Methodology

Last updated: February 25, 2026

FoodKnower provides two scores to help you understand both how industrially processed a food is and how likely it is engineered to encourage overeating.

Food Integrity Score

The Food Integrity score estimates how industrially processed a food is. It is anchored to the NOVA classification framework and extended with ingredient-level signals that often indicate heavy processing.

Key inputs

• Additive burden — emulsifiers, sweeteners, flavour enhancers, colours, and preservatives. Higher counts and higher-impact additive types reduce integrity.
• Industrial markers — ingredients commonly used in industrial processing (for example: maltodextrin, modified starch, glucose/corn syrups, hydrogenated fats, protein isolates).
• Functional role stacking — how many engineered functions are present, such as sweetener systems, texture engineering, preservation, colour, and umami enhancement.
• Category priors — adjustments based on product category to avoid over-penalizing categories where certain ingredients are structurally required.
• Guardrails — protects traditional and single-ingredient foods (for example: plain bread, plain yogurt, basic grains) from being over-scored as ultra-processed.

Portion Control Score

The Portion Control score estimates how easy or difficult it is to overeat a food. It combines nutrient thresholds with ingredient formulation signals associated with hyperpalatability.

Key inputs

• Nutrient analysis — evaluates nutrition data against the three hyperpalatable clusters from Fazzino et al. (2019): Fat + Sodium, Fat + Sugar, and Carbs + Sodium.
• Ingredient formulation analysis — scans the ingredient list for palatability signals, including: sugar/fat/salt positioning, synergistic reward combos (sugar x fat, fat x salt, sweet x salty), and snack-architecture patterns (chips/crackers, chocolate, flavoured-salty).
• Beverage penalty — liquid calories tend to bypass satiety signals, so beverages receive a penalty.
• Simplicity bonus — short, clean ingredient lists receive a small downward adjustment in risk.

Important notes

• Scores are estimates designed for consistency and user understanding, not medical advice.
• When data is incomplete (for example, AI-estimated nutrition), the score is conservative and may be less accurate.
• Methodology evolves as new research and validation data become available.