Food Health Score Algorithm & Methodology

General principle

The score starts at 100 and penalties are subtracted based on four independent factors. The final result is clamped between 0 and 100. A higher score means fewer health concerns were identified across all dimensions.

These weights reflect the relative impact each dimension has on long-term health outcomes according to current epidemiological literature. Nutritional composition remains the strongest predictor of diet-related chronic disease, which is why it carries the greatest weight. Additives carry the second-largest weight because certain substances have been linked to carcinogenicity, endocrine disruption, and inflammatory responses independent of a product's macronutrient profile.

score = 100 - nutritional_penalty - additive_penalty - processing_penalty - personal_penalty

1. Nutritional quality (50 points max)

This dimension is based on the Nutri-Score nutritional profiling algorithm. Nutri-Score was developed by a team of nutritional epidemiologists led by Professor Serge Hercberg at the French National Institute of Health and Medical Research (INSERM) and Université Sorbonne Paris Nord. It was officially adopted by the French government in 2017 and has since been endorsed by Belgium, Germany, the Netherlands, Spain, Luxembourg, and Switzerland.

How the algorithm works

The algorithm assigns points based on two sets of nutrients, calculated per 100 grams of product:

Negative components (0 to 10 points each, higher is worse):

• Energy density (kJ per 100g)
• Total sugars (g per 100g)
• Saturated fatty acids (g per 100g)
• Sodium (mg per 100g)

Positive components (0 to 5 points each, higher is better):

• Dietary fiber (g per 100g)
• Protein (g per 100g)
• Fruits, vegetables, legumes, nuts, and certain oils (% of composition)

The raw nutritional score is computed as: negative points minus positive points. This produces a value from -15 (best possible) to +40 (worst possible), which maps to letter grades A through E.

Raw Nutri-Score ranges for each grade

Conversion to our penalty system

We convert the raw Nutri-Score value to a 0-50 penalty scale using linear interpolation. A product with the best possible nutritional profile (raw score of -15, grade A) receives 0 penalty points. A product with the worst possible profile (raw score of +40, grade E) receives the full 50-point penalty. Everything in between is scaled proportionally.

When the exact numerical score is unavailable but the letter grade is known, we use fixed penalties: A = 5, B = 15, C = 25, D = 35, E = 45 points. When neither value is available for a product, a neutral penalty of 20 points is applied.

Scientific validation

The Nutri-Score algorithm has been validated in multiple prospective cohort studies. Research published in The BMJ (2020) demonstrated that consumption of foods with poor Nutri-Score grades was associated with higher all-cause mortality (HR 1.07, 95% CI 1.03-1.11 per 5-point increase). A 2022 study in The Lancet Regional Health - Europe showed Nutri-Score outperformed other front-of-pack labeling systems in predicting diet-related chronic disease risk across 22 European countries.

2. Additive risk assessment (30 points max)

Each additive present in a product is individually classified into one of four risk tiers. Classification is based on cross-referencing regulatory decisions and toxicological evidence from multiple authoritative bodies:

• The U.S. Food and Drug Administration (FDA) - GRAS status revocations, color additive delistings, market withdrawal orders
• The European Food Safety Authority (EFSA) - safety re-evaluations, revised acceptable daily intakes, scientific opinions
• The International Agency for Research on Cancer (IARC) - carcinogenicity classifications (groups 1, 2A, 2B, 3)
• EU Regulation (EC) No 1333/2008 on food additives
• Peer-reviewed toxicological literature on genotoxicity, endocrine disruption, microbiome disruption, and inflammatory markers

Risk tiers

High risk substances

This tier includes substances for which there is clear regulatory consensus on potential harm:

• Titanium dioxide (E171) - banned in the EU since August 2022 following EFSA's 2021 opinion concluding genotoxicity concerns could not be ruled out
• BHA / butylated hydroxyanisole (E320) - classified IARC group 2B (possibly carcinogenic to humans)
• Potassium bromate - banned in the EU, UK, Canada, Brazil, and China; classified IARC group 2B due to carcinogenicity in renal tissue
• Aspartame (E951) - classified IARC group 2B in July 2023
• Sodium nitrite (E250) - precursor to N-nitroso compounds, a class classified IARC group 2A (probably carcinogenic)
• Brominated vegetable oil - banned by the FDA in July 2024 after review of bioaccumulation data
• FD&C Red No. 3 / erythrosine (E127) - delisted by the FDA in January 2025 following long-standing evidence of thyroid tumors in animal studies

Moderate risk substances

This includes synthetic azo dyes such as Allura Red / Red 40 (E129), tartrazine / Yellow 5 (E102), and Sunset Yellow / Yellow 6 (E110). All three are subject to mandatory warning labels in the EU under Regulation (EC) 1333/2008 Annex V due to their association with hyperactivity in children (McCann et al., The Lancet, 2007). Also included: carrageenan (E407), linked to intestinal inflammation in animal models (Tobacman, Environmental Health Perspectives, 2001); BHT (E321), suspected endocrine disruptor; polysorbate 80 (E433), linked to gut barrier disruption (Chassaing et al., Nature, 2015); and sulfur dioxide (E220), a documented asthma trigger in sulfite-sensitive individuals.

