The study is real. Now read how it was built

Every product on this page tells you something about the study behind its headline number, and most of the shelf tells you nothing at all. That one fact changes the question worth asking. It stops being whether a test exists and becomes how the test was built.

Give the five brands below full credit before anything else. Naming a testing institution, printing a panel size, dating the study window, footnoting in small type that a figure came from a survey rather than from a machine: none of that is required of them, and every line of it hands a reader the material to argue with. A brand that publishes nothing cannot be argued with at all. That is the worse position for a shopper to be in, and it is the more common one.

So what follows is not a reason to avoid any of these products. It is what the disclosure makes possible. Four questions do almost all the work: how many people, over how long, compared with what, and measured by an instrument or by a questionnaire.

When 100 percent is a satisfaction answer

Realbarrier Ceramide Moisture Cleansing Foam

Four figures on this page read 100 percent: overall satisfaction, cleansing power, oil and water balance, and hydration. Beneath them, in small type, is a footnote most brands would not bother to print. It says these are questionnaire results, from a human efficacy evaluation at a named research centre, with 21 participants, over a printed date window.

So the numbers were not invented, and the brand told you what kind of numbers they are. What the page then does is typographic. A figure set in display type reads as a measurement, and this one is a tally of satisfied answers from 21 people who knew what they were using. There was no control arm and no blinding, and on a panel of 21, a single dissenting answer would have moved the figure to 95 percent.

One row shows the gap without any outside information at all. The same page prints instrumental cleansing results a few screens later: 93 percent for base makeup, 94 percent for sunscreen, 99 percent for particulate and 99 percent for pore debris. The brand's own instruments never reached 100. So the 100 percent cleansing row is a survey answer sitting a few screens above the measurement that disagrees with it, and only the small-type footnote tells them apart.

The oil and water balance row has a second problem: it names nothing anyone can measure. The page reports skin water content and barrier water loss, and neither of those is a ratio of oil to water. That row is a satisfaction score on a term with no instrument behind it, printed at the same weight as the two rows that do have one. The formula underneath is genuinely well built, with glycerin printed second and no soap base anywhere on the list, which is what makes the overselling unnecessary.

Twenty people, two days, nothing to compare against

Mamonde Amazing Deep Mint Clenasing Balm

The study behind this page is real and it is described: a named contract research company, 20 Korean women aged 20 to 35, run across two calendar days. The endpoints are reported separately rather than pooled into one flattering average, which is better practice than it looks. Pore debris cleansing comes in at 99.5 percent, blackheads and whiteheads down 24.6 percent, sebum down 30 percent, dead skin improved 55.2 percent, and five pore measurements down 26.5, 23.5, 20.8, 14.2 and 14.1 percent.

Now read the design. One arm, no untreated control, a single application, no statistical treatment stated and no instrument named. Washing your face changes every one of those readings on its own, so a change score with nothing to compare it against cannot separate the formula from the wash.

The oil figure is the clearest case. Surface oil fell 85.6 percent, measured after an oil-dissolving balm had been emulsified and rinsed off the face. Any product that cleanses at all moves that reading steeply. The number is close to a restatement of the fact that the balm was used.

Credit one more thing the brand printed itself: the accompanying pore image is labelled as the top three results by pore volume, with individual variation noted. That is the upper end of a 20-person panel rather than its middle, and the page says so.

A percentage attached to something that is not a quantity

Torriden DIVE IN Soothing Cream

Take the strongest footnote first, because it is genuinely strong. A human repeat insult patch test returned non-irritant on 50 subjects aged 20 to 60 over six weeks. That footnote names the design rather than only the result, and a six-week repeated occluded exposure is a harsher test than daily face use.

The efficacy block is built differently. Most of its figures come from 20 participants measured immediately after application: 145.02 percent improved immediate moisturizing, 15.55 percent improved deep hydration, a 3.46 degree drop in skin temperature, and 101.84 percent improvement in temporary redness from an infrared heat model. One figure, 30.85 percent improved skin water content, was measured after four weeks, and that timing makes it the most informative of the set.

