Selfie Skin Analysis in Your Brand's App: From Onboarding to the Right Notification

A brand's own app tends to be its most underused channel. The customer has already downloaded it, already accepted notifications, is often already logged in — and the experience is still a catalog with a login screen. Selfie skin analysis in your brand's app changes that: instead of asking for yet another sign-up, the app asks for a photo and immediately gives something back, a beauty profile that starts shaping the storefront, the search results and the messages that follow.

This article covers the mobile angle: where to place the capture, how to organize the architecture between app, backend and API, and what to decide before you ship to the stores.

Why the app is a good place for selfie skin analysis

Three practical reasons.

First, the camera. Asking for a selfie inside an app feels natural — camera permission is already part of the user's vocabulary, and capture quality tends to beat a desktop upload.

Second, identity. On e-commerce, much of the traffic is anonymous and the profile has to be tied to a cookie or a session. In the app, the person is usually logged in, so the beauty profile is born attached to a customer ID that already exists in the CRM.

Third, recurrence. The app gets opened again. A profile captured once keeps paying off in every new session, without repeating the question.

What the app gets back

MaIA Scan works as an API: the photo goes in, and back comes a JSON with visible attributes — skin tone and undertone, apparent skin type and condition, hair characteristics — with a confidence score per field. The model was trained on Brazilian skin, which matters for anyone selling here who has seen shade descriptions that do not match their real customer base.

Worth stating clearly in the product and in your communication: the reading describes visible attributes of a photo. It is not a medical assessment and does not replace guidance from a professional.

Mind the try-on format inside the app

If the plan includes makeup try-on, separate the routes. MaIA Try-On has a photo-based API that works well in a native app: the app sends the image, the product is rendered on it, and the image comes back. The SDK, in early access, runs in the browser, in live camera mode or photo mode, self-hosted by the brand — inside an app, that means a web surface within the flow. Decide this early, because it changes the screen design.

Where to place the capture in the app flow

In onboarding, as an explicit trade

The screen right after sign-up is the obvious spot, as long as the ask is honest: "take a selfie and we'll organize the app around your skin". Always leave a skip option. A profile captured under pressure tends to become an abandoned profile.

In a "my beauty profile" tab

A permanent area, with the result in customer-friendly language and the option to redo the analysis. That solves two things: it gives the person control and creates an entry point for anyone who skipped onboarding.

At the moment of doubt

On a foundation or concealer page, a "find my shade" button usually converts better than the same generic offer on the home screen — the question is in the person's head right at that second.

In in-app campaigns

A new collection, a seasonal quiz, a points club. The selfie comes in as a step in a mechanic that already has its own reason to exist.

Architecture: the app does not talk to the API directly

The API key should not travel inside the app binary. The standard path is: the app captures the photo → sends it to the brand's backend → the backend calls the API and stores the result → the app receives the processed profile.

That brings three benefits: the credential stays on the server, the backend can apply retry and image-quality rules, and the profile is persisted in the same place where orders and history already live. The API documentation covers the account, credits and key flow for anyone prototyping this during the week.

Guide the capture before spending a credit

Front lighting, centered face, no sunglasses, no camera filter. A framing mask and a simple on-device sharpness check avoid useless calls and low-confidence results. Also show what happens to the photo: if it is discarded after processing, say so on screen, not only in the privacy policy.

What changes in the CRM and in notifications

Once the profile is in the CRM, a push stops being an ad and becomes continuity. The message about a foundation launch can go only to people with a compatible undertone. The replenishment cadence can take the recorded skin type into account. Internal search can sort results by the profile instead of showing the same "best sellers" to everyone.

If the brand wants to turn this into a conversation, the virtual beauty advisor reads the same profile and recommends products from the brand's own catalog on WhatsApp, on the product page or inside the app — it is a project run with the B4A team, not a self-serve API.

Before go-live: a short checklist

• Consent copy reviewed by legal, stating purpose and retention period.
• Defined behavior for low confidence: ask for a new photo or show a partial result.
• Offline fallback and network error handling with a clear message.
• Profile version stored with a date, so you know when to suggest a new selfie.
• Agreed metrics: how many start the capture, how many finish, how many return to the profile tab, and what happens to conversion among people who have a profile.

Results vary with each brand's catalog, traffic and channel — which is why the pilot needs a comparison group from day one.

Want to test the API response with real photos from your audience? Create an account, buy credits and generate your key — or talk to the team to design the flow inside your app.