How a Product Score Works: Four Questions About the Ingredients

Reviewed by the LabReadAI medical team
How a Product Score Works: Four Questions About the Ingredients

Any score on a pack or in a catalogue is someone's opinion. An opinion is worth exactly as much as it can be checked. So here is our whole method in plain words: what the score in the catalogue and the ingredient check section stands for, where the numbers come from, what is left out and how to recalculate it yourself. The score is LabReadAI's opinion by an open formula — not medical advice and not a certificate.

The score is four questions about the ingredients

For every product we ask four questions. Each gets an answer from 0 to 1 and a confidence from 0 to 1 — how much the answer rests on exact data (a printed percentage, a reference intake, an absorption curve) rather than on an estimate from the ingredient order. We multiply the answers rather than average them: the score is their confidence-weighted geometric mean times 10. The difference matters. With a plain average, a product whose only proven active sits at trace level would score 7.5 out of 10 — three good answers hiding one failure. All four questions have to hold at once: the formula must work, carry the active at a working share, cover the job and not irritate — "good on average" does not count here. A balanced formula is unaffected: four answers of 0.85 still give 8.5. No question is "more important": only confidence sets the weight. A question we cannot answer (no data in the reference) gets no score at all — we say so in words. A question that has nothing to measure for this product (a cleansing base in a leave-on oil) is simply not asked.

The questions depend on what the product is. A magnesium capsule and a shampoo face different demands — measuring them with one ruler would not be fair.

Supplements, vitamins, sports nutrition

Question What we check Example
Does the substance work? Whether the substance has a proven effect as a supplement. Strong studies — 1.0; moderate — 0.75; weak — 0.4; none — 0.15 (levels follow the GRADE ladder) magnesium, vitamin D, creatine — 1.0; collagen — 0.75; biotin for hair — 0.4
How much will absorb? How much will really be absorbed from a serving compared with the daily intake — accounting for the form (some forms the body takes up almost whole, others by a quarter; data from studies that compared forms in people) and the absorption curve: the larger the tablet, the smaller the share it takes up. A form without comparative data is taken as "somewhere between the known forms of the substance" — with lower confidence 200 mg of magnesium chelate — 0.73; the same 200 mg as oxide — 0.29; 420 mg of citrate — 1.0
Does the formula cover the job? For every purpose — "iron deficiency", "sleep and stress", "bones", "hair and nails" — it is known which substances are needed and at what dose. We look at how much of that set the product covers. The purpose comes from the label; if the label is silent, we find it by the map of links between substances and purposes (see below) magnesium without vitamin B6 for "sleep and stress" — 71 %; iron without vitamin C — 57 %
Anything excess or risky? Each problem has its own probability: a serving above the upper level (the higher the excess, the higher), debatable additives by severity (say, the colourant E171), conflicting combinations. Probabilities multiply as independent — so one serious problem is not "diluted" by three minor ones pure creatine powder — 1.0; 420 mg of magnesium (above the upper level for supplements) — 0.35

Cosmetics and skin care

Question What we check Example
Do the actives work? Whether the main ingredients have a proven effect on skin. Strong studies — 1.0; moderate — 0.75; weak — 0.4; none — 0.15. An active that is clearly scarce in the product weighs less in the answer niacinamide, ceramides, retinol — 1.0; "banana extract" — 0.15
Is there enough active? If the maker states a percentage, we compare it with the working range. Above the range is no minus: no extra benefit is expected, but the product does not work worse for it; where a substance has a known tolerance threshold (niacinamide above 5 %, benzoyl peroxide above 5 %), the excess moves to the question "what may irritate" — where it belongs. If the percentage is not printed, we read the ingredient order as an interval: an active before the preservatives and fragrance (above the "1 % line") is present at 1 % to 100/k %, where k is its place in the list; after them — from traces to 1 %. The answer is the average over the interval, the confidence — its width. If the product's pH is known, we account for how much of an acid is actually active in it niacinamide 4 % — 1.0 (confidence 100 %); the same niacinamide fourth in the list with no percentage — about 0.7 at 60–70 % confidence; at the end of the list — about 0.1
Does the formula cover the job? Every skin purpose — acne, dark spots, wrinkles, dryness, sensitivity — has a set of actives with proven effect and working shares. Actives of one front (say, all exfoliating acids) are alternatives: any one is enough; different fronts complement each other. We look at how much of the set the product covers and point out what is missing and what could be added without conflicting with the formula. The purpose comes from the label, or from the map of links if the label is silent niacinamide for "acne" covers the "inflammation and sebum" front, a keratolytic is missing — a BHA or azelaic acid could be added
What may irritate? We do not invent the probability of a reaction — we take the measured one: how many people in a hundred had a positive skin patch test to that ingredient. Fragrance — 12.8 in 100, methylisothiazolinone — 13.8, formaldehyde releasers — 7.7, while parabens are 1.0 and iodopropynyl butylcarbamate 0.4. When the label names the fragrance allergens one by one (citral, isoeugenol, coumarin — 26 of them under EU rules), the risk is computed from the ones named. Also here: pore-clogging ingredients in leave-on face products, conflicting pairs (retinoid + acid), a formula too acidic for niacinamide, and exceeding a tolerance threshold. Probabilities multiply as independent no debatable groups — 1.0; a cream with fragrance — 0.87; a shampoo with fragrance and a harsh surfactant — 0.79

