How We Evaluate Every Product

Every product on DermApproved goes through the same evaluation — whether it's a $9 drugstore cleanser or a $300 luxury serum. For each concern a product claims, we answer three questions from its ingredient list and a versioned clinical-evidence table, and read a verdict off a published rule register — words, never a blended number. A deterministic scoring engine sits underneath, and a detailed editorial review grounded in published research sits beside it. Then we turn all of that into recommendations tuned to your skin.

This is the technical companion to How It All Works: that page tells the story of the whole system; this one shows the work — how a verdict is decided, where the evidence behind it comes from, what the engine underneath computes, and how confident we are in each reading — so you can decide for yourself how much weight to give our assessments.

How a Verdict Is Decided

For every product and every concern it claims to help, we answer three questions and read the verdict off a published rule tree — no weighted number, no blend. 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? Every verdict on the site names the rule that produced it, and every rule is listed here with its reasoning. A verdict is our assessment of how well a product’s ingredients fit a concern; it is never a claim about what the product treats.

The rule register rules 0.5.1 · evidence table ev-2026.09-r228

Rules are evaluated in order; the first that matches decides. The register and its anchor set are editorial decisions, each with a written rationale, and stand pending review by a dermatology advisory panel as a method — no advisor is, or will be, named on a rule or a product.

Triggers that make a formula an avoid for a concern are listed per concern in the same register; cautions — a photosensitising active, a strong acid, a single declarable allergen — add a caution line to a verdict and never decide it. A Tier 1 or 2 seal from an expert body (the National Eczema Association, the Canadian Dermatology Association and their peers) outweighs a trigger for the concern the seal covers.

Two Scores, Built From Evidence

Beneath the verdicts, the engine still computes an Overall Score (0–100) and, for each concern a product addresses, a Condition Grade (A+ to F). Since the verdict cutover neither is shown on a product page or a card — a verdict says what a number cannot, and a number invites a ranking the evidence does not support — but they still drive safety gates, the tolerability question inside every verdict, and the category "job" reading, and both unfold into a citable audit trail. They are computed independently: the Overall answers "how good is this product?", the grade answers "how good is it for this specific job?"

The Overall Score — six factors

No factor sits below 10% — anything that can't earn a tenth of the score becomes displayed information instead of a decorative weight. And every fact has exactly one scoring owner: fragrance is priced in safety, texture in usability, price in value. Nothing scores twice under two names.

Where the Numbers Come From

Underneath the engine sits a versioned table of ingredient–claim pairs — currently 228 reviewed rows, expanding continuously. Each row is not "niacinamide is good"; it is "niacinamide for post-inflammatory hyperpigmentation", as a separate row from "niacinamide for barrier support", because those claims have different evidence bases. Each row records what the primary literature actually measured — trial design, population, dose, duration, endpoint — and its evidence tier is computed from those findings, never typed in by judgement. Every row carries its sources, its reviewer, and its review date.

Three rules keep the table honest. Claims never transfer: a trial measuring post-inflammatory hyperpigmentation is evidence for PIH, and is never borrowed for melasma — which relapses and defeats treatments that clear PIH easily. Dose matters: a positive result at 4% for 12 weeks is not evidence for a product containing 0.5%. Unknown is a real value: an unverifiable concentration scores neutral and lowers stated confidence — never rounded up (which would credit marketing) and never rounded down (which would punish proprietary formulations for not disclosing).

Every citation on the site is confirmed against the source database before it publishes — the title, journal, and year of the paper at the link must match the reference that cites it. A plausible-looking link to the wrong study is worse than no link, so it's checked by machine, across every product review, ingredient reference, therapy guide, and article.

Reading Reviews Instead of Counting Stars

A 4.7-star average tells you a product pleased people; it can't tell you it failed everyone with rosacea. So the outcomes factor is built from review text, 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 with a specific concern actually reported, feeds the condition grades with concern-matched outcomes, and flags undisclosed promotional reviews — on some products we've found 36–50% of reviews were seeded placements, invisible in the star average everyone shops by.

Where reviews haven't been collected for a product yet, the grade still exists — it runs on formulation and clinical evidence, and is labelled accordingly rather than padded with a guess.

What No Score Can Outweigh

Weights price ordinary differences. Six gates handle the things that should disqualify a product no matter how well it scores. A gate can only be entered through its published criteria — never by editorial judgement — and every gate has exit criteria, a review schedule, a disputed state for findings under formal challenge, and a written appeal path.

How We Read the Evidence for an Ingredient

Every hero ingredient is graded against a published evidence table, one row per ingredient and concern — azelaic acid for rosacea is a different row from azelaic acid for acne. Each row answers two separate questions, and the product page shows both, because they are not the same claim.

How Confident We Are in Each Number

A score of 84 backed by 15 years on the market and 50,000 reviews is a different finding than an 84 on a three-month-old launch. So a confidence band sits beside the assessment on the product page, at equal visual rank — it is never absorbed into the number itself. Uncertainty lowers confidence, not the score.

The Same Grades Everywhere You Look

The scores aren't decoration on the product page — they are the ranking machinery for the whole site. Every "best for [concern]" list is ordered by the relevant condition grade with a hard floor at B−, so a list titled "best moisturizer for rosacea" is ranked by the rosacea grade, not by a general number that knows nothing about rosacea. And once you've set up a skin profile, product cards across the site carry letter badges for your concerns — computed by the same engine, from the same table, with the same audit trail behind them.

What's in Every Review

Beyond the numbers, every product gets an 800–1,500 word editorial review — a genuine analysis written by someone who understands formulation science, opening with a unique angle rather than the same template. Each one also carries the sections below.

Hype-Aware, Never Cynical

Not every product earns the attention it gets. We apply friendly skepticism where it's warranted — always assessing the formula on its merits, never punching down.

Where does all this data go once it's scored? The site personalizes itself around your skin — that story, from the skin profile to the full toolkit, is told in How It All Works.

We Show Our Work

Everything that goes into an assessment is on the page — and it goes all the way down. The headline grade opens into the factors that produced it; each factor opens into the individual measured facts (which ingredient, at what dose, judged against which claim); each fact opens into the studies themselves, with title, journal, year, and a link to the paper. If we're uncertain, the confidence band says so. You might weigh tolerability more heavily than we do, or care more about evidence strength than value — by exposing the breakdown, we let you apply your own priorities instead of just accepting ours.

And none of it is frozen. As formulas change, new research is published, or collected reviews reveal patterns we missed, the engine re-derives the numbers — every new row in the evidence table sharpens every grade that depends on it, across the whole catalogue at once. The goal isn't to be right the first time every time — it's to be honest, thorough, and continuously improving.