How the Grain Score is calculated.
A score is a starting point, not a verdict. Below is every input, every weight, and every assumption we make — in plain language. One scoring engine produces every score in the app; when it changes, this page changes with it.
A —The three components.
Nutrition quality — 70%
A Nutri-Score-style algorithm over per-100g data: energy, sugar, saturated fat, and sodium count against; fiber and protein count for. Drinks are judged on stricter thresholds than foods.
Additive quality — 20%
Every ingredient on the label is checked against three evidence tiers — hazardous (−25), moderate (−10), limited (−4) — plus a clean-ingredient bonus (+3 each, capped at +15). Details in section C.
Processing & certifications — 10%
NOVA processing group sets the base: NOVA 1 → 100, NOVA 2 → 75, NOVA 3 → 45, NOVA 4 → 15. Certified organic adds +15, non-GMO +10.
Personal overlay
Your allergies, ingredients you avoid, and preferred certifications are layered on top of the base score on your device. The base score is the same for everyone; the overlay is yours.
B —The formula.
base = nutrition × 0.70 + additives × 0.20 + processing × 0.10
(clean-ingredient NOVA 1–2 products: 0.45 / 0.40 / 0.15)
floors clean-ingredient products ........ never below 55
simple whole foods ............... never below 78
ceilings Nutri-Score E → max 35 · Nutri-Score D → max 60
NOVA 4 ultra-processed → max 75
per-serving sugar / sodium / saturated-fat caps
Products with genuinely clean ingredient lists — NOVA 1–2, no hazardous or moderate additives — shift to 45/40/15 weighting, so a dense whole food like dry oats isn't punished for calories alone.
C —The additive tiers.
Hazardous — −25 each, caps the additive component at 35
Reserved for direct regulatory action (bans, revoked authorizations) or genotoxic-carcinogen evidence at dietary relevance. Currently 11 ingredients: sodium nitrite, sodium nitrate, potassium bromate, propylparaben, butylparaben, ethoxyquin, titanium dioxide, brominated vegetable oil, azodicarbonamide, partially hydrogenated oils, and Red 3 (FDA revoked food use, January 2025).
Moderate — −10 each
Regulatory scrutiny or significant published evidence at dietary dose: high-fructose corn syrup, sodium benzoate, aspartame and other high-intensity sweeteners (acesulfame, sucralose, saccharin), Red 40, Yellow 5 & 6, Blue 1 & 2, BHA, BHT, TBHQ, carrageenan, polysorbates, added phosphates, Class III/IV caramel color, erythritol, maltodextrin.
Limited — −4 each
Recognized as safe but highly processed, or a concern only at high intake or for sensitive groups: dextrose, corn syrup, MSG, yeast extract, mono- and diglycerides, potassium sorbate, artificial flavor, generic caramel color.
Clean bonus — +3 each, capped at +15
Genuinely health-positive ingredients, not basic staples: olive and avocado oil, rolled oats, quinoa, lentils, chia and flax seed, nut butters, cocoa.
Tier placement tracks regulator evidence — FDA, EFSA, IARC, JECFA — and is re-graded as that evidence evolves. A 2026 science audit moved aspartame and the other high-intensity sweeteners, the common synthetic dyes, BHA/BHT/TBHQ, carrageenan, and polysorbates from hazardous down to moderate: a hazard label at any dose is not the same as risk at dietary dose. Of the dyes, only Red 3 — the one with a direct FDA revocation — remains hazardous.
Refined seed oils (canola, soybean, sunflower, corn, and similar) carry a separate penalty: −15, reduced to −5 when organic, expeller-pressed, or cold-pressed. We're explicit about what this is: no regulator classifies refined vegetable oils as hazardous, and some evidence cuts the other way. Flagging them is a deliberate, cautious Grain stance — shown to you as a stance, not as settled science.
EU-style labels are normalized before scoring: an ingredient panel that says "E621" scores exactly the same as one that spells out "monosodium glutamate."
D —Floors and ceilings.
Whole-food floors
Clean-ingredient products never score below 55. Simple whole foods — NOVA 1, three or fewer ingredients, zero flagged additives, like plain oats, rice, or frozen fruit — never score below 78.
Nutri-Score ceilings
A Nutri-Score grade E caps the final score at 35; grade D caps it at 60. A clean-looking ingredient list can't outrun bad nutrition.
Ultra-processed ceiling
NOVA 4 products cap at 75 — unless the product is certified organic with a near-spotless additive profile, the one rescue path.
Per-serving ceilings
Sugar ≥30g/serving caps a food at 50 (≥20g at 65; drinks stricter, from 12g). Sodium ≥1500mg caps at 45 (≥1000mg at 60, ≥700mg at 75). Saturated fat ≥15g caps at 50 (≥10g at 65).
E —A worked example.
Sweetened corn-flake cereal (illustrative)
Assumed inputs, per 100g: 375 kcal · 37g sugar · 470mg sodium · 2.4g fiber · contains BHT · NOVA 4 · Nutri-Score D
F —Sources and verification.
Cited, or honestly empty
Ingredient risk entries link to primary sources in the app — EFSA, FDA, IARC, NTP, WHO. Where no honest authority source exists, the citation list stays empty rather than padded.
Region-aware
Six regulatory regions — US, EU, UK & Ireland, Canada, Australia & New Zealand, Switzerland — with the matching food authorities (FDA, EFSA, FSA, Health Canada, FSANZ, FSVO) built into Grain's AI context. Five languages.
Grounded AI
AI product answers are grounded in live web search, and a "real" product that returns zero grounding citations is treated as a red flag — not shipped as fact. Responses are AI-generated and may not always be accurate.
Vision-checked images
Product photos pass an automated vision quality gate before they're shown to you.
G —What we don't know.
Nutrition science is not settled. Regulators disagree. Formulations change without notice. The Grain Score is our best attempt at synthesis, built from public data and AI assistance — it is an estimate, and it will be wrong sometimes. When an ingredient is genuinely contested — seed oils are the clearest example — we tell you it's a stance, not settled science. And when the formula changes, this page changes with it.