A skincare system built around your skin, not around a catalogue.

01 — Overview

What this is

A system for looking after your skin: deciding what to use, checking it’s safe for you, using it correctly, tracking whether it worked, and knowing when the answer isn’t a product at all.

Most skincare sites are shops with articles attached. The question they answer is “what’s popular?” This one is built to answer a harder and more useful question: “what is likely to work on my skin, and how would I know?” Those are different questions, and answering the second one requires machinery the first one never needs.

That machinery is four layers, and each is useful on its own:

They compound. The engine makes the content trustworthy; trustworthy content makes personalization meaningful; personalization makes the tools worth using; the tools build a history of your skin; and that history makes every future recommendation sharper. The system gets more useful the longer you use it, which is the opposite of how a product catalogue behaves.

02 — Why it needs to exist

The problem

Skincare is a category where the buyer is structurally outmatched. Five specific ways:

1. The label is not the product

   “Brightening” is at least five different clinical claims wearing one word — post-inflammatory hyperpigmentation, melasma, sun spots, general photodamage, or nothing but optical shine. They have different evidence bases and respond to different ingredients. Meanwhile “2.5% retinol” is often retinyl palmitate, and “botanical hyaluronic acid” is frequently a plant gum. None of this is visible from the front of the box, and much of it isn’t obvious from the back either.

2. Ratings measure popularity, not suitability

   A 4.7-star average tells you a product pleased people. It cannot tell you it’s a poor fit for rosacea, that its key active is under-dosed, or that a large share of those reviews came from a seeded promotion. When we began extracting review data properly, we found products where 36–50% of reviews were undisclosed promotional placements. Those reviews are invisible in the average everyone shops by.

3. Advice is generic; skin is not

   A rich ceramide balm is a gift for eczema and a liability for fungal-acne-prone skin. A 10% niacinamide serum is unremarkable on resilient skin and a flushing risk on reactive skin. A single score has to average that difference away — and that difference is the entire clinical reality of the product.

4. Nobody remembers anything

   Which product caused that reaction in March? When did the flare actually start? What did the last dermatologist change? This is precisely the information a clinician asks for and precisely the information that lives, unreliably, in memory and a camera roll.

5. And products have a ceiling

   Some conditions do not respond to anything sold without a prescription. Almost no commercial skincare site will say so plainly, because saying so ends the sale. That silence is why people spend years and hundreds of pounds cycling through products for something that needed a prescription in month two.

03 — The engine

How products are assessed

Most rating systems pick a number and construct a justification afterwards. This one runs the other way: it starts with evidence, reasoning and provenance, and derives the number at the end. That inversion is the whole design.

One verdict per concern, read off a published rule tree

For every concern a product claims to help, the engine answers three questions — does it contain an active with documented standing for this concern; is it present in a form that can work; is it likely to be tolerated for this concern — and a numbered rule maps the three answers to one of a handful of plain words: Strong choice, Good choice, Supporting role, Promising, Unproven, Not for this, Not yet assessed, Avoid (sunscreens read Full or Basic protection). No weighted number sits behind the word; every verdict names the rule that produced it, and every rule is published on the methodology page with its reasoning. Beside the verdict, never blended into it, sit the evidence strength (four dots, derived from the row’s trial record) and the collected reviews with their count. A verdict is our assessment of how well a product’s ingredients fit a concern — never a claim about what the product treats.