Hair and hygiene (shampoos, conditioners, shower gels, toothpaste) — the same questions as cosmetics (purposes — dandruff, oily scalp, damaged hair), but instead of "is there enough active" we ask what it cleans with: every cleansing agent has an irritation index relative to the harshest one — sodium lauryl sulfate (SLS = 1; laureth sulfate ≈ 0.5; betaines ≈ 0.3; glucosides and isethionates ≈ 0.2 — from comparative studies), and each one's share in the blend is estimated from its place in the list. The answer is one minus the weighted irritation. Conditioners, masks and leave-on oils are not asked this question — they have nothing to clean with. In rinse-off products irritants count more gently: they stay on the skin only briefly (each group has its own rinse-off severity).

Sun care — the first question is special: what it protects from — what share of the 290–400 nm spectrum (UVB + UVA) the filters in the formula cover: every filter has a known absorption range, and we measure the length of their union. Zinc oxide covers about 80 % (up to 380 nm), avobenzone + octocrylene — the whole spectrum, octinoxate alone — UVB only.

The colour of the score is not decoration: green means none of the four answers came out bad. A product with a high total but one failing answer stays blue — otherwise the colour would promise more than the bars beneath it show.

Prices are not in the score: they change daily and differ by seller, while the ingredients do not. Nor are "label transparency" or "claim honesty": they measured how nicely things were printed, not whether the formula is good.

Where the numbers come from

Every substance in our reference is described with links: reference intakes — from the Rospotrebnadzor guidelines (MR 2.3.1.0253-21), safe upper levels — from EFSA, form absorption and absorption curves — from comparative human studies, the evidence for cosmetic actives — from peer-reviewed work, comedogenicity — from Fulton's 1989 table, filter spectra — from SCCS and UV-filter reviews, acid pKa values — from chemistry handbooks, the share of reactions to debatable ingredients — from dermatology patch-test networks (NACDG, 4,121 and 50,799 people tested; Krijl 2022 for the 26 fragrance allergens). Even the method's "service" numbers — how much a level of evidence weighs, the severity of a debatable group, a comedogenicity rating — live as separate files with a source reference, not hidden in the program: they can be read and disputed. Where comparative data do not exist, the reference stores "no data", and the score says so in words instead of giving a zero.

How we read a cream's ingredient list

Cosmetics have no reference intakes, and makers rarely print the active's percentage. But by labelling rules, ingredients present at 1 % or more are listed in descending order, and everything below 1 % in any order. Preservatives, fragrance and thickeners are almost always below 1 % — so the first of them marks the boundary. An active before that boundary is present at no less than 1 % and no more than 100/k %, where k is its number in the list (k ingredients, each no smaller than it, cannot add up to more than 100 %); an active after the boundary — from traces to 1 %. Within the interval we do not know where the truth is, so we take the average over all possible shares and honestly show the confidence: the wider the interval and the closer to the threshold of action, the lower it is. Ceramides, retinoids and piroctone olamine work from hundredths of a percent — for them a place after the line changes little; niacinamide needs 2 %, and after the line it is almost certainly "for the label". If the brand publishes the product's pH, we account for it too: acids and vitamin C work only in their non-ionised share, which falls as pH rises by the Henderson–Hasselbalch law.

