On this page
- What error do the US label rules allow?
- Are calorie labels accurate in the UK and Canada?
- What do lab tests find in packaged and restaurant foods?
- Why can calories be off even when a label follows the rules?
- Why do calories not match macros?
- Should you subtract the thermic effect of food?
- Do labels show raw or cooked food?
- How accurate are calorie-tracking apps?
- What if a calorie-counting app has the wrong serving size?
- How much error should you plan for?
- Who should skip calorie counting?
- What should you do next?

What error do the US label rules allow?
The US Food and Drug Administration (FDA) uses different compliance margins for calories and protein. Under 21 CFR 101.9(g), checked November 11, 2026, FDA ordinarily tests a composite of 12 subsamples from 12 randomly chosen shipping cases. That is a batch-level test, rather than a guarantee about your package.
Calories fail the numerical criterion when the composite is more than 20% above the declared value. Reasonable calorie shortfalls are accepted. For naturally occurring protein, the minimum is 80% of the declaration. When protein is added as a nutrient supplement, the food’s total protein is Class I and must be formulated to reach the declared amount. The regulation also allows for recognized lab-method variability for calories and naturally occurring protein.
| Nutrient | Label per declared serving, food as packaged unless stated otherwise | Numerical criterion |
|---|---|---|
| Calories | 200 calories | No more than 240 calories |
| Naturally occurring protein (Class II) | 20 g | At least 16 g |
| Protein in a food supplemented with protein (Class I) | 20 g | At least 20 g |
Rounding adds another limit. Section 101.9(c) requires calories in steps of 5 calories through 50 calories, then 10 calories above that. A serving under 5 calories may be declared zero. You can check the serving size before multiplying; reading a nutrition label for protein explains that step.
Reference measurements also have uncertainty. The US National Institute of Standards and Technology (NIST), in a page updated May 27, 2026, says its measurements of fats, proteins and carbohydrates are accurate to within 2% to 5%. That describes NIST’s measurements, rather than a guarantee for every food label.
Are calorie labels accurate in the UK and Canada?
The UK Department of Health’s 2017 guidance describes calorie labels as average values. Its technical guidance on nutrition labelling (checked November 11, 2026) describes manufacturer analysis, ingredient calculations or established data as bases for average values and points to separate tolerance guidance. Natural and production variation are recognized; the US calorie criterion is not a universal UK rule.
In Canada, the CFIA nutrition labelling compliance test, checked November 11, 2026, sets a principal calorie criterion of no more than 120% of the declared value, adjusted for rounding, for the mean of three composite samples of at least four randomly selected consumer units each. Additional variability criteria and product exceptions apply. Neither country’s guidance guarantees an individual package’s exact energy.
What do lab tests find in packaged and restaurant foods?
Lab tests show variation between individual foods, even when the sample average is close to the stated value. These US studies from 2010–2013 compared laboratory energy estimates with labels or menus:
| Study and sample | Result |
|---|---|
| Jumpertz 2013: 24 packaged snack foods | Median calories were 4.3% above the label. |
| Urban 2010: 29 reduced-energy restaurant foods and 10 frozen meals | Restaurant foods averaged 18% above stated calories; frozen meals averaged 8% above. Neither difference was statistically significant. |
| Urban 2011: 269 items from 42 restaurants | Measured energy averaged 10 kcal per portion above stated energy, not statistically significant; 19% of items had at least 100 kcal more than stated. |
Restaurant menus follow a separate rule. Under 21 CFR 101.11(c), read in November 2026, covered establishments must have a reasonable basis for their numbers and take reasonable steps to match preparation and portions to that basis. Databases, cookbooks and lab analyses can qualify; the section sets no percentage tolerance. For portion estimates, see protein when eating out.
Why can calories be off even when a label follows the rules?
Calorie calculations depend on how much food energy becomes available to the body. The Food and Agriculture Organization (FAO), in its 2003 food energy report, describes general Atwater factors: about 4 kcal per gram for protein and carbohydrate, 9 for fat and 7 for alcohol. Food calories are kilocalories, abbreviated kcal; this page uses both names for the same unit. For drink examples, see calories in alcohol.