The rules the engine cannot break

• One scoring owner per fact. Fragrance is priced in safety, texture in usability, price in value. Nothing scores twice under two names — and no factor is computed from another. Value is never derived from popularity, which would let a fashionable product inflate its own price justification.
• Risk, not hazard. An ingredient is judged at the concentration it appears in, in the format it appears in. The same fragrance is a different proposition in a 30-second cleanser and an eight-hour night cream. Assessment that ignores dose and exposure is assessing vocabulary, not risk.
• No weakest link. One alarming-sounding ingredient never caps a product; ingredients aggregate in proportion to how much of the formula they actually represent.
• Claims are a closed vocabulary. Nothing is graded against a phrase taken from a label. Post-inflammatory hyperpigmentation is not melasma — PIH resolves, melasma relapses and defeats treatments that clear PIH easily — so a trial measuring PIH is evidence for PIH and is never transferred to melasma. That one rule invalidates a large share of the industry’s evidence borrowing.
• Unknown is a real value. It scores neutral and lowers stated confidence. Never rounded up (which would credit marketing), never rounded down (which would punish proprietary formulation). Uncertainty is absorbed by confidence, never by the score.
• Transparency can only help. A brand that publishes its concentrations gets measured precisely instead of estimated. The reverse is impossible by construction: nothing is penalised for non-disclosure, because a system that punishes secrecy rewards whoever hides the most.
• Same product, same data, same result. Every input comes from a reference table, a public record, an ingredient list or a collected review — never a fresh judgement made at scoring time. Two people running the engine on different days get the same number, or the number is an opinion wearing a lab coat.
• Judgement happens once, in the open. There is exactly one place human expertise enters: deciding what the published evidence says about an ingredient for a claim. That lives in a versioned table where every row carries its sources, its reviewer and its review date.

Hard gates: the disqualifiers

Weights price ordinary differences. Six gates handle things that should disqualify a product regardless of how well it scores: an active recall or verified contamination finding; a false clinical claim; documented review manipulation; a category integrity failure (a “sunscreen” without adequate UVA filtration, derived from its own ingredient list); a formulation that neutralises its own hero ingredient; and an expert consensus that it does nothing.

Every gate has published entry criteria, exit criteria, a review schedule, a disputed state for findings under formal challenge, and a written appeal path. A gate can only be entered through its published criteria, never by editorial judgement. That is the difference between a standard and a grudge.

Confidence is stated, not buried

Every assessment carries a confidence band alongside it, and the two are shown together. A product with thin data can still earn a good grade — it simply says so, and tells you what the grade currently rests on. Hiding uncertainty inside a confident-looking number is the most common failure in this category, and it’s designed out here.

04 — The evidence

The research behind it

The engine is only as good as what it reads. The research layer is the asset the whole system rests on — and it’s published in full, so any number on the site can be followed back to the study it came from.

Where the evidence comes from

Ingredient and claim research is built from the primary literature — controlled trials, systematic reviews and meta-analyses indexed on PubMed and PubMed Central — read at the level of the individual paper rather than summarised from secondary sources. Building the current tables involved working through well over a thousand papers, and the count rises with every research batch. Today the system carries 1,321 distinct verified studies cited across product reviews, ingredient references, therapy guides and articles, and more than 5,263 individual citations placed where they’re relevant. Both numbers grow continuously; the research pipeline runs in parallel with everything else.

How a claim becomes a score

1. Pair the ingredient with a specific claim. Not “niacinamide is good” — “niacinamide for post-inflammatory hyperpigmentation”, as a distinct row from “niacinamide for barrier support”.
2. Find what has actually been measured. Each claim publishes the endpoints an admissible study must have measured, so evidence can’t drift between similar-sounding conditions.
3. Read the primary sources. Trial design, population, dose, duration and endpoint are recorded — a positive result at 4% for 12 weeks is not evidence for a product containing 0.5%.
4. Derive the tier. The evidence tier is computed from what was found, never typed in by hand — so two researchers reviewing the same body of work land in the same place.
5. Record provenance. Every row carries its sources, the reviewer’s name and the review date, and is versioned. The current table holds 228 reviewed rows and is actively expanding — each new row sharpens every grade that depends on it, across the whole catalogue, at once.

The right study, cited precisely

Selecting the citation is part of the research, not a footnote added afterwards. For each claim we identify the studies that actually bear on it — the right population, the right endpoint, the right dose and duration — and prefer systematic reviews and controlled trials over single small studies or secondary summaries. A well-designed trial in the wrong population is not evidence for the claim in front of us, and it isn’t cited as though it were.

