Methodology

Testing methodology

Every score that appears on Nutrient Metrics derives from the protocol on this page. The protocol is documented, repeatable, and audited. Our editorial team owns it; our medical reviewer signs off on every clinical claim it produces.

Inclusion criteria

An app is eligible for evaluation in our general-evaluation rubric if it meets three minimum thresholds: a verified food database of at least 50,000 entries, an AI photo or barcode logging path, and a published privacy policy. Apps that do not meet inclusion are documented but not scored on the six-criterion rubric.

The six-criterion rubric

Criterion	Weight	Measurement protocol
Accuracy	30%	Mean absolute percentage error between app-reported energy and weighed reference value, measured against the DAI 2026 reference meal set (n = 240 meals across six dietary patterns) and our internal NM-REF-2026 corroboration set.
Database depth and verification	20%	Total verified entries, per-entry nutrient field completeness, and source attribution audited against USDA FoodData Central.
AI photo recognition	15%	Top-1 dish-identification accuracy and portion-estimation MAPE on the NM-IMG-2026 internal test set (n = 180 photos across six cuisines).
Macro and micronutrient granularity	15%	Number of nutrient fields tracked, configurability of macro targets, and presence of an extended micronutrient panel mapped against DRI categories.
User experience	10%	Friction-of-correction time, onboarding completion rate in our 30-participant usability cohort, sustained 30-day adherence in the testing pool.
Price and value	10%	Annual cost relative to category median, normalized for free-tier feature coverage and any usage caps.

Accuracy Index calculation

The Accuracy Index is the per-app score on the accuracy criterion, expressed on a 0–100 scale and aggregated across the six dietary patterns in the DAI 2026 reference set. The calculation is:

Accuracy Index = 100 - (MAPE × scaling_factor)
where:
  MAPE = mean( |reported_kcal - reference_kcal| / reference_kcal ) × 100
  scaling_factor = 4.5  (calibrated so the category-median MAPE maps to ~70/100)
  Floor = 0; Ceiling = 100.

The scaling factor is published and constant across the evaluation cycle. It is calibrated once per cycle so that the category-median MAPE produces a score of approximately 70 out of 100. An app with ±1.1% MAPE produces an Accuracy Index of approximately 95.0; an app with ±6.4% MAPE produces approximately 71.2; an app with ±9.4% MAPE produces approximately 57.7. The Accuracy Index is then weighted at 30% in the overall score.

Reference standard

The reference standard for accuracy measurement is the Dietary Assessment Initiative's 2026 validation meal set (DAI-VAL-2026-01), corroborated by our internal NM-REF-2026 set. The DAI 2026 set includes 240 weighed meals prepared and analyzed under controlled laboratory conditions, spanning six dietary patterns (omnivorous, plant-forward, low-carbohydrate, Mediterranean, low-FODMAP, and culturally diverse). Each meal has a documented per-component weight and a laboratory-verified nutrient profile.

Inter-rater reliability

Every measurement protocol on this site is structured to be repeatable by a second analyst. Inter-rater reliability is enforced at a Cohen's kappa of 0.85 or higher for categorical judgments and at a within-3% MAPE delta for continuous measurements. Any measurement that fails inter-rater reliability is re-run by a third analyst and flagged in the audit trail.

Conflict of interest policy

Nutrient Metrics accepts no sponsored placements, maintains no affiliate relationships with any app evaluated here, and accepts no financial compensation from any product developer. The full conflict-of-interest policy is documented on our editorial standards page.