Database Quality Tiers

Database quality tiers is shorthand for the stratified hierarchy of food-composition data sources by methodological rigour. An entry from USDA Foundation Foods, produced by documented sampling and analytical chemistry with per-nutrient uncertainty estimates, is a different epistemic object from an entry typed into a community-submission interface by a user who may or may not have owned a kitchen scale. Dietary-assessment work that treats the two interchangeably is producing figures with invisible quality variance.

A reasonable tier hierarchy

From strongest to weakest:

1. Analytical (Tier 1). USDA Foundation Foods. Documented sampling plan, AOAC-method chemistry, per-nutrient uncertainty estimates. The strongest available tier for most consumer foods.
2. Compiled-analytical (Tier 2). USDA SR Legacy, where entries are drawn from analytical work but the sampling and methodology metadata are less consistently documented. Still acceptable for most dietary-assessment purposes but with known staleness issues.
3. Modelled (Tier 3). FNDDS recipe entries, where mixed-dish composition is computed from ingredient proportions and ingredient-level nutrient profiles. Appropriate for aggregate dietary-recall studies; inherits any uncertainty in the underlying ingredient data.
4. Manufacturer-submitted (Tier 4). USDA Branded Foods and its commercial equivalents (Nutritionix product feed, Edamam branded entries). Label-derived figures inheriting FDA rounding and the ±20 per cent tolerance.
5. Professional-submitted (Tier 5). Restaurant-menu databases where a trained nutritionist or the restaurant's compliance team has entered per-dish figures. Quality varies; often calibrated against internal cooking specifications but without third-party analytical verification.
6. Community-submitted (Tier 6). User-typed entries in crowdsourced database (MyFitnessPal community, open-source Open Food Facts entries without verification). No provenance, no method, no uncertainty band. Should be treated as unverified.

What the tier mix looks like in major apps

Each consumer calorie-tracking app has a different mix of these tiers in its catalogue. Apps that prioritise verified entries (Cronometer, MacroFactor, some research tools) lean heavily on Tiers 1–2 for whole foods and Tier 4 for packaged items, exposing the provenance in the interface. Apps that prioritise catalogue size over provenance (MyFitnessPal, to the extent community entries are allowed; some aggregator-driven apps) include a larger Tier-6 fraction. A user's choice between these apps implicitly chooses a tier mix, and the per-meal accuracy consequences follow from that choice.

Tier metadata as a disclosure issue

The most transparent apps display per-entry tier information — "USDA Foundation" or "User-submitted, unverified" — so the user can weight the figure accordingly. Most consumer apps do not. A 2021 Nutrients review of food-database quality labelling in the ten most-downloaded calorie-tracking apps found that only two surfaced provenance to the user. The remaining eight returned database entries as undifferentiated results. An entry that says "chicken breast, 165 kcal/100g" in an app's search results carries no indication of whether that figure is a USDA Foundation analytical measurement or a five-year-old community entry typed by someone who rounded differently.