A shopper opens your skincare catalog, scrolls past twenty bottles and leaves. Brands know this problem, so many add a skincare quiz: four or five questions, a progress bar, then a bundle. It feels like help. For the shopper, it often is not, and the reason is worth understanding before you spend a budget on a better quiz.
How a skincare quiz really works
Under the interface, most quizzes are a lookup table. If someone answers combination skin, concerned about spots, normal budget, the quiz points to a pre-written set of products. Two shoppers with completely different faces and the same four answers receive the same bundle. People feel that shallowness even when they cannot name it, and they stop trusting the result.
A skin type quiz has a second weakness. It asks people to describe their own skin, and most of us are poor judges of it. Someone who says their skin is dry may be dealing with a damaged barrier. Someone who says oily may be reacting to a heavy moisturiser. The quiz records the belief, not the skin.
The honest limits of self-reported answers
- People use different words for the same thing. One shopper calls it breakouts, another calls it spots, another calls it texture.
- People answer for the skin they wish they had, or for how it felt last winter.
- Short quizzes cannot capture where on the face a concern sits, or how strong it is today.
What a photo adds, and what it does not
A selfie gives you something a questionnaire cannot: what is visible on the face today. Which concerns can be seen, how strong they look and where they sit. A good AI skin analysis combines both sources, what the shopper says they care about and what the photo shows. If the two disagree, that is useful information, not a bug.
But a scan only helps if it is true. Getting a vision model to answer is easy. Making the answer true is the real work, and it is where most tools quietly fail. They produce a confident score from a bad photo, an odd light or a face the model did not expect. Our own testing showed leniency in scoring, invented oiliness under flash, stubble read as blemishes and a person of about sixty-five told that fine lines were severe. Each one needed a fix inside the system, not a nicer prompt. You can read the full list on our technology page.
Three things an honest scan has to do
- Describe before it scores. The model should list what it sees, photo quality, features to exclude and observations by region, before it is allowed to put a number on anything.
- Let something else check the claims. A separate validator should discard findings that are not justified, such as shine under camera flash or a shadow read as pigment.
- Keep the headline number out of the model’s hands. A fixed, tested function should compute the overall score from the verified findings, so it cannot drift with the mood of the model.
This is a design principle, not a feature. A tool that never says I cannot tell is a tool that is guessing. When a photo cannot support a claim, the right behaviour is to drop it. Under flash, for example, one photo cannot separate shine from sebum, so oiliness should not be reported at all.
The routine has to come from your shelves
A skin result is only useful if it ends in something a shopper can buy from you. That means the routine has to be built from the brand’s own catalog, with a plain reason for each pick, so the recommendation reads as advice and not as a pitch. It also means safety rules must be real rules. A declared allergen removes a product. It is not a small penalty on a score.
The research on this is broad. Recommender systems are a long-studied field, and the general point is simple: a recommendation is only as good as the data and rules behind it (overview of recommender systems). For skincare, the extra rule is that safety comes before preference.
What this means for your brand
If your conversion depends on a quiz, look at what happens after the quiz. Do shoppers buy the bundle and come back, or buy and disappear? A bundle that does not fit is a returns problem and a trust problem. If you are considering replacing a quiz with a scan, ask the questions in our buyer’s guide to AI skin analysis first.
A quiz picks a bundle. A good scan builds a routine, and can say why.
A fair test for any quiz or scan
Ask one question of whatever you use today: if two very different people gave the same answers, would they get the same routine? If the honest answer is yes, you are looking at a lookup table. A better tool produces different routines for different faces, and can tell each shopper why.
That is the design behind Claré: five questions, one selfie, a check on every claim and four products from your own catalog. Nothing is stored, and the report your customer takes home contains no photo. If you want to see it on your own range, get early access for your brand.
Frequently asked questions
Is a skincare quiz the same as AI skin analysis?
No. A quiz maps a few self-reported answers to a pre-written set of products. An AI skin analysis also looks at a photo, checks what it sees and builds a routine from the brand’s own catalog.
Can a selfie replace a skin type quiz?
A selfie shows what is visible today, but it works best together with a short questionnaire about sensitivity, concerns, allergies and budget. Claré uses both.
Cosmetic analysis only. Not medical advice.