What "processed" actually means
"Processed" gets used as a synonym for bad, which makes it nearly useless as a word. The classification most often referenced in research is more specific than that, and more useful.
Published
Four groups, not two
The NOVA classification, developed at the University of São Paulo and referenced widely in nutrition research, sorts food by what was done to it and why - not by how healthy it is judged to be.
| Group | What it covers | Examples |
|---|---|---|
| 1 · Unprocessed or minimally processed | Whole foods, and foods altered only by cleaning, drying, grinding, freezing or pasteurising. | Rice, milk, plain oats, frozen vegetables |
| 2 · Processed culinary ingredients | Substances extracted from group 1 and used to cook with. | Oil, sugar, salt, ghee |
| 3 · Processed foods | Group 1 foods with group 2 ingredients added, usually to preserve them. | Cheese, salted nuts, canned pulses, most breads |
| 4 · Ultra-processed | Formulations made largely from industrial ingredients and additives used to make the result palatable and shelf-stable. | Soft drinks, instant noodles, most packaged snacks |
Group 3 is where a lot of ordinary food sits, and it is not a warning label. The distinction that carries the most weight in the literature is between group 4 and everything else.
How to spot group 4 on a pack
- Ingredients you would not find in a home kitchen - protein isolates, modified starches, invert syrup, hydrogenated oils.
- Additive classes whose job is sensory: colours, flavours, flavour enhancers, sweeteners, emulsifiers, thickeners.
- A long list where the recognisable food is not near the front.
None of these is disqualifying by itself. An emulsifier in a nut butter that would otherwise separate is doing a different job from a colour added to make a snack look like the picture on the front.
Why Filtizy scores it lightly
Filtizy reads two things off the ingredient list: how many ingredients are named, and how many flagged additive categories appear. Ingredient count is deliberately the lighter-weighted of the two, because count alone is a weaker signal than what the ingredients actually are. Neither is treated as a verdict on the product - they are two factors among eleven, and every one of them is shown broken out on the product page.
Filtizy scores are informational and are not medical or dietary advice. The scoring model is a documented heuristic, published in full on the methodology page, not a clinical assessment.
See it on real products
Every product page shows the same figures broken out, with the factor scores they produced.