Data sources & attribution

Nutrition data

USDA FoodData Central

The primary nutrition database for whole foods, single ingredients and amino-acid profiles. We use the SR-Legacy and Foundation Foods sub-datasets, which carry full macro- and micro-nutrient breakdowns including individual amino acids. fdc.nal.usda.gov

• Licence: public domain (US Government work).
• Where used: the whole-foods database (/foods), cuisine-dish reconstructions (/cuisine), the leucine source-factor model (/leucine), and the per-amino-acid figures in the methodology.

Note: USDA's Branded Foods dataset is not used for amino acids — branded entries do not carry verified individual amino-acid figures, so for leucine and similar we fall back to the protein-source factor table mapped to SR-Legacy commodity values.

Open Food Facts

Community-maintained product database — barcodes, packaged macros, ingredient lists, category metadata. Used as the primary source for branded snack products. world.openfoodfacts.org

• Licence: Open Database Licence (ODbL) for the database; individual contents under Database Contents Licence (DbCL). Attribution to "Open Food Facts contributors" is required and provided on every snack detail page.
• Where used: the snack database under /highest-protein/[category] and /snacks/[slug].
• When OFF is unavailable or returns degraded data, the ingest pipeline aborts rather than emit guessed nutrition — see methodology.

Manufacturer nutrition labels

For the narrow set of products that USDA does not carry and that Open Food Facts coverage is thin on, we use the manufacturer's own published nutrition panel and disclose the source on the row itself. This is the documented fallback path; it is never used silently. Banza chickpea pasta is one example called out by name on /foods.

Restaurant chains — official nutrition

National chains (McDonald's, Starbucks, Chipotle, Tim Hortons, etc.) publish their own nutrition data and we use it directly. Pages under /restaurants cite the chain's site or PDF as the source for each row.

Independent restaurants — USDA reconstruction

Independent restaurants have no nutrition label and no authoritative database. We reconstruct each dish deterministically from its ingredients using USDA SR-Legacy / Foundation Foods, with the recipe spec and FDC IDs documented. Every reconstructed dish is marked as an estimate in the UI; the methodology and limits are described in full at methodology » cuisine dish estimation.

Scientific sources behind the calculators

Every coefficient in the calculator math is linked to its peer-reviewed source on /methodology and (for leucine) on /leucine. Citations link out to the paper on PubMed or to Google Scholar where a free abstract is the most reliable landing point. The key papers behind the model are:

• Mifflin-St Jeor BMR equation accuracy — Frankenfield et al. 2005 (J Am Diet Assoc).
• Protein per kg targets for athletes & active adults — Phillips & Van Loon 2011; Helms et al. 2014; Morton et al. 2018 (BJSM meta-analysis).
• Older-adult anabolic resistance — Bauer et al. 2013 (PROT-AGE, JAMDA).
• Leucine threshold for MPS — Norton & Layman 2006; Witard et al. 2014.
• Plant- vs animal-source amino-acid composition — Gorissen et al. 2018.
• Collagen vs whey, leucine-matched — Oikawa et al. 2020 (AJCN); Aussieker & van Loon 2023 (PMC10487367).
• GLP-1 lean-mass component — Aronne et al. 2024 (SURMOUNT-1 body-composition analysis).

Keyword & SERP research

Decisions about which calculators and pages to build are driven by DataForSEO keyword and SERP data. The provenance for each round of research lives under research/dataforseo/FINDINGS.md in the repository. This is operational research, not nutrition data — it never appears on a user-facing page.

Found a number that looks wrong?

Please tell us. A photo of the label, a USDA FDC ID, or a paper citation is enough — we will check it and update the row, or note the disagreement publicly if the upstream source disagrees with itself. Contact us.