The row to stop on is 109.77 percent improved soothing from external physical irritation. Physical external irritation is not a defined dermatological endpoint. The footnote gives 20 participants and an immediate measurement, and does not say what was provoked or how soothing was scored. Panthenol and allantoin are on the list and make a soothing effect plausible. A figure carried to two decimal places, attached to an endpoint nobody has defined, cannot be checked against anything.

Unanimous on seven questions

AESTURA Atobarrier 365 Cleansing Foam

The disclosure here sits above the category norm and deserves saying so. A panel of 30 women aged 20 to 39, a dated one-month window, three cleaning endpoints reported separately at 98.1 percent for ultrafine dust, 88.9 percent for sunscreen and 97.9 percent for pore debris, and an explicit note that individuals vary. The missing piece is the institution: the brand does not name who ran the tests, which is the one item that would let anyone check the rest.

The satisfaction block is where the design becomes visible. Seven items were rated, among them overall cleansing power, oil removal, rinsing without residue, suitability for sensitive skin, low irritation, and a sense of being protected during the wash. All seven came back at 100 percent satisfaction from the 30 respondents.

Seven separate judgments answered favourably by every single person is a statement about the questionnaire at least as much as about the cleanser. No scale is printed, no question wording, no comparison group. And one of the seven, a sense of protection during the wash, is a feeling rather than a quantity, which is the same shape as the oil and water balance row further up this page.

A test named without being described

make p:rem Safe Me. Relief Moisture Cleansing Foam

Two lines from this brand's launch copy, which reach a reader through a press article rather than as a published report, are worth separating from each other.

The first says the formula uses only raw materials graded 0 to 2, green, in a well-known hazard-screening database. That database is a third-party screening tool, not a regulatory safety determination, and those two things get read as one. The brand does not print a per-ingredient grade next to the claim, so there is nothing beside it to check.

The second says an irritation test was completed, and moves from there to suitability for all ages and all skin types. Grant the first half readily: the surfactants belong to low-irritation classes, the product is rinsed off in under a minute, and the list carries no sulfate, no drying alcohol and no exfoliating acid. The sentence still gives no test type, no number of people, no institution and no date, and then extends a completed test into universal suitability in the same breath as it advertises bergamot fruit oil and sage oil as a natural fragrance. A panel that did not react is not a statement about everyone, and fragrance is precisely where that claim gets tested.

The reading order

Take these in order and most study blocks resolve in under a minute.

Find the asterisk before you read the number. The footnote is where a brand tells you what kind of figure it is. Satisfaction, questionnaire, self-assessment and agreed are the words that turn a percentage into an opinion poll.

Count the people. Twenty to fifty is normal for cosmetic testing and is not by itself a defect. It does mean one or two participants move the headline several points, which is why unanimity should read as a fact about the questions rather than about the product.

Ask what it was compared against. Untreated skin, a plain wash, a previous formula and no comparison at all are four different studies. Single arm means the effect of the routine and the effect of the product arrive as one number.

Ask whether the endpoint is a quantity. Skin water content is a quantity. Barrier water loss is a quantity. Oil and water balance, protection during washing and physical external irritation are not, and no instrument produces them.

Ask when the measurement was taken. Immediately after use is the window in which almost anything looks its best. A four-week reading is worth more than a triple-digit percentage taken thirty seconds after application. The same timing trap turns up across the shelf, and we walked through five shapes of it separately.

Common questions

Does a small panel mean the product does not work?

No. It means the number in front of you is weaker evidence than its typography suggests. Every product here is well made, and each has reasons to buy it that do not depend on the disputed figure at all.

Why do satisfaction surveys come back at 100 percent so often?

Because unblinded participants who know which product they are testing, answering questions written by the party that wants a favourable answer, is a design that produces agreement. Unanimity across several separate items is the clearest signal of it.

Is an unnamed testing institution a reason to discount a figure?

It is a reason to hold it loosely rather than to throw it out. Naming an institution is what makes a result traceable by someone other than the brand. Brands that name one, and several here do, have raised their own cost of publishing a bad number.

What is the single fastest check?

Look for the word measured, and then for the word asked. If the page never says which one happened, the figure has not told you what it is.

If a brand published its study and you cannot tell how far the design will carry, send us the footnote and we will read it through with you.