Pick up two loaves of bread. The bakery loaf lists four ingredients: flour, water, salt, yeast. The supermarket loaf lists twenty-two, including soybean oil, high-fructose corn syrup, monoglycerides, and calcium propionate. Flip both packages over and the nutrition panels are nearly identical — similar calories, similar carbs, similar sodium per slice. By the numbers on the back, they are the same food. They are not the same food.
The system built to see that difference is NOVA, a classification developed by nutrition researchers at the University of São Paulo and since used in reports by the World Health Organization and UN agencies. NOVA mostly ignores nutrients. It asks a different question: how much industrial transformation happened between the farm and your mouth? Every food lands in one of four groups, from an apple (group 1) to a toaster pastry (group 4).
That question turns out to matter. Ultra-processed foods — NOVA group 4 — now supply more than half of the calories American adults eat, and diets built on them are consistently linked to worse health outcomes even after researchers adjust for fat, sugar, and salt. Processing is not just a proxy for bad nutrition numbers. It carries its own signal.
What NOVA actually measures
NOVA classifies the purpose and extent of processing, not the nutrients that result. Washing, freezing, drying, fermenting, and pasteurizing all leave a food in group 1 — frozen broccoli and plain yogurt count as minimally processed because the processing exists to preserve the food, not to reformulate it. The line NOVA draws is at industrial formulation: products assembled largely from refined substances (protein isolates, modified starches, extracted oils) and cosmetic additives (colors, flavors, emulsifiers) rather than made from recognizable ingredients.
A useful shorthand: could a reasonably equipped home kitchen produce this? Bread, cheese, canned tomatoes — yes. A snack cake with a two-year shelf life and eleven ingredients you can't buy at a grocery store — no. That gap is what the four groups formalize.
The four groups, with everyday examples
| Group | What it is | Everyday examples | Nutrication points (of 35) |
|---|---|---|---|
| 1 — Unprocessed / minimally processed | Whole foods, possibly cleaned, frozen, dried, ground, or pasteurized | Apples, eggs, plain oats, frozen broccoli, milk, plain yogurt, dried lentils, fresh chicken | 35 |
| 2 — Processed culinary ingredients | Substances pressed, refined, or extracted for cooking | Olive oil, butter, sugar, salt, vinegar | 28 |
| 3 — Processed foods | Group 1 foods preserved or combined with group 2 ingredients, usually two to five in total | Canned sardines in olive oil, cheese, fresh bakery bread, canned beans with salt, jarred vegetables | 18 |
| 4 — Ultra-processed foods | Industrial formulations built from refined substances and additives | Soda, packaged snack cakes, instant noodles, chicken nuggets, flavored yogurt, most supermarket bread | 4 |
The last column shows how Nutrication's engine scores each group, and the shape is deliberate. Group 2 sits close to group 1 — olive oil and butter are concentrated, but they aren't formulations. Group 3 is a real step down. Group 4 falls off a cliff: it keeps 4 of 35 possible points, a roughly 89 percent penalty on this measure just for being ultra-processed.
Notice how one aisle can span groups. Plain yogurt is group 1; the fruit-on-the-bottom version with modified starch and flavorings is group 4. A bakery baguette is group 3; the soft sliced loaf with emulsifiers and preservatives is group 4. Same food category, different manufacturing philosophy.
Why processing level predicts health outcomes
Large cohort studies in France, Spain, the United States, and elsewhere keep finding the same pattern: the bigger the share of ultra-processed food in a diet, the higher the rates of obesity, heart disease, and type 2 diabetes. Crucially, the association survives adjustment for the obvious suspects — calories, sugar, sodium, saturated fat. Something about the category itself tracks with worse outcomes.
The most striking evidence is experimental. In a tightly controlled NIH feeding trial, adults lived on-site and were offered ultra-processed and minimally processed menus matched for calories, sugar, fat, fiber, and macronutrients. Told to eat as much or as little as they liked, participants on the ultra-processed menu ate roughly 500 more calories a day and gained weight. On the unprocessed menu, the same people lost it.
