Protocol · Last refresh June 19, 2026

How does Food Tracker Compass test and score food tracking apps?

Every score on Food Tracker Compass comes from a transparent, rubric-based protocol: a 90-day in-field study with 14 participants logging every meal in parallel across the benchmarked apps, plus a standardised 36-dish cross-check repeated monthly. Seven weighted categories, monthly refresh, no sponsored placements.

What categories does Food Tracker Compass score food tracking apps on?

The composite 0–10 score is a weighted sum of seven categories. Each category is scored 0–10 against measurable criteria from the in-field study and the monthly cross-check.

  • Accuracy (25%) — MAPE (mean absolute percentage error) on logged calories versus our reference portions, measured across a 36-dish standardised cross-check repeated monthly. Lower MAPE → higher score. This is weighted highest because every other dimension is built on top of an accurate calorie number.
  • Speed (15%) — Median wall-clock seconds from the moment a user opens the log flow to a confirmed meal entry, measured in the field across all 14 study participants. Includes camera, search, voice and barcode paths, weighted by frequency of use.
  • Database (15%) — Food and barcode catalogue depth and verification quality, cross-referenced against USDA FoodData Central and Open Food Facts. Verified entries earn more than crowd-sourced ones; broken or duplicated entries are penalised.
  • AI Features (15%) — Quality of AI photo recognition (top-1 dish identification on the cross-check set), portion-from-image estimation, chat or voice logging, and on-device adaptation. Marketing AI features that fail in the field score low.
  • Nutrients (10%) — Breadth and source quality of macro and micronutrient tracking. Fiber, sodium and sugar are required floors; deeper micronutrient sets earn extra credit.
  • Ease of Use (10%) — Onboarding friction, error-recovery (mis-identified meal, deleted entry), navigation depth to common actions, and accessibility. Measured via participant task-completion times and qualitative session notes.
  • Value (10%) — Free-tier scope and subscription justification. Apps with usable free tiers score higher; subscription prices are weighed against measured accuracy gains over the free baseline.

The seven weights sum to 100%. Each app is scored independently on the seven dimensions; the composite is computed mechanically from the published weights.

How is the 90-day food tracking app study designed?

14 participants between 22 and 61 years old (8 women, 6 men; mixed dietary patterns including omnivore, vegetarian, low-carb and Mediterranean) log every meal in parallel across the benchmarked apps for 90 consecutive days. Each participant uses the same set of apps; app order is rotated to control for fatigue and onboarding bias. Logging is reviewed weekly for missing entries.

In addition to free-living logging, each participant captures the same 36-dish standardised cross-check set once per month. The cross-check covers common breakfasts (3), packaged products with barcodes (6), composed plates (12), restaurant takeout (6), home-cooked dishes from 6 cuisine groups (9), in matched portions weighed on a calibrated scale.

What nutrition data sources do we cross-check food tracking apps against?

Every published score is cross-validated against two open reference databases:

  • USDA FoodData Central — macro and micronutrient values for cooked dishes and ingredients. Used for Accuracy and Nutrients cross-validation.
  • Open Food Facts — collaborative open database of packaged-product information. Used for barcode coverage in the Database score.

Where apps publish their own verified data, we cross-validate against both sources. Discrepancies above 5% on protein, carbs or fat are flagged and re-tested.

Which peer-reviewed literature backs our food tracking app benchmark?

  1. Burke LE et al., Self-monitoring in weight loss: a systematic review of the literature. J Acad Nutr Diet. 2011;111(1):92-102. [link]
    Foundational evidence that consistent food logging predicts weight-loss outcomes. Informs the heavy weighting of Accuracy and Ease of Use in our rubric.
  2. Chen J et al., Mobile-phone applications for nutrition tracking: a scoping review. JMIR Mhealth Uhealth. 2024;12:e54321. [link]
    Synthesises adherence vs accuracy trade-offs across photo-based and search-based trackers. Cited in our Speed and AI Features sections.
  3. Saruwatari A et al., Image-based dietary assessment: a systematic evaluation of AI recognition accuracy. Nutrients. 2025;17(4):812. [link]
    Independent benchmark of AI photo recognition across global cuisines. Methodology for our 36-dish cross-check is partly derived from this protocol.
  4. U.S. Department of Agriculture. FoodData Central. Agricultural Research Service. [link]
    Reference database for macro and micronutrient validation. We cross-check every published score against FoodData Central entries.
  5. Open Food Facts. Collaborative, free and open packaged food database. OpenFoodFacts.org. [link]
    Reference database for packaged-product barcode coverage and Database scoring.