Every citation is then confirmed against the source database before it publishes, so the reference you see resolves to exactly the paper it names — title, journal and year matching the record. That check runs automatically across products, ingredient references, treatment guides and articles, so the link under a claim is always the study behind it.

All of it is open on the page

This is the part most systems don’t offer: the evidence isn’t summarised for you and locked away. On any product page, the assessment expands.

• The headline grade opens into the chapters and factors that produced it.
• Each factor opens into the individual measured facts — which ingredient, at what dose, judged against which claim.
• Each of those opens into the studies themselves, with title, journal, year and a link to the paper.
• The provenance strip states where the grade’s data came from and how much of it there is.

It’s all loaded on demand, so the page stays fast for people who just want the answer — and goes all the way down for people who want to argue with it. The system is designed to be argued with, which is a deliberately different posture from a star rating.

Real-world outcomes, read properly

Alongside the literature, review text is collected from retailer and platform APIs and passed through a fixed extraction model — measured against human coding at κ = 0.851, a level of agreement generally considered strong. It reads what people actually said about their concern, rather than counting stars, and it flags undisclosed promotional reviews. Collection expands with every pass, and because it’s expensive to gather it’s treated as a durable asset: backed up separately, refreshed on a cadence, and never confused with the derived scores that are rebuilt from scratch each time.

05 — Personalization

The site rearranges itself around your skin

You describe your skin once. Everything after that is filtered through it — with the reason shown in words, every time.

The profile takes about a minute: skin type, up to three concerns you most want solved, how reactive your skin is, whether you’re pregnant or nursing — and then the refinements that sharpen the matching: your age band, your Fitzpatrick sun-response type (which drives white-cast and visible-light guidance for deeper skin tones), your experience with actives, and any ingredient families you never want to see, fragrance included. Every question is skippable; every answer makes the matching sharper. From that point:

• Product pages give you a personal verdict. Not a star average — “Graded A− for your rosacea”, or “You listed eczema; this product is flagged avoid for it”, with the reasoning written out.
• Every product card, everywhere, carries your grades. In listings, curated rankings, brand pages, routines and treatment pages, cards show up to two letter badges for your concerns. Safety-critical grades always surface, so a poor fit can’t hide behind a good one.
• Listings reorder around your profile instead of a generic ranking.
• My Skin becomes your skin dashboard. One page assembled entirely for you: top-graded picks for each concern, the therapy guide each one connects to, your routines, your shelf, and reading chosen for your skin — each card labelled with why it’s there. The homepage picks up the same dashboard.
• Tools prefill. Skin type, phototype, sensitivity and the products on your shelf flow between tools, so nothing is asked twice.
• Your exclusions are hard filters, not preferences. An ingredient family you excluded is checked against each product’s actual ingredient list — not a marketing claim on the front of the box — and the product is filtered or flagged accordingly, whatever grade it earned.
• Your reaction history overrides everything. A product you logged a reaction to in the Patch Test tool is flagged as yours to avoid — above any grade the engine gave it.

It sharpens the longer you use it

The profile is the start, not the whole picture. Every tool you touch quietly completes it: the products you add to your shelf are what the Conflict Checker audits and the Routine Order tool sequences; a failed patch test writes a permanent avoid flag; the Skin Type Quiz and Sun Safety Guide replace your guesses with measured answers that every other tool then inherits. Diaries, severity scores and progress photos build the record that Derm Visit Prep turns into a one-page appointment brief. Nothing is ever asked twice — and the recommendations get harder to distinguish from ones a person who knew you would make.

It also knows what it shouldn’t recommend. Something you reacted to stays flagged everywhere, forever, however well it grades for everyone else — and your shelf isn’t just stored, it’s audited: what clashes, what’s expired, what order it should go on in.