The map of links: how we tell what a product is for

We keep substances and purposes as a map: magnesium is linked to "sleep and stress" and "bones", vitamin C to "iron deficiency" (it helps iron absorb), and "hair and nails" builds on "iron deficiency". When the label does not say what the product is for, we "launch" its actives across this map and see which purposes they reach most often — the same principle search engines use to find the most connected pages. So the card honestly shows what the formula is really for, and whether it is narrow (one purpose) or spread thin (a little for several).

Pairs of substances count as numbers from studies, not as words: calcium from 165 mg in the same dose cuts iron absorption roughly in half, while vitamin C boosts it — 1.5× at 25 mg and almost 10× at 1,000 mg; piperine raises curcumin absorption twentyfold. These multipliers go straight into the dose calculation. There are triples too: iron + calcium + vitamin C counts only when all three are in the formula. Pairs without a number — "a retinoid and an acid irritate twice as much" — go into the tolerance question with their own severity. And "who it does not suit" can carry a dose threshold: vitamin B6 harms nerves from 100 mg a day, not at any dose — so the chip appears only on such products.

Where the ingredients come from: label, register, official records

  1. The label is the main document: by law the ingredient list on the pack must be complete and in descending order.
  2. Registers and official databases — for supplements the formula is fixed at registration (SGR), for imports — the NIH DSLD label database, for cosmetics — the full list on the brand's website.
  3. The leaflet and the maker's website — doses, contraindications, form.
  4. The seller's listing on a marketplace is the least trusted source: in our check of 192 listings almost half had an empty "Ingredients" field.

When sources disagree, we go by the label. If the ingredients are known only from the seller, we give no score and write "ingredients not confirmed".

What else is on the card

Every ingredient on the card says what it is and why it is there. For those with something to add beyond their role — fragrance, sulfates, preservatives, pore-clogging oils — we keep our own descriptions with sources, severity and who should care. The rest is ordinary formulation technology, and we name its role from the open European cosmetic ingredient inventory (CosIng, mirrored by Open Beauty Facts): thickener, emulsifier, preservative, solvent. Such an ingredient is marked as reference data and does not move the score — the inventory is there so that no line of the formula is left unexplained. If even the inventory does not know it, we say so.

We went through the questions and bad reviews of 555 products across eight marketplace categories and collected what people ask most. So besides the score the card answers: what the product is for, who it suits and who it does not (age, pregnancy, skin and hair type, conditions), what is not in the formula (fragrance, sulfates, alcohol, parabens), when and how to use it, when to expect an effect, what is normal in the first weeks and when to stop. All of it comes from the reference with sources, not from copy written for the product.

What is not in the score

  • Marketplace reviews and ratings. They are counted per brand and say nothing about the ingredients.
  • Advertising and paid placement. A brand can raise its score in one way only — by improving the formula.
  • The answer "does it suit me". The score is about the product, not you. Whether a supplement suits you depends on your tests, diet and other medicines: that is the match by blood tests; for skin and hair — the cosmetic check and the hair care match.
  • Medical claims. We do not write "cures" and do not replace a doctor: the blood magnesium level, ferritin or 25(OH)D is assessed by a doctor.

How to check our score yourself

Take the label and answer the four questions of your category. For a supplement: the form — from the tables on the pages of magnesium, iron, zinc and other substances; the dose — mg per serving against the intake; the purpose — is everything it needs in the formula; the excess — fillers, debatable additives, exceeded levels. For a cream: the evidence for the actives, the percentage or the place in the list, conflicting pairs, irritants. If your answer does not match ours, write to us: it is either a data error or a reason to refine the formula, and we will publish the result.

When the method changes

We revise the formula when new studies of forms, doses or pairings come out; the date this page was updated is the date of the current edition. In the edition of 1 September 2026 we removed from the score everything that was not about the ingredients and gave each category its own four questions. In the edition of 2 September 2026 every answer gained a confidence, an unprinted percentage became an interval from the ingredient order, a supplement's form moved into "how much will absorb" (it used to count twice), risks multiply instead of subtracting, actives of one front within a purpose became alternatives, the cleansing base and the filters are scored by irritation indices and spectra, and all the method's service numbers were moved into open files with sources. In the edition of 3 September 2026 we separated three different meanings that used to hide inside one "upper bound": a legal limit (a lawful product cannot exceed it — so it simply narrows our estimate of the concentration), a tolerance threshold (above it irritation is more common — a question of tolerance, not of strength) and a plateau (above it no extra benefit is expected — and that is no reason to cut the score). At the same time the probability of a reaction to a debatable ingredient stopped being a three-step expert ladder and became the measured share of positive skin tests — it turned out that parabens and benzyl alcohol are gentler than commonly believed, while isothiazolinones and formaldehyde releasers are markedly harsher. In the same edition the answers began to be multiplied instead of averaged: a failure on one of the four questions no longer hides behind three good ones. Old scores are recalculated, the change history is kept. We never change the formula for a particular product.

