Home › How scoring works
How YARAI scores a product
This page describes the actual method. The numbers below are read from the app's own configuration when this page is built, so what you read here and what the app computes cannot drift apart.
The short version
Every product starts from a clean sheet. Each ingredient the library flags subtracts from that, weighted by its level of concern and scaled by how much of it the ingredient list implies is present. What comes out is a score out of 100 and a band: Excellent, Good, Fair or Poor.
The exact weights and cutoffs are not published, because a published cutoff is an invitation to formulate to it rather than below it. What is published is everything that decides the result: which ingredients carry concern, at what concentration, and on what evidence.
Where concentration comes in
This is the part that differs most from a presence-based checker. Cosmetic labelling rules require ingredients present above one percent to be listed in descending order of quantity. That means the order of an ingredient list is not decoration. It carries real information about how much of each substance is in the bottle.
YARAI reads that order and scales each penalty by the concentration the position implies. An ingredient sitting near the end of a long list contributes far less to the score than the same ingredient sitting third. A checker that only asks whether an ingredient is present treats those two products identically. They are not identical.
The honest limit, stated plainly: order is not a percentage. Estimating concentration from position is an estimate. It is a better estimate than ignoring position entirely, and it is still an estimate. Ingredients below one percent may be listed in any order at all, so the tail of a long list is the least certain part of any reading.
What the library contains
| Ingredients scored | 203 |
| With a concentration threshold | 178 |
| With an EU concentration limit recorded | 38 |
| Published as public pages | 60 |
| Pairwise interaction rules | 8 |
| Pairs we deliberately do not flag | 4 |
Concern levels draw on published ingredient research and on EU and US regulatory positions, including Regulation (EC) No 1223/2009. Where the EU and the US disagree about an ingredient, the library records both rather than picking one.
Routine interactions are scored separately
A product score describes one bottle. It says nothing about what happens when that bottle is used forty minutes after another one. Interactions are handled by a separate set of pairwise rules, each carrying its own severity, mechanism, evidence tier and source, and each published in full in the conflict library.
That library also publishes 4 pairs we deliberately do not flag, with the reasoning for each. Popular advice treats several of these as incompatible. Publishing the pairs we decided against is the part of a method that is easiest to leave out, and the part that tells you most about how it was built.
What the score is not
- It is not a health rating, and it is not medical advice.
- It is not a prediction that a product will work for your skin. Nothing in an ingredient list can tell you that.
- It is not affected by who made the product, who sells it, or any commercial relationship. See our independence commitment.
- It is not a measurement of concentration. It is an estimate read from list position.
Common questions
Does a brand ever pay to change a score?
No. Scoring is computed from the ingredient list and the data in the ingredient library. There is no field in the scoring path that any commercial relationship can reach, and there never has been.
Why does the same ingredient score differently in two products?
Because position in the ingredient list differs, and position carries information about quantity. An ingredient listed near the end is present in a very different amount than the same ingredient listed near the start.
How accurate is the concentration estimate?
It is an estimate, not a measurement. Regulations require ingredients above one percent to be listed in descending order of quantity, so order is informative, but it is not a percentage. We publish this limitation rather than let the number look more certain than it is.
What does a high score mean?
It means the formula carries few or no ingredients that the library flags at any concern level, at the concentrations the ingredient list implies. It is not a claim about whether the product will work for your skin, which depends on your skin.
If you think a score is wrong
Ingredient data changes and we get things wrong. If a concern level or a threshold on this site looks incorrect, write to us with the source and we will look at it. Corrections change the app and this site at the same time, because both read the same data.
Every figure on this page is generated from the app's own configuration on 2026-09-09. Nothing here is medical advice.