Skin Analysis API for Pharmacy E-Commerce: Guiding Shoppers Across Dozens of Brands

Pharmacy and drugstore chains are one of the largest beauty channels in Brazil, and also one of the hardest to shop. A single category page can hold dozens of brands, each with its own vocabulary for the same thing: oily, oil-prone, matte control, shine control. A skin analysis API for pharmacy e-commerce solves that from the shopper's side — it reads a selfie and returns a structured beauty profile, so your site can reason about the person instead of asking them to decode the shelf.

Why pharmacy catalogs break generic personalization

Most recommendation engines in retail are built on behaviour: what people clicked, what they bought together, what is trending. That works well for a shampoo repurchase. It works badly for the first-time skincare or foundation decision, where the right answer depends on attributes the shopper may not be able to name.

Three characteristics make the pharmacy channel specific:

• Multi-brand competition. The same shopper sees five brands answering the same need, with incompatible shade naming and claim language.
• Thin product data. Pharmacy e-commerces often inherit supplier feeds, which means inconsistent attributes across brands for skin type, finish or coverage.
• Low-consideration traffic. A lot of visits come from search or an app push, with no intent to fill in a 12-question quiz.

A beauty profile fixes the first problem by giving every brand a common denominator. It makes the second problem visible, because you can only rank what you have tagged. And it reduces the third, because a selfie is one step instead of twelve.

What a skin analysis API for pharmacy e-commerce actually returns

MaIA Scan works as a plain API call: your front end or back end sends a photo, and the response is JSON describing visible attributes — skin tone and undertone, skin type and condition, hair attributes, and a confidence score per field. The model is trained on Brazilian skin, which matters in a channel whose customer base spans the full range of tones in the country.

Two points worth being precise about with internal stakeholders:

It describes, it does not diagnose. The API reports what is visible in a photo. It is not a dermatological assessment, it does not diagnose or treat anything, and it does not replace a professional. In a pharmacy environment, where shoppers may arrive with a health question, that boundary has to be written into your copy, not just your contract.

Confidence is part of the payload. Each field carries a confidence value, so you can decide what to do when a photo is backlit or framed badly: ask for a retake, fall back to a neutral ranking, or show the recommendation with softer wording.

Mapping attributes to a multi-brand catalog

The profile is only useful if your catalog speaks the same language. Before launch, agree on an internal taxonomy — skin type, undertone family, finish, coverage — and map every supplier's wording onto it. This is usually the longest task in the project and the one that pays off beyond the selfie: a clean taxonomy improves filters, search facets and merchandising rules at the same time.

Where it fits in the shopper journey

Product detail pages

On a foundation or base PDP, the profile answers the only question that blocks the purchase: which shade. Pair the analysis with MaIA Try-On, where the shopper sends a photo and gets the product rendered back on it, with palettes calibrated to the brand's real products. The analysis narrows the shade list; the photo-based try-on confirms the choice.

Category and search pages

On a crowded facial-care listing, use the profile to re-rank rather than to filter. Keep the full assortment visible, but push the SKUs that match the shopper's skin type and tone to the top, and label why: "selected for combination skin". Pharmacy shoppers compare brands — removing options reads as a trick, reordering them reads as help.

App, CRM and the pharmacist conversation

Store the profile against the customer record and it becomes a lifecycle asset: segmented campaigns, app home shelves, and a starting point for a human or automated consultation. If you want a conversational layer on WhatsApp or on the site that reads the profile and recommends products from your own catalog, that is the job of MaIA Beauty Advisor, delivered as a project with the B4A team rather than as a self-serve endpoint.

A realistic rollout

1. Pick one category. Facial care or base makeup, where attribute-based ranking has the clearest effect.
2. Tag that category first. Map 100–300 SKUs to your taxonomy before you write a line of integration code.
3. Call the API from a single surface. One entry point — a PDP module or a landing page — so the data is clean enough to read.
4. Define fallbacks. What the page shows with low confidence, with no photo, and when a shopper declines.
5. Measure against a holdout. Compare profiled sessions to a control group on add-to-cart and on exchange rate for the same SKUs. Results depend on your catalog, your traffic mix and your channel, so measure yours rather than importing someone else's number.

Privacy and consent in a health-adjacent channel

Ask for explicit consent before the photo is sent, explain in one sentence what will happen to it, and state your retention rule. Many retailers choose to store only the resulting attributes and not the image itself, which is simpler to explain and simpler to defend. Keep the wording cosmetic throughout: appearance, shade, finish, routine — never diagnosis.

Ready to test it against your own assortment? Create an account and an API key in the MaIA developer documentation, or talk to the B4A team about a pilot on one category.