FDA’s 21 CFR 101.9(c), read in November 2026, permits general factors, food-specific factors and other listed methods. FAO explains that factors differ by food and digestibility. Nut feeding trials illustrate the difference between predicted and metabolizable energy, the energy remaining after losses in feces and urine:
| Nut | Atwater prediction | Measured energy |
|---|---|---|
| Natural, whole almonds, Novotny 2012 | 168–170 kcal per 1 oz (28 g) | 129 kcal per 1 oz (28 g) |
| Walnut halves and pieces, Baer 2016 | 185 kcal per 1 oz (28 g) | 146 kcal per 1 oz (28 g) |
| Pistachios, Baer 2012 | About 159 kcal per 1 oz (28 g) | About 151 kcal per 1 oz (28 g) |
These controlled trials report group averages for the nut preparations tested in healthy adults. The results do not establish a personal calorie adjustment or extend to nut butters. You can keep the package or database value as a consistent logging reference.
Why do calories not match macros?
Calories calculated from displayed macros can differ from label calories without proving deception. Macros means protein, carbohydrate and fat. Under 21 CFR 101.9(c)(1), checked November 11, 2026, factors apply to nutrient amounts before rounding. The rules permit food-specific factors and separate treatment of non-digestible carbohydrate and sugar alcohols, also called polyols. Multiplying every displayed carbohydrate gram by the usual factor can therefore give a different result.
Other energy contributors include alcohol and organic acids, as the FAO’s 2003 food-energy report explains. Fiber and polyols need appropriate factors rather than being added again on top of carbohydrate. A small mismatch can reflect methods or rounding; after checking the portion and entry, you can ask the manufacturer to explain a large unresolved discrepancy.
Should you subtract the thermic effect of food?
The thermic effect of food is energy spent processing a meal. FAO’s 2003 report distinguishes metabolizable food energy from heat produced during digestion, absorption and metabolism. Keep these concepts separate:
- Label energy: an estimate of energy supplied by the food.
- Thermic effect: energy spent processing food, which varies with the food eaten.
- Personal energy expenditure: energy your body uses, including metabolism, activity and processing food.
For logging, you can retain the label value rather than subtract a fixed percentage from protein calories. The FAO energy framework does not provide a personal subtraction rule; label energy and your total energy needs answer different questions.
Do labels show raw or cooked food?
FDA’s 21 CFR 101.9(b)(9), read in November 2026, generally bases nutrition on food as packaged or purchased; exceptions include certain drained foods. An additional column may show the food as prepared.
USDA FoodData Central (SR Legacy, checked November 2026) lists 371 kcal per 3.5 oz (100 g) for dry enriched pasta and 158 kcal per 3.5 oz (100 g) for cooked enriched pasta without added salt. These are separate database entries, not the same portion before and after cooking. You can weigh and log food in the same state; how to count calories covers the method.
How accurate are calorie-tracking apps?
App accuracy depends on the entry and the task. Li and colleagues’ 2024 study in Nutrients entered dietitian-written sample menus into 16 manual food-logging apps and compared them with a professional nutrient database. Across apps, mean daily energy was about 249 kcal higher for the Western menu and about 363 kcal lower for the Asian menu. These were menu tests against a database, not measurements of what people absorbed.
Evenepoel and colleagues’ 2020 validation study compared app calculations with a Belgian food composition database using records from 50 participants. Participants’ four-day average energy intakes correlated strongly. Researchers supplied reference values for generic foods and detailed logging instructions; no energy values were removed during data cleaning. Correlation means values move together; it does not establish exact agreement. In practice, compare an unfamiliar entry with your package and check its portion and preparation state. The guide to AI calorie counter accuracy explains what to check before saving a photo estimate.
What if a calorie-counting app has the wrong serving size?
A calorie discrepancy is a reason to check the entry before changing your intake. You can work through this sequence:
- Match the portion and column. Compare the same food, serving mass and preparation state: raw or cooked, packaged or drained. Check per-serving versus per-package values. FDA’s label-reading guide, checked November 11, 2026, explains how amounts scale with servings eaten.
- Check the input and database record. Compare the package with the selected record, then check your serving multiplier and units. You can choose a matching record or correct an editable entry. App-provider database support guidance (2026) distinguishes wrong nutrition information, duplicate records and unavailable serving sizes.
- Check the saved diary. Repeated search results are alternative database records; they do not by themselves mean you logged the meal twice. Inspect the diary for the same eaten portion saved twice, and remove an accidental duplicate if present.
- Consider the calculation method. Once portions and records match, the factors, fiber, polyols and rounding described above may explain the remaining difference. Avoid forcing every label to match a simple macro calculation.
You can stop once you have a usable, repeatable entry. Checking a discrepancy does not require making every daily total exact.
How much error should you plan for?
