How ratings work
Every rating comes from facts with a source, run through the same written rules. This page explains those rules, where the facts come from, and how sure we are.
Counted from Formulary’s data on November 5, 2026.
The six ratings
A rating is about the ingredient itself: how good or bad it is for skin, based on the evidence.
Good studies show a large benefit, a difference most people would notice.
A real but modest proven benefit. Or a reviewed ingredient that does a job on the skin, such as cleansing, moisturizing, softening or protecting, with no notable concern.
Does a job in the formula, like thickening or preserving it, with nothing notable for or against your skin. Also used when studies show only a tiny effect.
A notable concern, such as a fragrance allergen the EU requires labels to name.
A serious concern: banned in EU cosmetics, judged not safe by the EU’s scientific safety committee, a common allergen (reactions in more than about 3% of people patch-tested in Europe), or a standard skin irritant such as SLS.
Not enough research or not rated yet. We’d rather say so than guess.
Where the facts come from
Each source has one job, and only certain sources can move a rating in a certain direction.
| Source | What it’s used for |
|---|---|
| CosIng (EU) | The EU’s official ingredient database: names, identifiers and what each ingredient does in a formula. |
| EU cosmetics law | Banned and restricted ingredients, and the fragrance allergens labels must name, along with the opinions of the EU’s scientific safety committee (SCCS). |
| Patch-test and irritation studies | How often an ingredient causes allergic reactions or irritation. Together with EU law, these are where every concern comes from. |
| Published studies | Studies found on PubMed that pass the quality bar below. They are the only route to Best and Good for a proven benefit. |
| Open Beauty Facts | Only to decide which ingredients to research first, by how often products list them. It’s never a source for a fact. |
US safety reviews (the Cosmetic Ingredient Review) aren’t used yet. Every reason Formulary shows is written in our own words, never copied from another site.
Which studies count
A study only supports a benefit if it is a systematic review or a controlled trial, done on human skin, with the ingredient applied to the skin, and it isolates the ingredient, comparing the same formula with and without it.
Narrative reviews, studies without a control group, lab or animal work, studies of the ingredient taken by mouth, and tests of a whole product are left out, and aren’t shown as support. The reviewer also checks that a study is peer-reviewed, not retracted, and used a realistic amount.
How sure we are
When a benefit earns Good or Best, the pill also says how strong the evidence is. This is worked out from the studies, never typed in by hand.
- Strong evidence: a systematic review, or two or more controlled trials with at least one independent.
- Moderate evidence: at least one good independent trial (blinded, about 20 or more people, 4 weeks or longer).
- Limited evidence: anything else that passes the bar, for example only company-funded trials.
“Independent” means the paper says it wasn’t company-funded, or it says nothing about funding and every author works at a university or hospital.
Evidence strength doesn’t set the rating; the size of the effect does. One study that passes the bar is enough to count, and the label tells you how much weight to give it.
From facts to a rating
Each fact is either a benefit or a concern, linked to at least one source. The rules are checked in this order, and the first one that fits sets the rating. Concerns come first, so a proven benefit never hides a real risk.
- A serious concern, with strong or moderate evidence → Worst
- A notable concern, with strong or moderate evidence → Bad
- A benefit with a large effect → Best
- A benefit with a modest effect → Good
- A reviewed ingredient that does a job on the skin, with no notable concern → Good
- A tiny effect, or a reviewed ingredient that works on the formula with nothing above → Neutral/Average
- Anything else → Unknown
Weak evidence of a concern, and mild concerns (such as an EU limit on how much may be used), are shown to you but never change a rating. An EU-listed fragrance allergen is always at least Bad, even where some studies suggest it rarely causes reactions.
Reviewer decisions
Rules can’t cover every case. A reviewer can set a different rating, higher or lower, but only with a written reason, which Formulary shows. A reviewer decision can always make a rating worse. It can never make a rating better than EU law allows: an ingredient with a notable EU concern stays at Bad or below, and one with a serious EU concern stays Worst.
Right now, 75 of the 156 reviewed ratings are reviewer decisions. We say so because it’s a judgment, not a rule, and you should be able to tell the difference.
Draft and approved
Every fact starts as a draft. Only approved facts count toward a reviewed rating. Facts taken straight from EU law are approved automatically; everything else is approved when a person reviews the ingredient. Today 106 facts are approved and 11 are drafts.
Who stands behind a rating
Every rating in Formulary carries one of these labels, so you always know where it came from.
| Label | What it means |
|---|---|
| Human reviewed | A person checked this ingredient’s facts and sources and signed off the rating. Ingredients are reviewed in order, the most common first. If new facts or rules later change a reviewed rating, it’s flagged to be reviewed again. |
| From EU regulation | Not reviewed yet, but set directly by EU law, like the 82 fragrance allergens labels must name. |
| Not Human Reviewed Yet | In Formulary’s data, but not reviewed yet. The rating is worked out by the same rules, counting draft facts and what the EU database says the ingredient does. Treat it as a first draft. |
| AI estimate | Not in Formulary’s data at all. Claude, an AI model, gives an estimate so the list isn’t left blank. No person has checked it. |
Research is gathered with the help of software and AI tools: they search PubMed, read the EU lists and draft facts. Those drafts don’t count toward a reviewed rating until a person approves them.
What ratings don’t tell you
- How much is in the product. Labels rarely say, so a rating describes the ingredient, not the amount in your product.
- How your skin will react. Everyone’s skin is different. An ingredient rated Best can still irritate you.
- Everything. Formulary only knows what has been studied, and many ingredients haven’t been reviewed yet.
Not medical advice. Ratings are general information about ingredients. Ask a doctor or dermatologist about allergies, reactions or skin conditions.
See a rating you think is wrong? Tell us, and include the ingredient’s name.