Most food scanner apps start from a public nutrient profile, usually Nutri-Score, then layer their own judgments on top: a penalty per additive, a bonus for organic certification, sometimes a processing class such as NOVA. The number on your screen is a weighted blend of those parts, and the weights and the additive list are where each app makes its own choices. When those choices are not published, a 45 out of 100 could mean "high in salt" or "contains one emulsifier the app dislikes", and you cannot tell which.
This article is educational information, not medical advice.
What goes into a food score?
Almost every score has three ingredients: a nutrient profile, an additive judgment and a processing or purity judgment. Apps differ in how they weight them and in how much of each they disclose.
The nutrient profile is usually borrowed from Nutri-Score, the front-of-pack label France introduced in 2017. It grades a product from A to E per 100 g or 100 ml by counting unfavorable elements, which the current algorithm lists as energy, saturated fatty acids, sugars, salt and non-nutritive sweeteners, against favorable elements: fiber, protein, fruit, vegetables and pulses. The label is used or recommended by a steering committee of seven European countries, and the updated algorithm took effect in France on 16 March 2025.
The processing judgment usually comes from NOVA, a classification proposed by Carlos Monteiro and colleagues that puts every food in one of four groups: unprocessed or minimally processed foods, processed culinary ingredients such as oils, sugar and salt, processed foods such as canned fish and fresh bread, and ultra-processed foods, defined as formulations made mostly or entirely from substances derived from foods and additives, with little if any intact group 1 food.
The additive judgment is the app's own. Each app decides which additives count against a product and by how much, and this is the part that is most often kept private.
How does Nutri-Score turn nutrients into a letter?
Nutri-Score is a points system. Unfavorable components add negative points, favorable components add positive points, and the net score maps to a letter. Because the inputs are the nutrition table on the pack and the fruit and vegetable content, two people with the same label data get the same letter. That reproducibility is the property a food score should have and rarely does.
The formula also changes in public. On 24 April 2023 the scientific committee published its beverage update: milk, plant-based drinks and milk-based drinks moved into the beverage algorithm, skim and low-fat milk sit in the most favorable categories, water stays at A, low-sugar drinks are better separated from sugary ones, and the update limits the benefit of swapping sugar for sweeteners. Operators were given until the end of 2025 to relabel.
An app that uses Nutri-Score inherits these changes. Whether it tells you which version of the algorithm it is running is another matter.
How does an app like Yuka build its number?
Yuka publishes its weights, which makes it a useful reference. According to its help center, nutritional quality is 60% of the score and is calculated with the Nutri-Score method, additives are 30%, and an organic label is worth the remaining 10%.
Each additive gets one of four risk levels: risk-free, limited risk, moderate risk or high-risk. The override is the part most users miss: if an additive Yuka considers high-risk is present, the maximum score for the product is set at 49 out of 100, and in that case, in the app's own words, this criterion can represent more than 30% of the score. A product with a Nutri-Score of A and one such additive cannot pass 49.
That is a defensible design choice. It is also a judgment about a single ingredient that decides more than the whole nutrition table, so the evidence behind each high-risk rating carries the entire score.
Where do two apps disagree about the same product?
They rarely disagree about the nutrient profile, because it is public. They disagree about additives, about how much an organic label should be worth, and about whether processing should count at all.
| Component | Typical source | Who sets the rule | Can you check it? |
|---|---|---|---|
| Nutrient profile | Nutri-Score points | Scientific committee of seven countries | Yes, the algorithm is published |
| Processing class | NOVA groups | Open Food Facts infers it from additives and product categories | Partly, the inference rules are public |
| Additive penalty | The app's own risk list | The app | Only if the app publishes the list and its sources |
| Organic bonus | National or EU labels | The app decides the weight | Only if the weight is published |
| Caps and overrides | The app | The app | Only if disclosed |
The additive column is where regulators and apps can part ways. In the EU, EFSA assesses additives and sets an acceptable daily intake where the data allow, and more than 300 substances are authorized, while the European Commission and member states decide on approval and conditions of use. An app that penalizes an authorized additive is disagreeing with that assessment, which it is free to do, but the disagreement should be visible and sourced.
There is also a difference between hazard and risk that many app ratings flatten. When IARC and JECFA reviewed aspartame in 2023, IARC did hazard identification, asking whether the substance can cause cancer at all, while JECFA did risk assessment at actual consumption levels and reaffirmed the acceptable daily intake of 40 mg per kg of body weight. A score that deducts the same points for "possibly carcinogenic" as for "exceeds the ADI in a normal serving" is mixing two different questions.
Why should the formula be public?
First, reproducibility. If the same inputs always give the same output, a score can be checked by anyone with the label. Nutri-Score has this property; a score with hidden weights does not, and a hidden formula can change between app updates without the number on the pack telling you.
Second, attribution. Each deduction is a claim about an ingredient. If the claim points to an EFSA opinion, an IARC monograph, an FDA rule or a peer reviewed study, you can read the source and decide whether you agree. If it points to nothing, you are trusting a product team's reading of the literature.
Third, the borrowed parts are already public. Nutri-Score's algorithm is published, NOVA's rules are published, EFSA's opinions are published. An app that keeps only its own layer secret is hiding the one part that was never reviewed by anyone else.
This is the standard we set for QScan: a planned score out of 100 with cited explanations. QScan is coming soon; its scoring formula is not yet published, and this article explains general scoring approaches rather than its implemented method.
How should you read any food score?
Find the nutrient profile behind it and check whether it is per 100 g or per serving. Find the additive list and ask what source rates each additive and whether that source is a regulator's risk assessment or a hazard classification. Check whether one ingredient capped the score, because a cap tells you more about the app's rules than about the food.
Then check the date. Nutri-Score changed its algorithm in 2023 with a transition to the end of 2025, and regulators re-evaluate additives continuously, so a score computed on last year's rules can be wrong on this year's evidence. A formula you can read is the only one you can date.
Frequently asked questions
Is Nutri-Score the same as the score in a scanner app?
No.
Nutri-Score is a public A to E nutrient profile computed per 100 g or 100 ml from energy, saturated fat, sugars, salt, sweeteners, fiber, protein and fruit and vegetable content.
Scanner apps typically use it as one input and add their own additive penalties, organic bonuses and caps, which is why an A product can still score poorly in an app.
Why do two apps give the same product different scores?
Because they agree on the public part and disagree on the private part.
The nutrient profile is usually the same Nutri-Score calculation.
The additive risk list, the weight given to an organic label, and any cap triggered by a single ingredient are each app's own choices, and unless they are published you cannot see which one moved the number.
What does NOVA measure?
NOVA measures degree of processing, not nutrition.
It puts foods in four groups, from unprocessed or minimally processed through culinary ingredients and processed foods to ultra-processed formulations made mostly from substances derived from foods plus additives.
Open Food Facts infers the group from the ingredient list, using marker additives and product categories, so it is an estimate rather than a measurement.
Does a high additive penalty mean the additive is banned?
Usually not.
In the EU more than 300 additives are authorized after EFSA assessment, and an app that penalizes one of them is applying a stricter standard than the regulator.
That can be reasonable, but the app should say which study or opinion it is relying on, and whether the concern is a hazard at any dose or a risk at realistic intake.