Selfie Skin Analysis In-Store: How to Equip the Counter With Real Data

Beauty retail still relies almost entirely on the visual read of whoever is behind the counter. An experienced consultant gets the foundation shade right by eye; a new hire, on a busy Saturday, not always. Selfie skin analysis in-store fixes that drift in a simple way: the customer takes a photo on a tablet, the system returns a structured profile in seconds, and the sales conversation starts from a data point shared by the whole team — not from each person's intuition.

This article is about how to build that flow in practice: what runs on the device, what comes back from the API, how to connect the result to the catalog and the CRM, and what to watch out for before rolling it out to dozens of points of sale.

Why the counter needs standardised data

Three problems show up in almost every own-store or kiosk operation:

• Variation between consultants. Two professionals suggest different shades for the same customer. The exchange becomes a return, and the return becomes a cost.
• Staff turnover. Teaching undertone and skin type takes time; the team changes before the training matures.
• Nothing gets recorded. The best beauty conversation of the month happens in person and disappears. The customer comes back three months later and starts from scratch, or buys online and is treated as anonymous.

A beauty profile captured at the point of sale tackles all three at once: it standardises the read, speeds up a new hire's autonomy and turns the service interaction into the brand's own data.

How selfie skin analysis works in-store

MaIA Scan is an API: the device sends a photo and gets back a JSON with the beauty profile — skin tone and undertone, skin type and condition, hair attributes and a confidence score per field. The model was trained on Brazilian skin, which matters in a country where the distribution of tones and undertones looks nothing like that of imported catalogs.

In the store, this becomes a single screen.

What runs on the tablet

The visible layer is a simple web page — a PWA open on a tablet at the counter, a self-service kiosk or the consultant's own phone. It does three things: captures the photo, calls your internal API (which in turn calls MaIA Scan with the API key kept on the server, never on the device) and shows the result in sales language, not raw JSON.

What the consultant sees

Don't show the payload. Translate it. Instead of undertone: warm, the screen says: "warm undertone — start with the golden family of foundations". Instead of displaying every field, show the three or four that change the recommendation in that category. The per-field confidence score is for your team, not for the customer: when it comes back low, the interface asks for a new photo in better light instead of pushing a weak suggestion.

Light, background and framing

A store is a hostile lighting environment: yellow spotlights overhead, a shop window with street light behind. It pays to define a fixed capture point, with a neutral background and diffuse frontal light, and to train the team to always use the same spot. Consistency of the capture point is what improves result quality the most in physical retail.

From profile to recommendation

A profile without a mapped catalog sells nothing. The next step is to link each combination of attributes to the brand's real SKUs: which foundation shades serve each tone and undertone range, which skincare lines match each visible condition, which hair products correspond to each attribute.

When that mapping needs to talk to stock, campaigns and brand language, it makes sense to use MaIA Beauty Advisor, the AI advisor integrated with the catalog — delivered as a project with the B4A team, not as a self-service API. It gives the store team the same recommendation the customer receives on WhatsApp or on the product page, which removes the classic contradiction between channels.

Virtual try-on as the second step

After the profile, the customer usually wants to see. MaIA Try-On comes in two forms: the photo API, which renders lipstick or foundation on the image with a palette calibrated to the brand's real products, and the browser SDK, in early access since September 2026, with thirteen products and Full Look in live camera mode or photo mode, rendered locally in the browser. At the counter, live mode has an obvious appeal; access is granted per account, so it is worth aligning this with the team before designing the journey.

Connecting the store to the CRM

The long-term value is not in that day's sale. It is in saving the beauty profile to the customer record, with explicit consent and a stated purpose, and using it later: email segmentation by undertone, replenishment reminders in the right shade, recognising the same profile when the person lands on the e-commerce site. The physical store becomes the entry point for first-party data, not an isolated channel.

One point that is non-negotiable: the analysis describes visible attributes in a photo. It does not treat anything, it is not a clinical assessment and does not replace a health professional. The team's script needs to reflect that.

What to measure in the first weeks

Start with a pilot in a few stores and track: number of analyses per store and per consultant, conversion rate of interactions with and without analysis, average order value, exchange rate due to the wrong shade and opt-in rate for the customer record. Results vary with catalog, store traffic and channel — which is why the comparison has to be internal, between similar stores, and not against third-party figures.

The integration itself is straightforward: an account, credits, an API key and one call. See the three MaIA products and create your account and API key to test the store flow with real photos of your audience — or talk to the team to design the pilot together.