'94% of Users Agreed': How to Read a Skincare Claim

“94% of users agreed skin looked more radiant in 4 weeks.”

You’ve seen this sentence, or something very close to it, on roughly half the products in your bathroom. It sounds like data. It has a number. It has a timeframe. It implies a study happened somewhere, conducted by someone, with results that mean something.

It almost certainly doesn’t mean what you think it means.

That’s not a cynical take — it’s just how cosmetic claims work. The gap between what a label says and what the underlying evidence actually shows is wide enough to drive a truck through. Learning to read that gap is more useful than any single ingredient guide we could write.

Here’s how to do it.

The Three Tiers of Skincare Evidence

Not all evidence is the same. Before you decode a specific claim, it helps to know what kinds of studies actually exist in cosmetic testing — and how they stack up.

Tier One: Consumer Perception Panels

This is the most common source of the “X% of users agreed” stat. A group of people — typically 30 to 50, sometimes fewer — use a product for a set period and answer a questionnaire. Did your skin feel more hydrated? Did you notice a glow? Did fine lines look improved?

The answers are self-reported. There is no measurement equipment involved. There is no placebo. There is no blinding — participants know exactly what they’re using and what the brand hopes they’ll experience. And because the sample is small, a single enthusiastic subgroup can easily tip the numbers.

This isn’t fraud. Consumer perception is a real metric. If 90% of users find a moisturizer comfortable, that’s worth knowing. The problem is when perception stats get presented as proof of a physiological effect. Feeling more radiant and measurably producing more collagen are different claims. They require different evidence.

Tier Two: Instrumented Studies

A step up: these use actual measurement devices. Corneometers measure hydration in the stratum corneum. Tewameters quantify transepidermal water loss. Cutometers assess skin elasticity. Mexameters can quantify melanin distribution for brightening claims.

Instrumented studies produce objective numbers. That’s better. But they still vary enormously in quality depending on sample size, duration, and whether they included a vehicle control — meaning a version of the product with the active ingredient removed, used by a separate group, to confirm the base formula itself isn’t driving the result. Without that control, you don’t know if the ingredient worked or if the humectants in the base are doing all the heavy lifting.

A 12-week instrumented study with a vehicle control and 80 participants is genuinely useful data. A two-week instrumented study with 20 participants and no control is not much better than the perception panel.

Tier Three: Randomized Controlled Trials

RCTs are the gold standard — randomized, double-blind, placebo-controlled, ideally peer-reviewed. Participants are randomly assigned to active or placebo groups. Neither they nor the researchers assessing outcomes know who got what. Results are published in journals where other scientists can scrutinize the methodology.

Most cosmetic brands do not run RCTs on finished products. It’s expensive, slow, and the bar for what counts as a cosmetic (versus a drug) limits what claims they can make anyway. RCTs exist mostly at the ingredient level — for retinol, niacinamide, vitamin C, established actives with decades of research behind them.

When a brand cites an RCT, it’s usually on an ingredient, not on their specific formula, concentration, or packaging format. That distinction matters.

Four Questions to Ask Every Claim

Once you know what tier you’re dealing with, four questions will tell you almost everything.

1. How many people?

Under 30: treat it as directional, not conclusive. Thirty to one hundred: better, but watch for other issues. Over 100 with a control group and peer review: now you’re getting somewhere.

The “94% of users agreed” stat is almost always a small panel. Occasionally you’ll find the actual number buried in the brand’s clinical literature or on the package insert. Look for it.

2. Agreed with what, exactly?

The exact wording of the question shapes everything. “My skin felt more hydrated” is a perception claim. “Hydration increased by 47% as measured by corneometry at 24 hours post-application” is an instrumented claim. They’re presented similarly on packaging. They’re not similar at all.

Watch for vague verbs: felt, appeared, looked, seemed. They signal perception data. Watch for specific mechanisms: collagen synthesis increased, elastin density improved, melanin transfer inhibited. Those require instrumented or histological evidence to be credible — and you should be skeptical if none is cited.

3. Compared to what?

No comparator means very little. Skin hydration compared to baseline, with no vehicle control, mostly tells you that applying any moisturizing product made skin feel more hydrated. That’s not news.

The most credible studies include: a baseline measurement, a vehicle control group, and ideally a comparison to an established benchmark (a percentage of tretinoin, a percentage of hydroquinone, however the ingredient performs against the standard of care). Without a comparator, the number floats free of any meaning.

4. Who funded it?

Brand-funded studies aren’t automatically invalid. But they’re more likely to feature favorable conditions — concentrations optimized for positive outcomes, short durations that capture early effects, endpoints chosen because they’re likely to move. Independent replication of the same finding is the thing that actually builds confidence.

If the only evidence for a $120 serum is two studies, both conducted by the brand’s own lab, both using proprietary measurement criteria that no one else can audit — treat the numbers accordingly.

