Food Database
Food Database — A food database is a structured collection of foods and their nutrient compositions used by calorie tracking apps, dietitians, and research instruments to convert food intake into calorie and nutrient values. Major sources include USDA FoodData Central (US), CIQUAL (France), McCance and Widdowson (UK), and proprietary app databases that blend public sources with user contributions.
What is a food database?
A food database is the underlying nutrient reference that every calorie tracking app uses to convert “I ate 100g of chicken breast” into calorie and macro values. A database entry typically contains:
- Food name and identifier
- Per-100g and per-serving nutrient values (calories, macros, micronutrients)
- Source provenance (lab-analyzed, manufacturer-supplied, user-contributed)
- For branded foods: UPC barcode, manufacturer name, product line
The major public databases:
- USDA FoodData Central (US) — the most authoritative public source, ~400,000 entries
- CIQUAL (France) — French food composition reference
- McCance and Widdowson (UK) — UK reference, longest-running national database
- Open Food Facts — international, community-contributed, ~3 million entries with mixed quality
How are food databases used?
Calorie tracking apps build their working databases from blends of public sources, manufacturer feeds, and user contributions. Three approaches dominate:
- Curated proprietary — high editorial standards, smaller catalog (e.g., Cronometer)
- Crowd-sourced — large catalog, variable quality (e.g., MyFitnessPal’s user database, ~14 million entries)
- AI-generated — machine-derived entries from package OCR or recipe parsing (newer apps)
Curated databases are more accurate per entry but may miss less-common foods. Crowd-sourced databases have wider coverage but require users to filter low-quality entries (entries without verified barcode, entries with implausible nutrition values, duplicate entries with different per-serving sizes).
Why food databases matter
Food database quality is the fundamental accuracy ceiling for any calorie tracking workflow. No matter how good photo recognition or barcode scanning is, if the database returns wrong values, output is wrong.
For users, practical implications:
- Apps with curated databases (Cronometer, MacroFactor) tend to have lower error rates
- Apps relying primarily on crowdsourced entries require user vigilance — confirm barcoded items match the package, prefer “verified” entries
- “Unknown” foods entered free-text often pull from low-quality entries; use weighed entries or USDA-sourced foundation foods when possible
Our six-app benchmark found a 4-5x accuracy gap between best and worst apps, partially attributable to database quality. See MAPE for accuracy framework, and USDA FoodData Central for the principal US reference.