Try it as someone else first — the Test Drive

The honest problem with a system like this is that on day one it’s empty: the trend lines, the photo comparisons, the skin dashboard that remembers what you’ve tried — all of it takes weeks of real use to appear. So we built a way to skip ahead. The Test Drive loads a fully lived-in sample person into your browser and lets you walk the entire site as them. Using it takes three steps:

1. Open the Test Drive page and pick a person. There are six, each built to show you something different: six months of eczema tracking through a real relapse, the week-three retinoid dip that makes people quit too early, a slow and honest melasma course with a summer rebound, a shelf full of clashes, expired products and cheaper twins.
2. Browse anywhere — the whole site is theirs. Product cards carry their grades, their skin dashboard is filled in, every tracker has months of entries, the photo tracker compares their timeline side by side, and Derm Visit Prep assembles their real brief. Nothing is walled off; whatever you’re curious about, open it.
3. Switch or leave whenever you like. A banner stays pinned to the top of every page with two doors: Start your own, which sets the demo aside and begins your real profile, and End test drive, which returns you to your own data and the scenario picker. Simply closing the tab does the same — every trace of the demo is gone.

Nothing you do in a test drive is saved, and your own data — if you have any — is never touched or shown. The progress photos are AI-generated illustrations, not real people or real results.

06 — The toolkit

42 tools, in six families

Every tool is free, needs no account, works instantly, and saves nothing to a server unless you ask it to. Each is useful on its own — and each one quietly makes the others smarter, because they share one profile and one library. More are in development.

Match — what should I use?

Check — is this safe for me?

Plan — how do I use it properly?

Sun — the highest-stakes category

Score — how severe is it, in clinical terms?

The newest family, and the most clinical: the severity instruments dermatology actually uses — the same indices that appear in trials and clinic notes — rebuilt for careful self-assessment. Each explains what it measures and where self-scoring has limits, keeps a dated history so treatment changes can be judged against a number rather than a feeling, and feeds straight into Derm Visit Prep. Arriving with a scored history is the fastest way to a productive appointment.

Track — is any of this actually working?

This family is what makes the site a management system rather than a shop: the running record — extent, symptoms, photographs — and the brief that carries all of it into an appointment.

07 — The honest ceiling

When products aren’t enough

Over-the-counter skincare has a limit. For several conditions that limit arrives early, and the responsible thing a skincare site can do is say so, then explain exactly what lies beyond it. That’s what the therapy guides are.

These are not blog posts about prescriptions. Each guide is a structured, clinically grounded map of a condition — built from the current recommendations of the professional bodies that write them, and published as education rather than as advice to any reader.

What’s inside a guide

Every guide is assembled from the same twelve parts, because the questions people arrive with are always the same twelve questions.

1. Triage first. About six questions, and the page rearranges around your answers — which branch applies to you, which checkpoint you’re standing at, and what would have to change for the next step to be right. Five presentations may be mapped on a page; only one is probably yours.
2. Staging, as a way to describe a presentation. Clinicians don’t grade by a single number — they grade by which lesions are present, how widespread they are, and what the condition is leaving behind. The staging framework is drawn from published clinical staging, and it exists so that a presentation can be described accurately, not assessed at home.
3. Red flags, with urgency. The specific findings that mean skip the ladder and see someone — deep nodules, any scarring already forming, rapid change — each with the reason it matters and how quickly to act.
4. The over-the-counter ceiling, stated honestly. What OTC care genuinely achieves for this condition, for whom, and by when — including the realistic timeframe, which is almost always months rather than weeks.
5. The OTC pathway itself. Where products do work, the route is laid out as a ramp rather than a series of experiments: one regimen built up gradually as tolerance allows, not a carousel of products swapped every three weeks.
6. Branches and ladders. Different presentations of the same condition need different sequences. Each branch is a numbered ladder — rung by rung, OTC through prescription — with the rationale for each step and what has to be true before you climb.
7. Therapy cards. Every option, OTC and prescription alike, gets the same card: what it is, what the evidence says, its tier, how it is used and what to expect. Currently 163 therapies are documented across the guides, spanning over-the-counter actives, prescription topicals, oral medications, procedures and devices.
8. Supporting care — the part nobody tells you. Prescriptions fail for the same reason routines fail: people stop. Usually because the skin around the treatment got too dry, too raw or too sore to continue. Each guide covers the supporting routine that keeps a course survivable, which is often the difference between a therapy working and a therapy being abandoned.
9. The visit. How to actually get seen (including when primary care is a legitimate first step), what to bring, and what will likely be asked — paired with the Derm Visit Prep tool so you arrive with it already assembled.
10. Expectations, in weeks. What each treatment realistically takes to work, and what the adjustment period looks like — including the ones where skin looks worse before it looks better. Knowing that in advance is frequently what determines whether someone finishes the course.
11. The pipeline. What’s genuinely coming — restricted to treatments in phase 3 trials or later, or newly reaching market. Earlier-stage research is deliberately excluded, because early promise often doesn’t survive larger trials and reporting it as if it were imminent is exactly the hype this site exists to counter. Nothing here changes what to do now, and the page says so.
12. The guideline anchors. Each guide names the specific professional guidelines it is built on, with citations and links to the papers.