The leading explanations are mechanical rather than mysterious. Ultra-processed foods tend to be calorie-dense, soft, and fast to eat, so more goes down before fullness registers. They pair refined carbohydrates with fat and salt in combinations rarely found in nature, which makes stopping harder. And the additives themselves function mostly as a marker: an ingredient list carrying emulsifiers and artificial flavors reliably signals a product engineered for shelf life and palatability. For a deeper look at the category, see our guide to ultra-processed foods.
Where the classification gets messy
NOVA is a lens, not a verdict. A group number says nothing about portion size or nutrients — group 1 white rice can still spike blood sugar, and group 3 cheese still carries plenty of saturated fat and sodium. That is exactly why nutrient-based systems like Nutri-Score exist, and why no serious grading system should rely on processing level alone.
There is also a data problem. Food databases are crowd-sourced, and their NOVA tags overshoot. A tin of sardines in olive oil — sardines, oil, salt — is a textbook group 3 food, yet it frequently gets tagged group 4 simply because it comes in a can. An app that takes those tags at face value will punish some of the most defensible foods in the store.
How Nutrication uses this
NOVA processing is the largest of the six measures behind every Nutrication grade, worth up to 35 points before the grade is normalized to 100 — more than Nutri-Score (30), ingredient flags (20), nutrient density (20), glycemic load (15), or fat quality (15). The mapping is fixed: group 1 earns 35 points, group 2 earns 28, group 3 earns 18, and group 4 earns 4. You can see how all six measures combine in how we grade.
When the food database already carries a NOVA tag, the engine uses it. When it doesn't, the engine infers the group from the ingredient list: two or more recognized industrial additives means group 4; exactly one means group 3; a culinary marker like salt, sugar, oil, butter, or vinegar means group 2; a clean list means group 1. For a group 4 food, the app names up to four of the specific additives that drove the call and invites you to tap each one for a plain-English explanation — our ingredient red flags measure separately penalizes the genuinely concerning ones. And when a group 4 product contains no cataloged additives at all, a list running past eight ingredients is presented as the tell: characteristic of an industrially formulated product.
The engine also second-guesses the database in one specific direction. If a product arrives tagged group 4 but has five or fewer ingredients, zero recognized additives, and matches a recognizable whole food — canned fish, jarred vegetables — it gets reconciled down to group 3. The app says so openly:
Some databases tag it "ultra-processed," but a recognizable, short ingredient list is more consistent with a processed food — so we've graded it that way.
Chips and similar snacks are deliberately never rescued this way; a short ingredient list doesn't undo frying and extrusion. And when there is no ingredient list at all, NOVA contributes nothing — the measure is excluded and the grade is computed from the remaining evidence rather than a guess.
The group also echoes through the other measures. In a group 4 food, micronutrient credit is cut in half, because vitamins in ultra-processed products are usually synthetic fortification — a fortified sugary cereal shouldn't read as nutrient-dense as spinach. Fat-quality credit is likewise halved when the unsaturated fat comes from refined oil in an ultra-processed product: a bag of chips shouldn't earn the same fat credit as a handful of almonds just because the fryer used sunflower oil.
What to actually do
- Read the ingredient list before the nutrition panel. Length and recognizability tell you the NOVA group faster than any nutrient number.
- Use the kitchen test. If you couldn't plausibly assemble it at home, treat it as group 4 and buy it with your eyes open.
- Don't fear groups 2 and 3. Olive oil, cheese, canned fish, and real bread are how people have eaten well for centuries.
- Swap within the aisle. Plain yogurt over flavored, bakery bread over soft sliced, oats over puffed cereal — same shelf, one or two groups better.
- Watch the share, not the exceptions. The research links harm to diets dominated by ultra-processed food; an occasional group 4 item in a mostly group 1–3 diet is noise.