Low risk substances

Substances generally recognized as safe with minor documented concerns. This includes sodium benzoate (E211, can form benzene in combination with ascorbic acid under specific conditions per FDA's 2006 benzene survey), potassium sorbate (E202), calcium propionate (E282), emulsifiers like lecithin (E322) and mono- and diglycerides (E471), and artificial sweeteners including sucralose (E955) and acesulfame potassium (E950).

No risk substances

Naturally derived substances with extensive safety records and no documented adverse effects. This includes organic acids (citric, lactic, malic, tartaric, ascorbic acid), natural pigments (beta-carotene, riboflavin, curcumin, annatto, carmine, beet juice, chlorophylls, spirulina), plant-based thickeners (pectin, guar gum, xanthan gum, gellan gum, gum arabic, agar, cellulose), and flavor enhancers (monosodium glutamate, disodium inosinate, disodium guanylate).

The total additive penalty is the sum of all individual penalties found in the product, capped at 30 points. Our database covers over 80 substances common in American grocery products (Walmart, Costco, Target, Kroger, Whole Foods), using both E-number and FDA common names for accurate matching. The database is updated within 48 hours of any new FDA ruling, EFSA re-evaluation, or IARC classification change.

3. Processing level (10 points max)

This dimension uses the NOVA food classification system, developed by Professor Carlos Monteiro and colleagues at the University of São Paulo (NUPENS). NOVA classifies foods by the extent and purpose of industrial processing. The system is referenced by the World Health Organization, the Food and Agriculture Organization of the United Nations, and the Pan American Health Organization.

The four NOVA groups

Group 4 products are characterized by ingredients not used in home cooking: hydrolyzed proteins, modified starches, hydrogenated oils, and cosmetic additives (flavors, colors, emulsifiers, humectants, glazing agents). When NOVA data is not available for a product, a neutral penalty of 5 points is applied.

Why processing matters independently of nutrients

A 2024 umbrella review published in The BMJ (Lane et al.) analyzed 45 meta-analyses covering nearly 10 million participants. It found consistent associations between high ultra-processed food intake and increased cardiovascular disease mortality (RR 1.50, 95% CI 1.37-1.63), type 2 diabetes (RR 1.12, 95% CI 1.06-1.17), depressive outcomes (RR 1.22, 95% CI 1.16-1.28), and all-cause mortality (HR 1.14, 95% CI 1.08-1.20). These associations persist after adjustment for BMI, energy intake, and individual nutrient content, suggesting that industrial processing contributes to health risk through mechanisms that nutrient profiling alone cannot capture.

4. Personal dietary profile (10 points max)

During onboarding, you indicate what you personally try to avoid. These responses create a personalized penalty layer capped at 10 points total.

Personalization rules

Seed oil avoidance (+5 if triggered). Applied when the ingredient list contains canola oil, rapeseed oil, soybean oil, sunflower oil, corn oil, cottonseed oil, safflower oil, rice bran oil, grapeseed oil, or generic "vegetable oil." This reflects concerns over the omega-6 to omega-3 ratio in industrially refined seed oils (Simopoulos, Biomedicine & Pharmacotherapy, 2002; DiNicolantonio & O'Keefe, Open Heart, 2018).

Added sugar sensitivity (+5 if triggered). Applied when sugar exceeds 10g per 100g, the threshold set by the WHO 2015 guideline on free sugars intake.

High sodium watch (+5 if triggered). Applied when salt exceeds 1.5g per 100g, the "high" threshold per EU Regulation 1169/2011 Annex XIII. The American Heart Association recommends no more than 2,300 mg sodium per day.

Gluten-free diet (+5 if triggered). Applied when the product contains wheat, barley, rye, spelt, kamut, or triticale. Detected through allergen declarations and ingredient text analysis. Serves users with celiac disease (~1% of U.S. population per Rubio-Tapia et al., American Journal of Gastroenterology, 2012) and non-celiac gluten sensitivity.

Interpreting the final score

The score is continuous from 0 to 100. For interpretation:

• 80 to 100: Minimal concerns. Strong nutritional profile, no high-risk additives, low processing.
• 60 to 79: Minor flags in one dimension. Still reasonable for regular consumption.
• 40 to 59: Notable concerns. Best consumed occasionally rather than daily.
• 20 to 39: Significant issues across multiple dimensions.
• 0 to 19: Critical health flags in most or all dimensions.

Scores below 50 may also result from the automatic cap triggered by the presence of a high-risk additive.

Limitations and transparency

No scoring system fully captures the complexity of human nutrition. We are transparent about the following:

• Nutri-Score evaluates per 100g, not per serving. Products consumed in very small quantities (mustard, hot sauce) may score lower than their real dietary impact warrants.
• Additive risk reflects population-level evidence. Individual sensitivity varies.
• NOVA depends on accurate processing level identification, which is not always deterministic from the ingredient list alone.
• The score does not account for dietary context. A food scoring poorly in isolation may fit into a balanced overall diet when consumed rarely and in small amounts.
• Unknown additives receive a conservative low-risk classification. Our database is comprehensive for the U.S. market but does not cover every possible substance.

We continuously refine our methodology as new evidence emerges. When regulatory agencies update their positions or new meta-analyses are published, we review and adjust our database accordingly.