Built on the guidelines clinicians use

These guides aren’t assembled from our own opinion of the literature. They’re anchored to the current recommendations of the bodies that set practice, and updated as those bodies publish. They currently draw on guidance from:

• American Academy of Dermatology
• American Academy of Dermatology / National Psoriasis Foundation joint guidelines
• British Association of Dermatologists
• National Rosacea Society Expert Committee
• Global ROSacea COnsensus (ROSCO) panel
• EADV Mycology Task Force position statements
• North American Contact Dermatitis Group
• Cochrane systematic reviews
• US FDA over-the-counter drug monographs
• International expert consensus on pigmentation disorders (JEADV)

Where a condition has no formal guideline — which is genuinely the case for several pigmentation and rosacea sub-questions — the guide says so rather than implying a consensus that doesn’t exist, and falls back to systematic reviews with their limitations stated.

Where clinical review currently stands

Being straight about this matters more than the claim it replaces. There is a hard publication gate in the build — a guide marked as a draft cannot reach the site, the sitemap or a deploy, no matter what else happens — and every guide carries a dated changelog of what changed and when. That gate is editorial, not clinical. It records that a guide is finished and checked against its sources; it does not record a dermatologist’s signature.

The published guides — Acne, Eczema, Melasma, Rosacea, Fungal acne, Contact dermatitis, Hyperpigmentation, Dandruff, Nail fungus, Facial seborrheic dermatitis, Keratosis pilaris, Perioral dermatitis, Ringworm, jock itch & athlete's foot, Tinea versicolor, Warts — are written from the professional guidelines named above and the studies cited on each page, and are with a consulting dermatologist for review. None of them currently carries an individual clinician’s sign-off, and none claims one. When a guide has been reviewed, the reviewer is named on it by name and credential, with the date; until then, what stands behind it is the published guidance it cites and nothing more. Stating that plainly is the only version of this section worth publishing.

Where the catalogue stops

One rule governs the commercial boundary: prescription and procedure cards never carry product links. The catalogue stops at the over-the-counter ceiling. Where products genuinely help alongside a prescription — a bland moisturiser during a retinoid course, a mineral sunscreen during melasma treatment — they appear as supporting care, in their real supporting role, not as an upsell dressed as treatment.

The site that tells you when to stop buying is the one worth believing when it tells you what to buy.

08 — Continuity

Your library, and why it has to survive

A management system that forgets is not a management system. Everything you build here is inventoried in one place, portable, and yours to take or destroy.

My Library

One page lists everything the system holds for you — skin profile, product shelf, wishlist, comparisons, photo log, flare diary, POEM/itch/UAS7 scores, patch tests, mole checks, cycling and ramp schedules, phototype, appointment briefs: around three dozen distinct records, growing as tools are added. Each row links to the tool that owns it, shows whether it’s backed up, and can be cleared on its own. The whole library exports to a single file and re-imports safely.

