USDA FoodData Central (Branded Foods + Foundation + SR Legacy + FNDDS)

One identifier for everything Americans eat. USDA FoodData Central (Branded Foods + Foundation + SR Legacy + FNDDS) is the federal food-composition corpus, packaged by Datadory for commercial work: about 1.7 million branded grocery products keyed by barcode, sitting alongside laboratory-analyzed Foundation Foods samples, the frozen SR Legacy reference tables, and the FNDDS survey foods underneath national dietary research. One record per food or product, one schema across all five data types, and nutrients stated per 100 grams or per the serving size on the pack.

Every branded record carries what the label shows - brand owner, brand name, ingredient statement, serving size, household measure, package weight - plus a full nutrient vector whose values arrive tagged with their derivation, so an analytical measurement never masquerades as a calculation. That mix of shelf-level commerce and bench-level chemistry in a single table is why it anchors the Packaged Foods & Meats catalog. Get a sample of this dataset.

One product, thirteen visible columns, one nested nutrient array. A branded cheese record exactly as it ships:

fdcId                   : 2057648
description             : CHEDDAR CHEESE
dataType                : Branded
gtinUpc                 : 094395000172
brandOwner              : Grafton Village Cheese Co, LLC
brandName               : GRAFTON VILLAGE
ingredients             : UNPASTEURIZED MILK, SALT, TRUFFLES (TRUFFLE, WATER,
                          SALT, AROMAS), TRUFFLE OIL, CULTURES, ENZYMES.
servingSize             : 28 g
householdServingFullText: 1 ONZ
packageWeight           : 8 oz/227 g
foodCategory            : Cheese
marketCountry           : United States
publishedDate           : 2021-10-28

The gtinUpc column is the join key - point it at retail scans, recall notices or e-commerce listings and the label travels with the barcode. dataType separates Branded shelf goods from Foundation, SR Legacy, Survey (FNDDS) and Experimental records, so one query holds supermarket SKUs and laboratory commodity samples side by side.

Sixteen documented fields ride on every record, typed and defined in full in the dictionary below - nothing held back for the sales call. Three conventions matter. Identity is layered: a permanent fdcId, a human-readable description, and - for packaged goods - the gtinUpc barcode plus brandOwner and brandName. Label fidelity: ingredients, servingSize, servingSizeUnit, householdServingFullText and packageWeight reproduce the pack copy rather than paraphrasing it. Provenance: publishedDate and modifiedDate timestamp each record, marketCountry names the market, and the nested foodNutrients array carries a derivation code beside every value.

Geography - the United States market. marketCountry rides on branded records and reads predominantly United States; this is the aisle American groceries walk down.

Temporal - four clocks running at once. Foundation Foods samples run April 2019 to present. SR Legacy is preserved as finalized in 2018, deliberately unchanged. FNDDS documents the 2021-2023 What We Eat in America survey cycle. Branded records carry their own publishedDate and modifiedDate across the modern era, so recency is a property you can query rather than a promise you take on faith.

Granularity - one row per food or product, with nutrient vectors per 100 grams or per the stated serving size.

API, files, or your warehouse. Daily, weekly, or hourly.

• Shelf-level competitive intelligence - brandOwner, brandName and gtinUpc turn a category into a corporate map: who owns how many SKUs of cheddar, in which foodCategory, with which ingredient statement.
• Reformulation and label-drift tracking - diffing ingredients against modifiedDate catches recipe changes, allergen additions and clean-label swaps product by product.
• Nutrition modeling and ML - about 1.7 million labeled products with nutrient vectors and category tags is training-set-sized raw material for estimation, classification and recommendation work.
• Dietary research replication - FNDDS survey foods reproduce the food coding behind What We Eat in America, so intake studies reconcile against the same definitions.
• Compliance and recall screening - joining barcodes against enforcement records puts a recall exposure figure on every assortment before the regulator calls.
• Assortment and market sizing - SKU counts by category and brand give retailers and consultants a bottom-up census instead of a syndicated estimate.

Competitive intel & product teams get the corporate map of the shelf - who owns what, under which label, formulated how. E-commerce operators enrich listings with authoritative ingredient and nutrient copy keyed on barcode. Data scientists & ML engineers inherit a clean one-row-per-product schema with typed enums and nested nutrients, ready without scraping. Market researchers & consultants size categories bottom-up from SKU censuses. Journalists, academics & students cite the canonical answer to what is actually in the American shopping cart.

Four things worth knowing upfront. The branded universe skews heavily to the United States market - marketCountry says so row by row, and international assortments need a companion dataset. Ingredient statements are literal pack copy: upper-case, allergen-styled, occasionally wrapped across lines, so NLP work wants a light normalization pass. The four non-branded data types answer different questions - SR Legacy is frozen at its 2018 form and FNDDS describes survey composites, not SKUs - so name the question and we will pull the right dataType. And nutrient values arrive with derivation codes because they are not all the same kind of number; models that ignore the distinction inherit its noise.

• USDA FoodData Central API - the platform-side companion record for the same corpus, useful when the question needs record-level lookup rather than the whole table.
• Open Food Facts - crowd-sourced labels alongside federal curation; the head-to-head sets out where each wins.
• openFDA Food Enforcement (Recalls) API - recall events that join to branded barcodes, turning the shelf census into an exposure register.
• Spoonacular Food API - Product Search - a grocery-product search layer for applications built on top of the same products.
• USDA ERS Food Expenditure Series - the dollars spent where FoodData Central counts nutrients; quantity meets spend.
• Best packaged foods meats datasets - where this sits in the wider category stack.