Framing
Facts, opinion & right of reply
Two different kinds of statement appear on every product page, and we label them differently on purpose.
FactsIngredient lists, SPF and PA labelling, sizes, prices and availability. These are sourced, dated and correctable. If one is wrong, we fix it.
OpinionVerdicts, scores, confidence tiers, skin-type fit and value calls are our honest editorial opinion, reached by applying the criteria set out on this page to those facts. They are not statements of fact about a brand, not laboratory findings, and not medical advice. Reasonable people reading the same INCI can land somewhere else.
Right of replyAny brand can respond to an opinion about its product. Send a statement via brands: correct our data and we publish it beside the rating, attributed and unedited except for length and legality. While a substantive dispute is open, the disputed line is marked as such.
No brand pays for placement, a score, or the removal of one. Correcting a factual error never improves a verdict on its own — it just re-runs the same published criteria over better inputs.
Trust
Confidence tiers
Every score you see is tagged with one of these badges.
Community-verified
When≥3 real users have submitted a micro-review for this product on this axis.
StrengthHighest-trust signal on the app. Community votes fully override the AI formula estimate.
CaveatEarly samples can skew. We show sample size + agreement so you can judge how settled the verdict is.
Web evidence
WhenOur review-mining bot found mentions in blogs, forums, or retailer reviews.
StrengthReal-world signal from many voices — often more informative than a single opinion.
CaveatWeb-scraped text can be marketing spin, sponsored, or about a different SKU. We filter, but noise remains.
Formula estimate
WhenNo user reports and no web evidence — the score is inferred from the ingredient list and label claims.
StrengthA useful heads-up, and near-deterministic where INCI decides the answer — UV filters, barrier lipids, which actives a serum contains.
CaveatSensory axes (stickiness, whitecast, sting) are guesses until real people weigh in. Never treat as verdict.
Brand claim only
WhenWe have no ingredient list for this product at all — the score is inferred from the brand's own description, category and format.
StrengthWeak. It tells you what the product is positioned to do, not what's actually in it.
CaveatNothing here is checked against a formula. Add a label photo or a brand source URL and these scores get re-derived properly.
Not yet rated
WhenThe ingredient list hasn't been scanned yet AND no reports exist.
StrengthNone. We show this openly rather than fake a number.
CaveatRescan usually resolves within a day.
Weight
Sample size & agreement
A score built on 1 report is not the same as one built on 40. We publish both numbers inline — the raw sample size and how much people agreed.
Strong · 80%+Most people saw the same thing. Verdict is reliable.
Mixed · 50–74%Majority view holds but a real minority disagrees.
Split · <50%No consensus yet. Read the individual reports before deciding.
1 reportDirectional only. We don't hide it, but we don't dress it up either.
Fairness
Skin-tone floor
Across every category, whitecast is the only axis where skin tone materially changes the verdict. We only publish a per-tone verdict once we have at least 5 reports for a given Fitzpatrick bucket (light I–II, medium III–IV, deep V–VI). Below that the cell shows a "help us rate" prompt instead of a coin-flip verdict. We ask everyone for their Fitzpatrick type (1–6, not just light/medium/deep) so buckets don't quietly collide.
Sourcing
Per-rating sourcing
How each individual axis is sourced today, grouped by what it applies to.
All categories
Ingredient list (INCI)
DeterministicParsed from the pack or brand page and matched against our ingredient dictionary. Everything below depends on it, so a product with no verified INCI shows 'unknown' rather than a score.
Fragrance-free
DeterministicWhole-word match over the INCI list, including botanical fragrance sources. Deterministic once ingredients are attached.
Alcohol / drying content
DeterministicPosition of denatured alcohol in the list drives the skin-type fit call. Fatty alcohols are explicitly excluded.
Comedogenic / breakout risk
AI-estimatedRough estimate from INCI, weighted by how high the culprit sits in the list. Skin varies wildly — treat as directional.
Pregnancy advisory
DeterministicFlagged from ingredient class (retinoids, high-dose salicylic acid). An advisory to raise with your clinician, never a safety verdict.
Grounded prices & price per use
Web-scrapedCross-checked across the brand site and major retailers. Bundles, multi-packs and marketplace resellers are excluded.
