions; the oil content of a tuna can varies by catch location and season. A narrower tolerance would require either much more expensive per-batch analytical labelling or routine label violations. The 20 per cent figure splits the difference, balancing regulatory practicality against consumer information quality.