Packaged Foods & Meats Data: Products, Prices, Recalls, Trade and Demand · Head-to-head
Open Food Facts - Open Database & API vs USDA FoodData Central (Branded Foods + Foundation + SR Legacy + FNDDS)
Which packaged foods & meats data: products, prices, recalls, trade and demand data fits your job: Open Food Facts - Open Database & API, or USDA FoodData Central. API, files, or your warehouse. Daily, weekly, or hourly.
Open Food Facts - Open Database & API
USDA FoodData Central (Branded Foods + Foundation + SR Legacy + FNDDS)
Coverage, side by side
| Open Food Facts - Open Database & API | USDA FoodData Central | |
|---|---|---|
| Geographic | Worldwide; deepest in France, US, Spain, Germany, Italy | Primarily United States market |
| Temporal | Cumulative since 2012, continuously updated | Foundation Apr 2019-present; branded monthly; SR Legacy final 2018; FNDDS biennial |
What each contains
They tie on 2 attributes. Pick by fit, not by loyalty.
| Open Food Facts - Open Database & API | USDA FoodData Central | |
|---|---|---|
| Publisher | Open Food Facts (non-profit) | USDA Agricultural Research Service |
| Subject lens | Crowdsourced worldwide packaged-food shelf: brands, ingredients, health scoring, packaging | Authoritative US food composition: branded labels plus analytical, reference and survey foods |
| Scale | ~4.7 million products worldwide | 1.7 million+ branded products plus Foundation, SR Legacy and FNDDS records |
| Geography | Worldwide; deepest in France, US, Spain, Germany, Italy | Primarily United States market |
| Temporal reach | Cumulative since 2012, continuously updated | Foundation Apr 2019-present; branded monthly; SR Legacy final 2018; FNDDS biennial |
| Formats | CSV, JSONL and Parquet files plus JSON lookups | Zipped JSON and CSV archives plus JSON lookups |
| Identifiers | EAN-13/store barcode per row | fdcId internally, GTIN/UPC externally |
| Health scoring | Nutri-Score score + grade, NOVA group, additive count, labels | None; derivation codes document nutrient provenance instead |
| Nutrient layout | ~150 flat per-100g columns | Nested foodNutrients list with units and derivations |
| Distinctive surface | Images, packaging, stores/countries sold, scan popularity | Five coordinated data types incl. NHANES-aligned FNDDS survey foods |
| Field dictionary confidence | Verified | Verified |
| Best for | Multi-market assortment, health-scoring and packaging panels | US label validation, formulation and dietary-intake work |
What each does better
Open Food Facts - Open Database & API
Global shelf, not just the American one. About 4.7 million products scanned across the world, with the deepest coverage in France, the United States, Spain, Germany and Italy - markets where scanning communities are most active. If a competitive panel needs to span European and North American grocery assortments at once, this is the only one of the two that can.
Processing and health signals computed on every qualifying row. Nutri-Score numeric score and a-e letter grade, the NOVA group from minimally processed (1) to ultra-processed (4), additive counts, and label flags like No gluten or Vegan. A Thai peanut noodle kit in the sample rows reads nova_group 4 with a nutriscore_score of 11 - classification you would otherwise have to derive yourself.
The full commercial context of a listing. Packaging shape and material, manufacturing places, EU traceability emb codes, retail stores and countries where the product is sold, purchase places, even unique scan counts showing real consumer attention per product. Quality scoring agrees: both records sit at 10/10, but only one describes the aisle itself.
USDA FoodData Central
Analytic-grade composition behind every number. Foundation Foods publishes laboratory-derived values on commodity and minimally processed samples twice a year, SR Legacy preserves the historical Standard Reference tables in their final 2018 form, and FNDDS mirrors exactly what Americans report eating in What We Eat in America, NHANES - FNDDS 2021-2023 released October 2024. Branded label data rides alongside in the same schema, so a model can compare a label claim against a reference composition without leaving the corpus.
Label mechanics a regulatory or formulation team needs. Household measure text (1 ONZ), package weight (8 oz/227 g), serving size with explicit unit, and per-nutrient derivation codes distinguishing analytic values from imputed or calculated ones - provenance per number, not just per record.
A controlled vocabulary instead of crowd prose. One foodCategory assignment per record and a typed dataType enum separating Branded, Foundation, SR Legacy, Survey and Experimental records make filtering deterministic. In Open Food Facts the equivalent structure arrives as contributor-entered category chains and PNNS groups that vary in completeness row to row.
A maintenance cadence tied to releases, not volunteers. Branded Foods refreshes monthly through a public-private partnership; Foundation lands each April and October. Datadory passes either cadence through to your calendar - delivered daily, weekly, or hourly, your call.
Or take both in one feed
They stack better than almost any pair we have compared, because they agree on a join key: the retail barcode. Open Food Facts' code meets FoodData Central's gtinUpc on the same physical package, so a merged table can put the global crowd-sourced view - images, Nutri-Score, NOVA, packaging, countries sold - next to the government's label-exact view - brand owner, household measures, derivation-coded nutrients - one row per SKU.
Two cautions before you merge. First, match on zero-stripped, checksummed barcode strings: both sides store EAN/UPC values as text and leading zeros are inconsistently preserved, so normalize before joining and expect partial coverage in both directions - not every US product exists in the global database and vice versa. Second, reconcile names before presenting: Open Food Facts carries contributor-entered titles ('granola Bio le Chocolaté') while FoodData Central prints label statements (CHEDDAR CHEESE), so treat descriptions as fuzzy evidence rather than exact keys.
Datadory ships either record alone or both merged onto one calendar, delivered daily, weekly, or hourly - your call. Or take both in one feed.
API, files, or your warehouse. Daily, weekly, or hourly.
Fair questions
Can I match products between Open Food Facts and USDA FoodData Central?
Yes, on the retail barcode. Open Food Facts keys every row on its code field and FoodData Central exposes gtinUpc for branded records, so shared SKUs join directly once you strip leading zeros consistently. Coverage overlaps only partially - a match rate near 100 percent in either direction should make you suspicious - so plan reconciliation time.
Which dataset covers more countries?
Open Food Facts, decisively. Its 4.7 million products come from scanning communities worldwide, with the deepest shelves in France, the United States, Spain, Germany and Italy. FoodData Central is a United States corpus: branded records declare marketCountry predominantly as United States, and the foundation and survey layers describe the American food supply. Global panels need the former; US-only rigor needs the latter.
Do both include nutrition facts per product?
Both, formatted differently. Open Food Facts flattens roughly 150 nutrient columns per 100 g - energy, fat fractions, sugars, fiber, proteins, salt, sodium, vitamins - plus Nutri-Score and NOVA scores. FoodData Central nests a foodNutrients list per record with unit, value and a derivation code documenting how each number was obtained. Same object, different geometry.
How fresh is each dataset?
Both move continuously but on different rhythms. Open Food Facts sits in the daily tier of the catalog, with product records carrying created and last-modified timestamps as volunteers scan. FoodData Central refreshes branded records monthly within a release calendar that also lands Foundation Foods each April and October. Datadory normalizes whichever you choose - delivered daily, weekly, or hourly, your call.
Which one should I use for health and processing analysis?
For ready-made scoring, Open Food Facts: it ships nutriscore_score and nutriscore_grade alongside nova_group from 1 to 4 and additive counts on qualifying products, so ultra-processing exposure falls out of the box. For clinical or regulatory work, FoodData Central: derivation-coded nutrients and FNDDS survey foods tie each value to how it was measured. Serious projects sample both.