The Language of Soft Claims

Brands are legally savvy. The claims that are hardest to challenge aren’t the big ones — they’re the carefully hedged ones.

“Clinically tested” means a test occurred. It says nothing about what was found, what the test measured, or what the results were.

“Dermatologist-tested” means at least one dermatologist used or evaluated the product. It is not an endorsement of efficacy.

“Clinically proven” is slightly stronger in implication but still legally loose. Proven by what study, at what threshold of significance, in what population? The claim doesn’t say.

“Up to X% improvement” means the best-performing participant or measurement saw that result. Most participants saw less. You might see much less.

“Targets” and “helps support” are hedging language. They allow a brand to associate a product with a benefit without committing to causing it.

None of this is illegal. Some of these products genuinely work. But the language is built to survive regulatory scrutiny, not to communicate scientific precision. Reading it charitably is a mistake.

Where Ingredient Evidence Lives (and Its Limits)

The strongest evidence in skincare is at the ingredient level, not the product level. Retinol (the OTC conversion precursor to tretinoin) has decades of peer-reviewed data supporting collagen upregulation, cell turnover, and fine-line reduction. L-ascorbic acid has solid mechanistic and clinical evidence for photoprotection and hyperpigmentation. Niacinamide has convincing data on barrier function and sebum regulation.

But ingredient evidence doesn’t automatically transfer to a product. Concentration matters enormously. Formulation context — pH, delivery system, stability — can make or break whether an ingredient even reaches the target layer of skin. A serum containing 0.025% retinol and a prescription for 0.1% tretinoin both “contain retinol.” The outcomes are not comparable.

Our oil-soluble vs water-soluble vitamin C guide covers exactly this problem for one of the most studied actives in skincare — same ingredient, radically different stability and delivery depending on the form. The ingredient evidence doesn’t tell you which form is in the bottle you’re holding.

And if a brand cites ingredient-level evidence for a proprietary extract, an untested complex, or an ingredient at concentrations too low to find in the literature — that evidence doesn’t apply. The research on bakuchiol at 0.5% doesn’t validate a product using 0.01% in ingredient position 23 of 24. For a full breakdown of how the evidence for bakuchiol specifically compares to retinol at different concentrations, see our bakuchiol vs retinol deep dive.

The “Natural” and “Clean” Claim Problem

Perception claims get even murkier when “natural” or “clean” enters the picture. These words have no legal definition in most markets. They read as quality signals. They function as marketing categories.

A product can be entirely plant-derived and irritate your skin significantly. A synthetic ingredient can be more stable, more bioavailable, and better-studied than its botanical equivalent. The naturalness of an ingredient and its efficacy or safety are separate questions — and the evidence for one doesn’t speak to the other. We’ve covered this in detail in our piece on the clean beauty myth around “chemical-free” skincare.

App-based scoring systems like Yuka have made this worse in some ways. A product that scores poorly because it contains phenoxyethanol (a well-studied, low-risk preservative) isn’t less safe than one that replaces it with botanical extracts that haven’t been tested for irritation potential. The Yuka score explainer gets into why high app scores and actual skin safety aren’t the same thing.

A Note on Where Your Advice Comes From

The sources driving most skincare decisions in 2026 are not peer-reviewed journals. They’re short-form video, brand-funded influencer content, and algorithm-curated posts optimized for engagement, not accuracy. We looked at this problem directly in our piece on where skincare advice actually comes from — the findings are not encouraging.

That doesn’t mean every recommendation from a content creator is wrong. It means the credibility filter has been removed from the pipeline. A convincing before-and-after and a properly controlled clinical study require the same scroll time, but they are not equivalent.

The fix isn’t to trust nothing. It’s to apply the questions above consistently: perception or measurement, how many participants, compared to what, and who paid for it.

Putting It Together

Skincare marketing has gotten very good at sounding like science. The vocabulary is right — “clinically proven,” “dermatologist-tested,” specific percentages, specific timeframes. But vocabulary isn’t methodology.

The short version:

“X% of users agreed” is self-reported perception from a small panel. It’s not nothing, but it’s close to the bottom of the evidence ladder.
Instrumented studies with vehicle controls and reasonable sample sizes are meaningfully better.
RCTs are the standard — and they mostly exist for well-established ingredients, not specific commercial formulas.
Ingredient evidence doesn’t automatically transfer to a product. Concentration and formulation context change everything.
Funding source and comparator are the two most useful things a citation can tell you — and brands rarely volunteer either.

You don’t need to read primary literature to buy good products. You do need to recognize when a number is doing work it isn’t qualified to do.

The next time a serum tells you 94% of users agreed — ask what they agreed to, how many of them there were, and what they were compared against.

Usually the answer is: a small group, feeling things, compared to nothing.