Methodology

How We Measure What's In Your Food

Not every number on FoodData Explorer was arrived at the same way. Some come from a food physically tested in a laboratory; others are calculated or estimated by USDA from related data. We think that distinction matters, so instead of hiding it behind a single average, we keep USDA's own record of how each value was determined. This page explains what that record means.

The two kinds of data on this site

Foundation Foods are directly lab-analyzed. Each of the 363 Foundation Foods on this site was physically tested using one of 275 named laboratory methods in our reference data — real analytical techniques like HPLC (used in 40 of those methods), microbiological assay (27), gravimetric analysis (16), and gas-liquid chromatography (15), each cited by its specific method name (for example, "AOAC 968.06 + 992.15" for a combustion-based nitrogen/protein method).

FNDDS ingredients are derived, not individually re-tested. The 1,882 ingredients in USDA's Food and Nutrient Database for Dietary Studies (122,330 nutrient values in total) represent "as consumed" foods — cooked rice rather than raw rice, for instance — built by USDA from a mix of direct analysis, calculation, and adjustment for how cooking changes a food's nutrient content. We keep USDA's own record of which applies to which value (see below), rather than presenting all values as equally direct measurements.

What "derivation" means, and why we show it

Every nutrient value in FoodData Central carries a derivation code — USDA's own record of how that specific number was produced. Our data carries 64 of these codes (plus 49 additional FNDDS-specific ones), rolling up into 10 broader source types. A few examples: "Analytical" means the value came from direct laboratory measurement; "Summed" means it was calculated by adding related measured components; "Based on another form of the food" means USDA adjusted a related food's measured value (with or without a retention-factor adjustment, described below) rather than testing this exact form directly. [VERIFY: confirm whether the derivation code/description is currently surfaced on individual food or ingredient pages — if not yet displayed, soften "we show" to "we retain in our data" until that UI ships.]

We don't treat this as fine print. A value derived by calculation from a well-measured related food is still useful — but it isn't the same claim as "USDA put this exact food in a lab," and we'd rather you be able to tell the difference than have us flatten it into one undifferentiated number.

Retention factors: why a cooking method can change the number

How a food is cooked changes its nutrient content — frying adds fat, boiling leaches out water-soluble vitamins, and so on. USDA accounts for this with retention factors: our reference data includes 270 of them, each tied to a specific food-and-cooking-method combination (USDA's retention factor table distinguishes, for example, "EGGS, FRIED, SCRAMBLED" from "EGGS, POACHED" from "EGGS, HARD COOKED" as separate entries, each with its own adjustment). When an FNDDS ingredient's derivation code indicates a retention factor was applied, the nutrient value reflects the food as prepared — not the raw ingredient before cooking.

What "per 100g" actually means

Every nutrient value on this site, regardless of source, is standardized to grams of nutrient per 100 grams of the food or ingredient as described (raw, cooked, or otherwise — the food's own description says which). That standardization is what makes it possible to directly compare, say, spinach and steak on the same nutrient — the numbers are on the same basis. Portion-level amounts (a "medium apple" versus 100g of apple) are a separate, secondary conversion layered on top of the 100g value, using USDA's own recorded portion weights where available.

Limitations — what this data can't tell you

FoodData Central reports nutrient composition for the specific samples USDA measured or modeled — not a guarantee about the exact item in your kitchen. Two real limitations worth naming plainly: natural variation (soil, season, and variety can shift a fresh food's actual nutrient content from the published average), and scope (a derived or calculated value is only as good as the related measured food it was built from). None of this is personalized dietary advice — see our note on that on the About page.

Something look off?

If a value here doesn't match what you'd expect, or doesn't match USDA's own FoodData Central listing for the same record, tell us: [email protected]. We trace every report back to the underlying USDA record and correct our side if we've made an error in how we imported or displayed it. For how current our overall dataset is and what's not included yet, see data sources.