# How accurate are nutrition labels, really?

July 16, 2026 · Calorie Tracking · 8 min read · https://burnweek.fit/blog/how-accurate-are-nutrition-labels/

> The 20% a label may legally hide is the least of it. Rounding turns cooking spray into a zero, and the tolerance leans one way: toward the friendlier number.

**Key takeaways**

- A US food is misbranded only if measured calories exceed the label by more than 20% — a "200-calorie" bar can legally deliver 240 (21 CFR 101.9).
- The tolerance is asymmetric: naturally occurring nutrients must hit 80% of the declared value, while calories and fat may run 20% over — errors skew low, not random.
- Anything under 5 calories per serving is legally declared zero, which is how cooking spray and creamer enter your log as free.
- Packaged snacks check out: 24 products measured a median of just +6.8 kcal (+4.3%) over label (Jumpertz 2013).
- Label accuracy tracks manufacturing, not law: sealed snacks ran +4.3%, frozen trays +8%, restaurant plates +18% — the gap widens as the process moves from machine to hands (Urban 2010).

## A label is a declaration, not a measurement

How accurate are nutrition labels? Accurate enough to steer by, and legally permitted to be off by a fifth. Under US federal regulation, a packaged food carrying a calorie declaration "shall be deemed to be misbranded" only if the measured nutrient content "is greater than 20 percent in excess of the value for that nutrient declared on the label" ([21 CFR 101.9](https://www.law.cornell.edu/cfr/text/21/101.9)). A bar printed as 200 calories can be laboratory-measured at 240 and remain fully compliant. Nothing was faked. The number on the box was never a claim about the bar in your hand.

That is the crucial reframe. A label is not the output of someone measuring your specific package; it is a declaration about a product, computed from a formula, rounded by rule, and enforced against a tolerance band. Once you know how the number is manufactured, the right way to read it becomes obvious — as the middle of a range rather than a fact. This article walks the four stages where slack enters: the legal tolerance, the rounding rules, what labs actually find when they burn the food, and your own digestion.

## The tolerance, and why it's asymmetric

The 20 percent figure is not a scandal — it is an engineering necessity. Broccoli varies, flour varies, the machine that squirts filling into a cookie varies. Demanding an exact match would make compliance impossible. What's genuinely interesting is that the law's slack runs in *opposite directions* depending on the nutrient, and the asymmetry tells you what the regulation is actually for.

The rules split nutrients into two classes ([21 CFR 101.9](https://www.law.cornell.edu/cfr/text/21/101.9)):

| Nutrient type | Rule | Effect |
| --- | --- | --- |
| Class I — added nutrients in fortified foods | Content must be at least equal to the declared value | Can't overpromise fortification |
| Class II — naturally occurring nutrients (fiber, vitamins) | Content must be at least 80% of the declared value | May under-deliver by up to 20% |
| Calories, sugars, fat, saturated fat, cholesterol, sodium | Misbranded if more than 20% *in excess* of the declared value | May exceed the label by up to 20% |

Read the third row against the second. For things you want more of, the label is a floor you must roughly hit. For things you want less of — calories included — the label is a ceiling you must not badly overshoot. The regulation is protecting you from overstatement of benefit and understatement of cost, which is sensible. But the practical consequence for a calorie counter is one-directional: when a label is wrong, the law's structure means it is likelier to be wrong *low* than high.

> A "200-calorie" bar can legally deliver 240. The label isn't lying — it's declaring a range and printing the round number in the middle.

## Rounding: the quiet second error

Before the tolerance ever applies, the number has already been blurred by rule. Federal regulation requires caloric content to be "expressed to the nearest 5-calorie increment up to and including 50 calories, and 10-calorie increment above 50 calories, except that amounts less than 5 calories may be expressed as zero" ([21 CFR 101.9](https://www.law.cornell.edu/cfr/text/21/101.9)).

That last clause is the one worth internalizing. **Anything under 5 calories per serving is legally zero.** A cooking spray with a serving size defined as a fraction-of-a-second spray rounds to zero calories — and is pure fat. Six seconds of spraying is not zero. The same mechanism operates on every "zero-calorie" additive, creamer splash, and chewing gum you log as free. None of it is deceptive; the rounding rule is doing exactly what it says. It just means your log inherits a floor of small, systematically-downward errors that never round back up. This is one of the more reliable sources of [hidden calories stalling weight loss](/blog/hidden-calories-stalling-weight-loss), and the mechanics are worth knowing in full — see [do food labels round calories](/blog/do-food-labels-round-calories).

## What the calorimeter says: packaged food checks out, plated food doesn't

So much for what the law permits. What happens when researchers put the food in a bomb calorimeter and burn it? The results sort cleanly by food category, and the sort — not any one number — is the lesson.

| Study | What was tested | Measured vs. label |
| --- | --- | --- |
| [Jumpertz et al., 2013](https://pmc.ncbi.nlm.nih.gov/articles/PMC3605747/) | 24 popular packaged snacks, 8 categories | Median **+6.8 kcal (+4.3%)** — most within FDA tolerance |
| [Urban et al., 2010](https://pubmed.ncbi.nlm.nih.gov/20102837/) | 29 reduced-energy restaurant items | Averaged **+18%** over stated |
| [Urban et al., 2010](https://pubmed.ncbi.nlm.nih.gov/20102837/) | 10 supermarket frozen meals | Averaged **+8%** over stated |

The headline finding for everyday tracking is the top row, and it is reassuring: across 24 candy bars, chips, cereal bars, cookies, crackers, ice creams, nuts, and yogurts, measured metabolizable calories ran a median of just 6.8 kcal — 4.3 percent — above the label, and most products sat comfortably inside the 20 percent allowance ([Jumpertz et al., 2013](https://pmc.ncbi.nlm.nih.gov/articles/PMC3605747/)). Carbohydrate content and serving-size deviation together explained 95 percent of the excess. Standardized packaged food is, on the whole, labeled honestly.

