Crowdsourced Database
Crowdsourced Database — A crowdsourced database is a food database whose entries are submitted, edited, or verified primarily by app users rather than by paid editorial staff or by reference institutions like the USDA. Crowdsourced databases offer enormous coverage — millions of entries — at the cost of variable quality, with substantial error rates on user-submitted nutrition values.
What is a crowdsourced database?
A crowdsourced food database is one whose entries come primarily from app users. The pioneer example is MyFitnessPal, which built a database of more than fifteen million entries by allowing users to add foods that weren’t already there — and to copy entries created by other users. Cronometer, Lose It, and most other large consumer trackers run hybrid models: a curated core of editorially-verified foods plus a much larger periphery of user-submitted entries.
Crowdsourcing solves a coverage problem that no curated database can match. The USDA’s FoodData Central has high-quality entries for whole foods and many branded products, but it is not going to have a separate entry for “Joe’s Coffee Shop blueberry muffin, downtown Boston.” A crowdsourced database, given enough users, will. The trade-off is quality: when a user submits a calorie value, no reference institution checks it against the manufacturer’s label. The submitter may have read the label correctly, may have eyeballed the size, may have entered values for a different brand by mistake.
How is it measured in our testing?
Calorie Tracker Lab’s database-quality criterion (20% of the 100-point rubric) explicitly evaluates how an app handles the crowdsourced-vs-verified distinction. Four sub-dimensions:
- Coverage. A 50-item search panel including grocery SKUs, restaurant chain items, regional dishes. Coverage measures whether the app has the food at all.
- Verification. We sample 20 entries per app and check whether displayed values match the manufacturer label or the USDA FoodData Central entry. Apps that flag verification status (so the user knows whether the entry is from a verified source or from another user) score higher.
- Freshness. Restaurant menus rotate. We check whether the app’s chain-restaurant entries reflect the current menu values.
- Noise resilience. We submit ambiguous queries (“pizza”) and score how the app surfaces canonical entries vs. dumping low-quality user submissions.
Apps that mix user-submitted and verified entries without distinguishing them in the UI are penalized. Apps that surface user-submitted entries first by default — with no verification flag — are penalized more heavily.
Why it matters in calorie tracking apps
For users, the crowdsourced-database trade-off translates into a daily-use risk: the app you trust to log lunch may be returning a calorie value that a stranger on the internet entered three years ago, with no fact-check. In MyFitnessPal’s database, for example, the same food can have ten or twenty different user-submitted entries with calorie values varying by 30% or more. The user who picks the first result without checking introduces a systematic error into their tracking.
The mitigation, in current apps, is to: (1) always prefer entries flagged as verified or sourced from the USDA, where the app surfaces this; (2) for branded foods, scan the barcode rather than searching by name (the barcode lookup goes to the manufacturer’s verified label); (3) cross-check the calorie value against the package label or USDA FoodData Central when in doubt. For a verified food database, the trust calculus is different — see that entry. See our free-tier entry for the related question of whether the verified database is locked behind a paywall.