It answers “what do you have on me?” completely, in one place, with a delete button beside each answer.

The account — memory, not a wall

There is no signup wall, on anything, ever. Every tool works anonymously and completely. An account exists for one reason: browsers delete data. Safari erases a site’s stored data after seven days without a visit — which would silently destroy someone’s shelf, profile and three months of flare tracking. For a system built on continuity, that’s fatal, so the account solves it:

• Sign up with a name (a pseudonym is fine), an email and a password. Everything you created before signing up is merged in — nothing you did as a visitor is lost.
• Profile, shelf, diaries, scores and schedules follow you to any device and any browser you sign into.
• Progress photos back up too, with generous per-user storage, and deleting one on any device deletes it everywhere rather than letting it reappear.
• A persistent chip tells you the truth continuously — “saved on this device only”, “backed up to your account”, or a prominent warning if your session lapses. Silent data loss is designed out.
• Conversion is never a modal or a popup. It’s an inline note at a moment when it’s genuinely relevant — a second diary entry, a completed profile — naming your actual data (“your 12-product shelf”), capped to one at a time and silenced for a fortnight if you dismiss it.

Underneath, the merge logic is careful in a way that matters: diaries and logs merge entry by entry rather than whole-file, so a phone that’s been offline for a week can’t overwrite entries made on a laptop, and deleted entries stay deleted instead of resurrecting from an old device. Full data export and complete account deletion are first-class functions, not buried requests.

09 — Scale

The content estate

Everything above runs on a large and deliberately disciplined body of content. Every number here grows continuously — the editorial and research pipelines run constantly.

10 — Principles

What we refuse to do

Constraints enforced in code and process, not promised in a paragraph.

• No paid placement, no sponsored seals, no house brand, and no mechanism by which a brand could influence a score. Revenue is affiliate-based and invariant to what any score says.
• No score computed from popularity.
• No citation without the study behind it — every reference resolves to the paper it names, confirmed against the source database.
• No treatment claims. Grades describe suitability — “A− for rosacea-prone skin” — never “treats rosacea”.
• No brand politics scored as product quality. Cruelty-free status and parent-company ownership are displayed as facts; they never move a number.
• No personal data sent to a server in order to personalize anything.
• No signup wall on any tool.
• No modals, no popups, no interstitials, no countdown timers.
• No medical advice, anywhere. Clinical tools carry their educational framing and point toward a professional, and nothing on the site assesses anyone’s skin or recommends a therapy to anyone.
• No pretending the catalogue is the answer to every problem.

11 — The experience

How it’s built to feel

A system you’re meant to use every day has to feel effortless, so the whole thing is built around two promises: never make you wait, and never get in your way.

• Made for your phone first. Most skin care happens at the bathroom sink and in the shop aisle, not at a desk — so every page and every tool is designed for the phone in your hand, and works just as well on a laptop.
• Fast everywhere. Pages open near-instantly, even on slow mobile data, whatever the size of the catalogue behind them.
• Nothing to install, nothing in your way. It runs in any browser — no app, no signup wall, no popups, no “please rate us”. You open a page, and the page answers.
• Dependable. The site stays up, links keep working, and the library you build here survives every update — it’s treated as yours, not as ours.

12 — Direction

Where it goes next

The system is built to widen rather than to be finished. Everything below is in progress, not aspiration.

• More therapy guides. Additional conditions are drafted and moving toward clinical review. The rate limit is dermatologist review capacity, deliberately.
• A deeper evidence table. Every new ingredient-claim row sharpens every grade that depends on it, retroactively, across the whole catalogue. This is the highest-leverage work in the system and it runs continuously.
• Broader outcomes collection. More products with real extracted review data rather than category priors.
• More tools. The toolkit is designed so a new tool inherits the profile, the library, the storage guarantees and the sync layer for free.
• Continued catalogue growth, with new products assessed by the same engine on arrival — no backlog of un-scored items.