StratumSkin science, cited

Methodology

How to read a study, and how we grade one

Every grade on this site is a judgment call, made using the same handful of questions each time. This page is those questions, written out, so you can apply them yourself the next time a bottle promises something a study never quite said.

Start with who ran it, and who paid for it

The single fastest useful question. A study funded by a company that sells the ingredient it's testing isn't automatically wrong. Several of our entries rest on exactly that kind of study, graded accordingly. But it tells you how much independent scrutiny the result has actually had, and how surprised to be if an unfunded lab finds something smaller later. We name the funder in every reference we publish, for that reason.

Ask what was actually measured

‘Improved skin texture’ can mean a dermatologist's visual scale, a machine reading of elasticity, or two researchers agreeing a photograph looked better. None of those is fake, but they're not the same claim, and only one is close to what you'd notice in a mirror. The word ‘endpoint’ or ‘outcome measure’ in a study tells you what was actually tracked, as against what a press release implies was tracked.

Check who was in it, and whether there was a control group

A trial of 14 people is a hint, not a verdict. Several of our Grade C entries rest on studies exactly that size. A trial with no control group can't separate the treatment's effect from time passing, a change in weather, or the placebo effect, all of which move skin measurements on their own. ‘Randomised’ and ‘controlled’ in a study's description are doing real work, and their absence is worth noticing.

Significant is a statistics word, not a mirror word

When a paper calls a result ‘statistically significant’, it means the result was unlikely to be chance. It doesn't mean it's large, and it doesn't mean you'd see it. Our own niacinamide piece cites a wrinkle study that was significant on measuring instruments and never asked whether anyone could see the difference. Both facts can be true of the same study: the effect is real, and may be too small to matter to you.

One study is a data point, not a verdict

Individual trials go wrong for ordinary reasons: a small sample, an unusual population, plain bad luck. That's what systematic reviews and meta-analyses are for: pooling several trials to see whether an effect holds up across them. We grade a claim higher when independent studies agree, and say so plainly when a grade rests on one trial that hasn't yet been repeated.

Watch for the sorting trick

Our collagen peptides correction is the clearest example on this site. Pooled together, the trials showed a benefit. Split by who funded them, the benefit lived almost entirely in the industry-funded studies. The pooled number wasn't fabricated. It was real, and it was also hiding something a funding breakdown revealed immediately. Whenever a pooled result surprises you, it's worth asking whether it survives being split by quality or funding.

How that becomes a Stratum Grade

The four grades below are the same ones attached to every ingredient, claim and article on this site. They describe how well something is known, never how well it works.

A
Well established

Backed by systematic reviews or several separate randomised controlled trials that point the same way. Different research groups, different patients, same result. This is as certain as skincare gets.

B
Good evidence

At least one properly run human trial with a control group, but it hasn't been widely repeated yet. Probably true. Worth acting on. Not yet carved in stone.

C
Suggestive

Small studies, no control group, or research paid for by whoever sells the ingredient. None of that makes it wrong. It does mean the result could shrink or vanish when someone independent checks.

D
Lab only

Tested on cells in a dish or on animals, never on actual people using it on actual skin. Plenty of ingredients look brilliant in a petri dish and do nothing on a face.

The Sunday study

One paper a week, explained properly

We pick a single new study, tell you what it measured, how many people it measured it in, and whether it should change anything you're doing — plus the two articles we published that week. No panic, and no hype.

We don't recommend products yet. When we do, they'll be things you can actually buy in India, any affiliate link will be labelled as one, and a commission will never buy a recommendation.

One email each Sunday. Unsubscribe in one click.