The breakage sits outside the package. The same team's 29 reduced-energy restaurant items averaged 18 percent above stated ([Urban et al., 2010](https://pubmed.ncbi.nlm.nih.gov/20102837/)) — a gap that has nothing to do with labeling law and is worked through in [how accurate restaurant calorie counts are](/blog/how-accurate-are-restaurant-calorie-counts).

It would be a mistake to file those two results as a disagreement. A packaged snack at +4.3 percent and a restaurant plate at +18 percent are not rival answers to one question; they are two different questions, and the numbers land exactly where the manufacturing would predict. The row that proves it is the third one, and it is the most overlooked finding in the set: the 10 supermarket frozen meals — packaged goods, but assembled into a tray rather than sealed by weight — came in at 8 percent over, sitting neatly between the two ([Urban et al., 2010](https://pubmed.ncbi.nlm.nih.gov/20102837/)). Label accuracy is not really a property of labels. It is a property of how repeatably the food was made, and it degrades smoothly as the process moves from a machine toward a pair of hands.

## Even a perfect label describes the food, not the transaction

Suppose the tolerance were zero and the rounding exact. The number would still be an approximation, because of what it is computed from. Label calories come from the Atwater system — 4 calories per gram of protein and carbohydrate, 9 per gram of fat — devised at a USDA station in the 1890s. The FAO's expert consultation notes these general factors return values roughly 5 percent higher than food-specific factors across a mixed American diet, with gaps of 20 to 38 percent for individual foods like snap beans, cabbage, and lemons ([FAO, 2003](https://www.fao.org/4/y5022e/y5022e04.htm)).

And then there is you. In a tightly controlled inpatient study, participants ate identical measured diets while researchers burned both their food and their stool to calculate what was actually absorbed. Energy lost in stool ranged from 2.1 to 9.2 percent of intake on a 2,400-calorie diet across the group — a fourfold spread between individuals eating the same food. Shifts in gut bacterial composition tracked it: a 20 percent increase in the proportion of Firmicutes was associated with roughly 150 additional calories absorbed per day ([Jumpertz et al., 2011](https://pmc.ncbi.nlm.nih.gov/articles/PMC3127503/)). A label cannot know which end of that spread you sit on, because a label describes the food and not the transaction between the food and your gut. It is the same reason [calorie counts are ranges](/blog/why-calorie-counts-are-ranges) at every level of the stack.

## How to actually use a label

None of this argues for ignoring labels. It argues for grading your confidence by source. Packaged staples with a barcode are the most trustworthy number in your day — a median 4.3 percent overshoot is better than any other estimate you will make. "Reduced-energy," "light," and restaurant figures deserve genuine suspicion, since those are exactly the items with an incentive to sit at the friendly end of the tolerance. Serving size is where you should aim your attention: it drove much of the measured excess, and it is the one variable a scale settles instantly.

The useful posture is to treat the printed number as the center of a band a few percent wide, log it, and move on — while remembering that the band widens dramatically the further you get from a sealed package. A fuller audit of every layer of this stack lives in [how accurate calorie counting is](/blog/how-accurate-is-calorie-counting).

## FAQ

### Can a food label legally be wrong?

Yes, within limits. A food is misbranded only when measured calories, fat, sugars, or sodium exceed the declared value by more than 20 percent ([21 CFR 101.9](https://www.law.cornell.edu/cfr/text/21/101.9)). Below that threshold the label is compliant, so a 20 percent gap is a legal outcome rather than a violation.

### Are "zero calorie" foods really zero?

Not necessarily. Federal rounding rules permit amounts under 5 calories per serving to be declared as zero ([21 CFR 101.9](https://www.law.cornell.edu/cfr/text/21/101.9)). For a single serving that is negligible; for cooking spray, creamer, or condiments used many times a day, the rounded-away calories accumulate silently in a direction your log never corrects.

### Should I trust a packaged label more than a restaurant menu?

Much more, and the reason is not enforcement — both are legal declarations under the same tolerance. Packaged snacks measured a median 4.3 percent above their labels ([Jumpertz et al., 2013](https://pmc.ncbi.nlm.nih.gov/articles/PMC3605747/)), while reduced-energy restaurant items averaged 18 percent over ([Urban et al., 2010](https://pubmed.ncbi.nlm.nih.gov/20102837/)). The menu side is taken apart in [how accurate restaurant calorie counts are](/blog/how-accurate-are-restaurant-calorie-counts).

## Sources

- [21 CFR 101.9 — Nutrition labeling of food. Code of Federal Regulations (Cornell Legal Information Institute).](https://www.law.cornell.edu/cfr/text/21/101.9)
- [Jumpertz R, et al. Food label accuracy of common snack foods. Obesity (Silver Spring). 2013.](https://pmc.ncbi.nlm.nih.gov/articles/PMC3605747/)
- [Urban LE, et al. The accuracy of stated energy contents of reduced-energy, commercially prepared foods. J Am Diet Assoc. 2010.](https://pubmed.ncbi.nlm.nih.gov/20102837/)
- [Jumpertz R, et al. Energy-balance studies reveal associations between gut microbes, caloric load, and nutrient absorption in humans. Am J Clin Nutr. 2011.](https://pmc.ncbi.nlm.nih.gov/articles/PMC3127503/)
- [FAO. Food energy — methods of analysis and conversion factors, Chapter 3: Calculation of the energy content of foods. 2003.](https://www.fao.org/4/y5022e/y5022e04.htm)