Regional availability
Web-scrapedRetailer scraping across US/EU/UK/KR/JP/AU. Marketplace hits need brand plus local currency or script to count.
Reformulation tracking
Web-scrapedWe re-read the INCI periodically and flag the product when the list changes materially since our last check.
Texture, stickiness, greasiness, finish
Community-firstAI-estimated from the formula until ≥3 community ratings arrive on that axis, then the community verdict overrides the estimate.
Derm-tested
Web-scrapedVerified via brand pages and derm-lab sites when available; absent evidence is shown as absent.
Sunscreens
SPF / UVA filter system
DeterministicRead from the filter list and label rating. Filter quality, UVA coverage and photostability pairings are rule-based, not opinion.
Water / sweat resistance
DeterministicTaken from the label claim and the minutes tested. Marked 'unverified' when no claim can be sourced.
Whitecast by skin tone
Community-firstPer-tone verdict only publishes once ≥5 reports exist for that Fitzpatrick bucket — below that we ask for more ratings instead of guessing.
Eye sting
Community-firstEven one 'a little' user report overrides the formula estimate on this axis.
Reef-safe
DeterministicFilters checked against the Hawaii and Palau ban lists.
Moisturisers
Barrier & humectant load
DeterministicCeramides, cholesterol, fatty acids, glycerin and other humectants are identified by INCI and weighted by list position.
Occlusivity / richness
AI-estimatedEstimated from the occlusive and emollient stack, then sanity-checked against how users describe the wear.
Pilling under makeup or SPF
Community-firstEstimated from known film-formers (silicone/gum combinations) until real reports arrive, which then override.
Hydration claims
Web-scrapedOnly shown when the brand publishes a corneometry or TEWL study we can cite; otherwise the claim is marked undisclosed.
Serums
Active identification
DeterministicEvery active is matched to a known family — retinoids, vitamin C forms, acids, niacinamide, peptides — so derivatives aren't treated as the parent molecule.
Concentration / dose
Web-scrapedBrand-disclosed percentages first. Where none exist we publish a bound from list position, labelled 'dose not disclosed' rather than a guessed number.
Third-party lab assays
Web-scrapedIndependent lab breakdowns are ingested where published and always outrank our own estimate.
Clinical evidence for the active
DeterministicStudy ranges come from published literature on the ingredient, not from the brand's marketing copy.
Layering & irritation risk
AI-estimatedRule-based pairings (retinoid plus acid, vitamin C plus niacinamide) plus known irritants, then adjusted by community sting reports.
Limits
Openly disclosed blindspots
Things we can't guarantee today — better to say so than to fake it.
In-vivo SPF verification
We trust the number on the label. Independent SPF re-testing is outside our scope today, as is measuring how filters degrade under UV over several hours.
Undisclosed active concentrations
Most serum brands never publish a percentage. Where none exists we publish a bound from list position and label it undisclosed — we will not invent a number.
Long-term efficacy
We can tell you whether an active is dosed within its studied range. We cannot tell you it will work on your skin over six months — no ingredient list can.
Batch reformulations
Brands quietly tweak INCI. We re-scan periodically but a fresh batch may differ from what's on file.
Regional formula variants
US/EU/KR versions of the 'same' product can have very different filters. We flag known variants but can miss silent regional swaps.
Comedogenicity & irritation
Breakouts and stinging are highly personal. Our estimates are directional, not a promise, and never a substitute for patch-testing.
Manufacturing quality
Stability, pH and how well a formula survives its shelf life are invisible from an ingredient list unless the brand or a lab publishes data.
Accountability
Corrections & how to reach us
We get things wrong. When we do, we want to hear it — from anyone.
Submit a correction — readers, labs and brands. Send the sentence that's wrong and a source; we review every one and reply with the outcome, including when we decide not to change anything.
Brands: correct our data or reply on the record — a direct line for brand, PR and legal teams, including trademark and disputed-claim concerns and the right of reply described above. Free, and never tied to a score.
Corrected pages state what changed and when. Email corrections@skinparency.com if you'd rather not use a form.
Early access
Help us get sharper
The fastest way to move a score from formula estimate to community-verified is a micro-review after you've tried a product. Join the early-access list to start rating.
Free to join · no spam · one launch email