Foundation Foods vs. Branded Foods vs. Survey Data: What's the Difference?

This site currently holds two of USDA's five FoodData Central data types: 469 Foundation Foods (lab-sampled, single-ingredient items) and 1,882 Survey/FNDDS ingredients, the foods behind national dietary surveys. It does not include Branded Foods — USDA's packaged-product database — at all yet. Here's what that split actually changes about a number you look up.

What are the different kinds of data USDA actually calls "FoodData Central"?

USDA doesn't run FoodData Central as one dataset — it's five, each built a different way. According to USDA FoodData Central's own data type documentation: Foundation Foods covers "data and metadata on individual samples of commodity/commodity-derived minimally processed foods," with full sample-level detail. SR Legacy is "derived from analyses, calculations, and published literature" and received its "final release April 2018" — it's frozen, not updated. FNDDS (Survey) provides "data on nutrients and portion weights for foods and beverages reported in What We Eat in America, NHANES," refreshed "every two years in concert with WWEIA, NHANES release." Branded Foods comes from "labels of national and international branded foods collected by a public-private partnership," updated monthly. A fifth type, Experimental Foods, covers peer-reviewed research data added a few times a year. This site currently draws on two of those five.

What's actually in this database right now?

Querying this project's own tables directly: the food table — Foundation Foods — holds 469 rows. The fndds_ingredient_nutrient_value table holds 122,330 individual nutrient measurements across 1,882 distinct ingredient codes, which is this site's Survey/FNDDS coverage. There is no Branded Foods table populated here, no standalone SR Legacy dataset, and no Experimental Foods data. That's a deliberate, documented scope limit, not an oversight: per this project's own Content Team Playbook, a fresh USDA bulk download that would add Branded Foods (and full SR Legacy/Survey coverage) is listed as a future evaluation, not a completed import.

Why does the same food show up twice, under two different ids, with slightly different numbers?

This is where the three-way distinction stops being abstract. Tomatoes, grape, raw exists in this site's Foundation Foods as fdc_id 321360. The same food, under the same name, also exists in the Survey/FNDDS ingredient table as ingredient code 100147 — querying that record shows its protein (0.83g), potassium (260mg) and vitamin C (27.2mg) values are tagged with source "Foundation," meaning FNDDS pulled them straight from the Foundation record rather than running a separate analysis. But the FNDDS energy value for that same tomato is 31 kcal, while the Foundation record itself states 27 kcal — a real discrepancy between two live numbers for the same food, in the same database, both currently sourced as "Foundation" or close to it. The gap almost certainly comes from FNDDS applying its own energy-calculation convention on top of the borrowed macronutrient values rather than copying Foundation's stated kcal figure directly — a synthesis point neither record states on its own. Hummus, commercial and its FNDDS counterpart, Hummus, commercial, form a second matched pair worth comparing the same way.

Does "Survey" data mean it's less accurate than Foundation data?

Not inherently — it means something different was being measured, and this project's own data shows exactly how that plays out. Grouping all 122,330 rows in the fndds_ingredient_nutrient_value table by their nutrient_value_source field gives a clear picture: 116,333 of them — about 95% — are sourced to "SR Legacy," versus just 3,407 (under 3%) sourced to "Foundation." The tomato example above, where the value was pulled from Foundation, is actually the exception rather than the rule. The overwhelming majority of this site's 1,882 Survey/FNDDS ingredients carry nutrient figures inherited from SR Legacy — the database USDA says received its "final release April 2018" and will not be updated again. In practical terms: a number on an ingredient page labeled "FNDDS survey nutrient database" is, about 19 times out of 20 by row count, actually an SR Legacy value wearing an FNDDS label, not a fresh Foundation analysis.

That doesn't make the number wrong — SR Legacy's values were real lab analyses and literature figures in their own right when they were compiled — but it does mean an FNDDS value's effective age and chain-of-custody depends entirely on which underlying dataset backs that specific figure, and that underlying source is usually the older one, not the newer one.

Does that 95% figure mean Foundation Foods barely matters to this site's Survey data?

Not quite — it means Foundation's influence shows up concentrated in a minority of rows rather than spread evenly. The remaining non-SR-Legacy rows break down further: 1,235 are sourced to "SR28" (an even earlier, specifically-numbered SR release), 294 are labeled "Assumed zero" — USDA's label for a nutrient assumed not to be present in a food at all, rather than one that was measured and came back empty — and roughly 1,061 come from label- or other-source-based derivations. Foundation's 3,407 rows are concentrated in foods that happen to overlap with the 469 items this site's Foundation program has actually gotten around to sampling — Tomatoes, grape, raw is in that overlap; most of the 1,882 FNDDS ingredients are not, simply because Foundation Foods hasn't analyzed a matching raw commodity for them yet.

What's actually missing because Branded Foods isn't here yet?

No UPC lookups, no packaged-product nutrition-facts-panel data, and no manufacturer-specific label values for the vast majority of name-brand products — that entire category of search ("brand X cereal calories") isn't covered by this site today. There's one nuance worth being precise about: a handful of branded-sounding products, like protein bars and shake mixes, already appear inside the Survey/FNDDS ingredient set — because NHANES respondents reported eating specific commercial products, FNDDS named them directly. Those rows still sit inside the 1,882 FNDDS ingredients counted above; they are not part of USDA's separate Branded Foods database, which is a much larger, continuously updated catalog of UPC-coded packaged items that this site has not imported.

How do I know, on this site, which dataset a given number came from?

On an ingredient detail page, check the "Source" column next to each nutrient value — it will read something like "Foundation," "SR Legacy," or a specific legacy code such as "SR Legacy code 11529." That column only exists on ingredient pages because every value on this site's food pages is, by definition, Foundation data; there's no equivalent ambiguity to flag there. The ingredient page's own intro text states plainly that its numbers come from "the USDA FNDDS survey nutrient database," which itself derives from "USDA's Standard Reference and Foundation Foods analyses" — the same two-source blending the tomato and hummus examples show directly.

What should this distinction change about how you use a number here?

If you're cross-referencing a nutrient value against another source, check whether you're looking at a Foundation record or an FNDDS record for that food — they can legitimately disagree, as the tomato energy figures show, and neither one is simply "wrong." If you're searching for a specific packaged product by brand name, this site's current scope won't have it unless that exact product happens to appear inside the FNDDS survey set. And if a number matters enough to act on, the "Source" column on the ingredient page (or the absence of one on a Foundation Food page) tells you in one glance which of USDA's five data types you're actually reading.

Will this site eventually add Branded Foods and full SR Legacy coverage?

That's an open, documented question rather than a settled roadmap item. This project's own Content Team Playbook lists a fresh USDA bulk download — one that would cover SR Legacy, the full Survey/FNDDS dataset, and Branded Foods together — as a Phase 3 item to be evaluated, with a decision memo and import plan scoped only if it's approved. Nothing about that expansion has happened yet, so for now the honest answer to "does this site have Branded Foods" is simply no, and the honest answer to "is this site's Survey data mostly fresh FNDDS analysis" is also no, for the reasons the sourcing breakdown above shows directly. Both limits are worth knowing before you treat a search here as exhaustive.