Every figure on GetFoodFacts originates in the USDA's FoodData Central database. We download the raw files USDA publishes — the Branded Foods dataset (manufacturer-submitted nutrition labels), SR Legacy and Foundation Foods (laboratory-analysed whole foods), and the Survey (FNDDS) foods — load them through a documented, version-controlled data pipeline, and render them into food, category, nutrient, and ranking pages using shared templates. No profile is hand-written, and no nutrient value is typed in by an editor. Each number you see is read directly from the official USDA record for that food's FDC ID.

Our editorial team is responsible for the parts a pipeline cannot decide on its own: which datasets to use, how each metric is defined and labeled, what the methodology says, which derived measures (such as a national percentile or grams-of-protein-per-100-calories) are computed and how, and what we will not publish. The pipeline then applies those decisions uniformly across every food, so the rule that governs one food's page governs all of them.

We publish only data that comes from official government sources, and we name the source on every page. Our datasets are:

• USDA FoodData Central — Branded Foods: manufacturer-submitted Nutrition Facts label data for packaged products sold in the United States, including serving sizes, ingredients, and allergens.
• USDA FoodData Central — SR Legacy & Foundation Foods: USDA laboratory analysis of whole and minimally processed foods, the reference set we use for nutrient distributions and rankings.
• USDA FoodData Central — Survey Foods (FNDDS): the foods and portion data used in national dietary surveys.
• FDA Nutrition Facts label format: the regulatory standard we follow when rendering each food's label panel and daily-value percentages.

We do not scrape third-party sites, we do not republish proprietary nutrition scores or diet-app ratings, and we do not generate any nutrient values ourselves. Where a figure is derived from the official data (for example, a percentile, a category average, or grams of protein per 100 calories), the page links to our methodology, which sets out exactly how it is calculated.

Because the data is read straight from USDA files, the most common source of error is the upstream record itself: Branded Foods values are self-reported by manufacturers and can be incomplete, and USDA stores macronutrient amounts per 100g (or 100ml), which we convert to the package serving size where the label declares one. Our pipeline applies systematic checks before a value is published. It screens out implausible outliers, shows a value as unavailable rather than guessing when the source omits it, reconciles the per-serving label against the per-100g USDA basis, and computes derived measures (percentiles, averages) against the full reference population they are drawn from.

When we find that a displayed number is wrong, we fix the cause, not the symptom. We trace the value back to the data layer, correct the derivation or screening rule there, and regenerate the affected pages, so the same class of error is resolved everywhere at once rather than patched on a single page.

GetFoodFacts does not accept payment, sponsorship, or promoted placement from food manufacturers, brands, supplement companies, or any commercial entity. Our only revenue source is contextual display advertising served by Google AdSense. Advertisers have no influence over which foods we cover or how their data is presented, and they receive no preferential placement. We publish no subjective "good food / bad food" verdicts; we present the official USDA data, the derived context that helps you read it, and the caveats that come with it, and leave the dietary judgment to you and your clinician.

USDA releases FoodData Central on its own cadence — the Branded Foods dataset is refreshed periodically as manufacturers submit updates, and SR Legacy and Foundation Foods are revised across USDA releases. We refresh our database to incorporate each new USDA release and re-stamp the affected pages so the published date reflects when the content genuinely changed. The data vintage in effect is named on every data page, on our About page, and in our methodology.

If a figure looks wrong, please tell us. We treat data-error reports as a priority and follow the same process every time:

1. Report. Email hello@getfoodfacts.com with the page URL and the figure you are questioning.
2. Verify. We check the value against the official USDA FoodData Central record for that food's FDC ID, which is printed on every food page.
3. Fix at the source. If the figure is wrong on our side, we correct the underlying data or derivation rule and regenerate every page it affects.
4. Note it. If the figure is correct but reflects a known upstream reporting quirk (for example, a manufacturer-submitted Branded Foods value that is incomplete), we explain the caveat rather than silently changing it.

Some apparent errors trace back to the USDA source itself. Because Branded Foods data is self-reported by manufacturers, a serving size, ingredient list, or nutrient value may be incomplete or out of date at the source. When that is the case, we will tell you so and point you to the official USDA record so you can verify it directly. Manufacturers who need to correct their own product data should do so through the USDA's FoodData Central program; we mirror the public file and cannot change a manufacturer's submitted information on their behalf.

Questions about our standards, methodology, or a specific figure are welcome at hello@getfoodfacts.com. For more on what the data covers and how it is processed, see our About page and methodology.