What the score cannot do

The score describes the ingredients by the label, not how the product will act on you, and it does not replace a doctor. We do not score medicines — they have a leaflet and a prescription. The service is informational: make any decision about taking or applying a product with a doctor, especially with chronic conditions, in pregnancy and when taking other medicines.

Frequently asked questions

  • Only by improving the formula: choosing a form that absorbs, putting in a working dose, adding the substances the purpose lacks, removing debatable additives. Prices, reviews, advertising and payment do not affect the score, and we have no paid placement. If a maker believes the data on its product are wrong, it can send the label — we will recheck and publish the result, whatever it is.

  • Because they are not about the ingredients. On marketplaces the rating and the review count are aggregated per brand: different products from one seller show the same "4.9" and the same texts. We use reviews differently — we went through thousands of questions and complaints so that the card answers what people actually care about: does it suit me, when to expect an effect, is it normal that it stings.

  • That the ingredients are known only from the seller — no label photo, no register record, no maker's page. Then we give no score: it is honest to calculate one only from data that can be checked. Once a label appears, we recalculate.

  • Because the score answers four questions, not one. Magnesium's effect is proven — 1.0. But even from an excellent chelate the body will not absorb all of 200 mg, and that is less than what the daily intake would deliver — 0.73. For "sleep and stress" magnesium lacks vitamin B6 — 0.71. No excess and no debatable additives — 1.0. The average is 8.6. That is fair: a very good product, but not a perfect one.

  • How much the answer rests on exact data. A printed percentage, a reference intake, an absorption curve — confidence 100 %. No printed percentage, so we estimated it from the active's place in the list — lower, because the interval is wide. A supplement form without comparative studies — lower too. It works inside the score as a weight: answers with less confidence weigh less, and if a question cannot be answered at all, no score is shown — we say which data are missing. We do not show it as a separate number on the card — it describes the label's completeness, not the product's quality.

  • We look at the formula not as "what is in the jar" but as "what the body needs for the purpose". Iron deficiency needs enough iron plus vitamin C; bones need calcium, vitamin D, magnesium and K2; sleep and stress need magnesium and B6. A product covers part of that set — that is its completeness. Plain magnesium gives 71 % for "sleep and stress", magnesium with B6 — 100 %. So the same honest magnesium gets a different score for different purposes — and that is true: for bones it is not enough on its own.

  • By the map of links between substances and purposes: the product's actives "lead" to the purposes they are needed for, and we take those they lead to most often. Iron leads to "iron deficiency" and through it to "hair and nails"; magnesium to "sleep and stress" and "bones". If the label names a purpose, we use it.

  • With the physiological requirement norms of Rospotrebnadzor (MR 2.3.1.0253-21): for adults magnesium 420 mg, zinc 12 mg, iron 10/18 mg, vitamin D 15 mcg, omega-3 1 g — and with the EFSA safe upper levels. For substances without a norm — creatine, collagen, inositol — we take the working range from clinical studies. All figures with references are on every substance page, for example vitamin D.

  • A boundary in the ingredient list. Ingredients at 1 % and above go in descending order; everything below in any order. Preservatives, fragrance and thickeners are almost always below 1 %, so the first of them in the list is the boundary. An active before it is at a noticeable share, after it — "for the label". For substances that work at small shares (ceramides, retinoids, piroctone olamine) a place after the line is not a minus.

For informational purposes only

This article is for informational purposes only and does not constitute medical advice, diagnosis, or treatment. Please consult a healthcare professional for medical guidance.

Is this supplement right for you?The label tells you whether a product is honest. Whether you need it, your blood tests tell: upload the form, answer a few questions and get a straight answer.
Supplement matchSkin care matchHair care matchCosmetic ingredient review
Still have questions about your health?Ask the AI assistant in plain words — about symptoms, how you feel, sleep, or what a value means. No files needed, first question free. Ask AI about health