How accurate are nutrition labels? What the FDA allows

A frozen dinner says 340 calories, not "about 340." The number looks exact, but it's better read as a regulated estimate. FDA rules leave room for error on purpose, and the way calories are calculated and rounded adds a little more.

The short version: packaged foods have tested close to their labels in published studies, often within single digits on average. Restaurant numbers are looser, and that's where the large misses appear.

The 20% rule

The compliance rules are in 21 CFR 101.9(g). When the FDA tests a food, it compares a composite sample against the label, and the check depends on the nutrient.

Three details are easy to miss.

• The limit is one-sided. For calories, the rule targets foods that contain more than the label says. It's built to protect you from understated numbers.
• It's an enforcement line, not a typical error. A 400-calorie meal would have to test above 480 calories to cross it. That doesn't mean most meals are 20% off.
• It applies to a composite. The test looks at a combined sample, so one package in your kitchen can sit further from the label than the composite does.

Where the calorie number comes from

Manufacturers don't have to burn food in a lab. Section 101.9(c)(1)(i) lists several accepted methods, including:

• Specific Atwater factors for particular foods
• General factors of 4, 4, and 9 calories per gram of protein, total carbohydrate, and total fat
• The same general approach after subtracting non-digestible carbohydrates and sugar alcohols from total carbohydrate
• Bomb calorimetry, with a correction for protein

This is why multiplying the grams on a label often won't give you the exact calorie figure. If a product has a lot of fiber or sugar alcohol, a straight 4-4-9 calculation will come out high, and the label may still be correct.

Rounding adds its own gap

For a normal meal, this barely matters. It matters more when the serving is very small. A serving with 4 calories and 0.4 g of fat can read 0 calories and 0 g of fat. Use ten servings and you've added 40 calories and 4 g of fat from a food that reads zero.

Serving size is the other variable

An accurate label doesn't help if your portion doesn't match the serving. FDA's 2016 serving size rule ties labeling to the Reference Amount Customarily Consumed (RACC), which reflects how much people typically eat at one time.

• A package holding less than 200% of the RACC must be labeled as a single serving.
• A package holding 200% to 300% of the RACC must show two columns, per serving and per package.

These requirements applied from January 1, 2020 for manufacturers with at least $10 million in annual food sales, and from January 1, 2021 for smaller ones. It's why a 20-ounce soda is labeled as one serving, while a 24-ounce bottle can carry two columns.

For foods without an obvious serving, such as cereal, nut butter, or oil, the fastest way to find your real portion is to weigh it a couple of times.

What lab tests have found

Few studies have checked these numbers directly. Packaged foods have stayed fairly close to their labels, while restaurant food has varied far more.

• Snack foods. Jumpertz and colleagues measured 24 common snack foods by bomb calorimetry. Calories came in about 4% above the label and carbohydrates about 8% above, with no significant difference for fat or protein. The authors concluded the gaps fell within FDA limits.
• Frozen meals. Urban and colleagues found that 10 reduced-calorie frozen meals from supermarkets averaged 8% more calories than stated.
• Restaurant food. The same study tested 29 reduced-calorie restaurant foods. They averaged 18% more than stated, and some items reached 200% of the stated value. The averages weren't statistically significant because the results varied so widely, which is itself a useful warning: a typical gap tells you little about the plate in front of you.
• Free sides. In that study, complimentary side dishes brought the food served to an average of 245% of the stated value for the entrée alone.
• Menu calories in England. Finlay and colleagues tested 295 items from 60 cafes, pubs, fast food outlets, restaurants, and leisure venues by bomb calorimetry. In their main analysis, 35% of items were more than 20% away from the menu figure. The menu was too high for 23% of items and too low for 11%. The average gap was small, about 17 calories, which shows again how little an average says about a single item. This is a UK study, so it's a guide to how menu numbers behave rather than a test of US menus.

Restaurant calories work differently

Chains with 20 or more locations that serve substantially the same menu must disclose calories under 21 CFR 101.11. A covered restaurant needs a "reasonable basis" for its numbers and must give the FDA the method and data behind them on request. The regulation itself doesn't define that phrase in detail.

A number built from a standard recipe also can't account for a heavier pour of oil or a bigger scoop of rice. Portion variation is a practical reason to treat restaurant calories as a starting point.

How to use labels in a food log

• Trust packaged food for everyday logging. Single-digit gaps aren't worth worrying about, and over time they're small compared with how much logs tend to miss.
• Log what you ate, not what the serving says. Use grams when you can.
• Leave some margin at restaurants. Log sides and sauces separately, because that's where a meal can quietly grow.
• Look at the week. Small errors matter less when you're reading an average.

Home-cooked and restaurant meals usually have no label at all, which is where AI photo estimates come in, with their own limits.

Where CaRoot fits

CaRoot's barcode scan looks up packaged foods in Open Food Facts, a community-maintained database, so you get saved product data rather than a visual guess. Community data can be outdated or wrong. If it doesn't match the package in your hand, you can correct it from the label, and the fix applies to everyone who scans that barcode. You can also use the nutrition panel on a package to create an entry.

Sources

• 21 CFR 101.9, Nutrition labeling of food. https://www.law.cornell.edu/cfr/text/21/101.9
• 21 CFR 101.11, Nutrition labeling of standard menu items in covered establishments. https://www.law.cornell.edu/cfr/text/21/101.11
• FDA. Food Labeling: Serving Sizes of Foods That Can Reasonably Be Consumed at One Eating Occasion; Dual-Column Labeling. Federal Register, May 27, 2016. https://www.govinfo.gov/content/pkg/FR-2016-05-27/html/2016-11865.htm
• FDA. FDA Extends Nutrition Facts Label Compliance Dates, 2018. https://www.fda.gov/food/hfp-constituent-updates/fda-extends-nutrition-facts-label-compliance-dates
• Jumpertz R, et al. Food label accuracy of common snack foods. Obesity, 2013. https://pubmed.ncbi.nlm.nih.gov/23505182/
• Urban LE, et al. The accuracy of stated energy contents of reduced-energy, commercially prepared foods. Journal of the American Dietetic Association, 2010. https://pubmed.ncbi.nlm.nih.gov/20102837/
• Finlay A, et al. Accuracy of menu calorie labelling in the England out-of-home food sector during 2024: assessment of a national food policy. British Journal of Nutrition, 2025. https://doi.org/10.1017/S0007114525105217

This article is general nutrition information, not legal or medical advice. If you need to manage sodium, sugar, or another nutrient for a health condition, talk to your doctor or a registered dietitian.