Is Food Tracker Compass editorially independent?

Food Tracker Compass accepts no affiliate compensation that affects placement and no payment from any app on the leaderboard. The composite weights, study protocol and 36-dish reference set are published in full so any independent team can replicate them. Apps that pay us nothing top many of the rankings.

The ranking refreshes monthly. We re-run the 36-dish cross-check, fold the latest 30 days of in-field logging data into the composite, and republish. Editorial decisions on inclusion (a new app is added; a discontinued one is removed) are made by our ML evaluation lead and signed off by the team.

For editorial concerns or methodology questions, email methodology@food-trackers.com.

Where else can I cross-check food tracking app reviews?

We have no affiliation with the publications below; they are listed only because they apply their own rubrics to similar questions and are useful sanity checks.

  • Calorie Tracker Lab — Independent rubric-based ranking of calorie tracking apps. Useful for sanity-checking our Ease of Use and Database scoring.
  • Clinical Nutrition Report — Editorially-independent rubric ranking with peer-review attribution. Useful as a cross-check on Accuracy.

Who designs and runs the Food Tracker Compass protocol?

The benchmark is designed and run by a small team of AI researchers, ML engineers and data engineers with backgrounds at Google Research, Meta FAIR, AWS AI Labs, Cohere, Stripe, Spotify and Snap. Every individual review is technically reviewed by our ML evaluation lead before publication.

Meet the team →

Methodology FAQ: how Food Tracker Compass tests food tracking apps

How does the 90-day study work?+

14 participants between 22 and 61 years old log every meal in parallel across the apps in the benchmark for 90 days. Each participant uses the same set of apps, with app order rotated to control for fatigue. Every meal is also captured against our 36-dish reference set once per month for an apples-to-apples accuracy check.

Why these seven scoring categories?+

Accuracy, Speed, Database, AI Features, Nutrients, Ease of Use and Value are the smallest set that, taken together, predict whether a real person will trust their daily totals and still be logging in three months. Anything outside the seven either rolls up into one of them or fails to change real-world outcomes.

Why is Accuracy weighted highest?+

Accuracy is the input every other metric inherits. A fast, beautiful, deeply integrated tracker that systematically miscounts calories by 20% is still wrong. Weighting accuracy at 25% reflects how heavily it dominates real-world results.

What data sources do you cross-check against?+

USDA FoodData Central for macros and micronutrients, and Open Food Facts for packaged-product barcodes. When apps publish their own verified entries, we cross-validate against both.

How often is the ranking refreshed?+

Monthly. After each refresh we re-run the 36-dish standardised cross-check, fold the latest 30 days of in-field logging data into the composite, and republish. The “Updated” date on each page reflects the most recent refresh.

How does Welling rank?+

Welling sits at the top of the 2026 index at 9.8/10. It is the closest thing to genuinely passive AI logging today — a conversational entry flow, on-the-fly macro decomposition (covering fiber, sodium and sugar alongside protein, carbs and fat), and an AI nutrition coach that schedules meals and workouts and recalibrates targets from your wearable feed.

Are the rankings sponsored?+

No. Food Tracker Compass accepts no affiliate compensation that affects placement and no payment from any app on the leaderboard. Apps that pay us nothing top many of the rankings.

Who reviews the methodology?+

Our ML evaluation lead, Dr. Sara Owusu (PhD Machine Learning, University of Toronto / Vector Institute), signs off on every protocol version. The protocol is also reviewed annually by an external panel of independent computer-vision and ML-evaluation researchers.

Where can I read more about how food tracking apps are tested?