Where our nutrition data comes from

Search “banana”, “grilled chicken” or “white rice” and the numbers come from three databases in USDA FoodData Central:

• SR Legacy, USDA's laboratory-analyzed reference database (7,700+ foods).
• Foundation Foods, USDA's current lab-analyzed data, which USDA updates twice a year (300+ foods).
• FNDDS (Survey Foods), the database USDA uses for its national NHANES diet survey: 5,200+ foods as people actually eat them, grilled, scrambled or with the skin on.

We rewrote the names in plain English (“Raw Ground Beef 93% Lean” instead of “Beef, ground, 93% lean meat / 7% fat, raw”). The numbers are USDA's, unchanged.

Scan a barcode and the numbers come from USDA's Branded Food Products Database, part of FoodData Central (400,000+ products in our catalog). That's the Nutrition Facts label data manufacturers submit to USDA through industry data partners. It is label data, not USDA lab testing. Each barcode shows one product, not a list of look-alike duplicates.

When we don't have a barcode yet, the app photographs the product's Nutrition Facts label, reads it, and you check every number before saving. Numbers outside sane ranges, or calories that don't add up from the protein, carbs and fat, are flagged. When two different people add the same product on their own with matching numbers, it's marked verified. Anyone can report a wrong entry: a reported entry loses its verified mark, and three reports hide it.

Type in “my mom's lasagna, 600 calories” and it stays in your own log. It never shows up in anyone else's search. That's the main difference from crowdsourced databases, where one person's typo can become everyone's search result.

A small set of common foods (150+, like a protein bar, a spicy tuna roll, an oat milk latte or string cheese) have no generic USDA entry, so we researched them. Each one either copies a cited USDA food or is built from at least two cited sources (USDA ingredient values, standard recipes, the median of comparable USDA branded products, manufacturer labels), and is checked so its calories add up from its protein, carbs and fat. None of them come from another tracking app.

For 140+ restaurant chains (200,000+ US locations), we use each chain's official published nutrition, with the full menu and every customization: swap the bread, add bacon, light sauce, extra meat. A customized order still adds up from the chain's own numbers.

When a chain publishes no nutrition (10 chains today) or leaves gaps, we estimate those numbers and mark every one: “Estimated” with a ~ and a range in the app, and a ~ here on the website. We never present an estimate as official.

Each restaurant page shows when its menu data was last updated and the references we recorded for it.

A photo of your plate is an AI estimate, and it's shown as one, with a ~. When the photo shows a branded product, we swap in that product's published label numbers.

Calories, protein, carbs, fat, fiber, sugar, sodium and cholesterol. Not vitamins and minerals, and that's on purpose.

US restaurant chains have to publish calories, fat, saturated and trans fat, cholesterol, sodium, carbs, fiber, sugars and protein, and nothing more (a few add vitamins A and C, calcium and iron). Packaged-food labels add only vitamin D, calcium, iron and potassium. So when an app totals dozens of vitamins and minerals, any food that doesn't list one adds nothing to that total, and a day with a restaurant meal or most packaged foods comes up short.

We track the nutrients every restaurant and food label publishes, so a day's totals cover every meal in it.