No single error percentage fits every food, menu or app. The studies above do not support adding a blanket correction to your log. Instead, you can choose a repeatable method:
- Check portions you repeat. A kitchen scale can reduce portion-estimation error.
- Reuse checked entries. Match the food, portion and raw or cooked state before saving a favorite.
- Include additions. The guide to hidden calories in oils, sauces and drinks covers commonly missed ingredients.
- Look across days. A weekly average can be easier to interpret than an isolated daily total, while still remaining an estimate.
Who should skip calorie counting?
Calorie counting is optional. If you have a history of an eating disorder or disordered eating, skip counting and talk to a clinician. Is counting calories healthy? discusses suitability, and tracking protein without counting calories offers another approach. Ask your prescriber before cutting calories if you take GLP-1 or diabetes medicines. If you are pregnant, breastfeeding, under 18, have kidney disease or follow a medical diet, talk to your care team first.
What should you do next?
Choose a food you log often, check its portion against the package, and save the matching entry. For the full routine, see how to count calories.
YolkPal is an iPhone protein tracker with optional calories. AI estimates are editable before saving, and saved foods let you reuse meals you repeat.
Sources
- Code of Federal Regulations, Title 21, section 101.9, “Nutrition labeling of food”: paragraphs (b)(9), (c)(1), (e) and (g)(2)–(6). Read on the eCFR, November 2026. ecfr.gov, 21 CFR 101.9
- Code of Federal Regulations, Title 21, section 101.11, “Nutrition labeling of standard menu items in covered establishments”: paragraph (c). Read on the eCFR, November 2026. ecfr.gov, 21 CFR 101.11
- National Institute of Standards and Technology. How do you know your food’s nutrition facts label is accurate? How Do You Measure It? Page last updated May 27, 2026. nist.gov
- Food and Agriculture Organization of the United Nations. Food energy: methods of analysis and conversion factors. FAO Food and Nutrition Paper 77, Rome, 2003, sections 3.2–3.5 on energy flow, thermogenesis and the Atwater systems. fao.org/4/y5022e/y5022e04.htm
- Jumpertz R, et al. Food label accuracy of common snack foods. Obesity 2013;21(1):164–169. doi:10.1002/oby.20185
- Urban LE, et al. The accuracy of stated energy contents of reduced-energy, commercially prepared foods. J Am Diet Assoc 2010;110(1):116–123. doi:10.1016/j.jada.2009.10.003
- Urban LE, et al. Accuracy of stated energy contents of restaurant foods. JAMA 2011;306(3):287–293. doi:10.1001/jama.2011.993
- Novotny JA, et al. Discrepancy between the Atwater factor predicted and empirically measured energy values of almonds in human diets. Am J Clin Nutr 2012;96(2):296–301. doi:10.3945/ajcn.112.035782
- Baer DJ, et al. Walnuts consumed by healthy adults provide less available energy than predicted by the Atwater factors. J Nutr 2016;146(1):9–13. doi:10.3945/jn.115.217372
- Baer DJ, et al. Measured energy value of pistachios in the human diet. Br J Nutr 2012;107(1):120–125. doi:10.1017/s0007114511002649
- Li X, et al. Evaluating the quality and comparative validity of manual food logging and artificial intelligence-enabled food image recognition in apps for nutrition care. Nutrients 2024;16(15):2573. doi:10.3390/nu16152573
- Evenepoel C, et al. Accuracy of nutrient calculations using the consumer-focused online app MyFitnessPal: validation study. J Med Internet Res 2020;22(10):e18237. doi:10.2196/18237
- USDA FoodData Central, SR Legacy: pasta, dry, enriched (FDC 169736); pasta, cooked, enriched, without added salt (FDC 169737). fdc.nal.usda.gov
- UK Department of Health. Technical guidance on nutrition labelling. March 2017, pages 19–20. Checked November 11, 2026. Technical guidance (PDF)
- Canadian Food Inspection Agency. Nutrition labelling compliance test. Checked November 11, 2026. inspection.canada.ca
- US Food and Drug Administration. How to Understand and Use the Nutrition Facts Label. Checked November 11, 2026. fda.gov label-reading guide
- MyFitnessPal Help. How to report incorrect food items in the database. Updated October 1, 2026. Used for database-error categories, not nutrition evidence. Provider support guidance
This article is general information, not medical advice. If you have a health condition, are pregnant, or follow a diet set by your doctor, talk to them or a registered dietitian before changing how you eat.


