Skin Analysis API for CRM Segmentation: From Selfie to Audience

Beauty CRMs are full of behaviour and empty of biology. You know what a customer bought, when she opened the last campaign and how much she spent — but not her undertone, her skin type or whether her hair is fine and wavy. A skin analysis API for CRM segmentation closes that gap: the customer sends one selfie, and your database receives structured attributes it can filter, score and personalise on.

This article is about the operational side of that idea — what data you actually get back, where it lives, and which segments become possible once it does.

What the API returns and why it fits a CRM

MaIA Scan works as a simple request: your application sends a photo and receives a beauty profile as JSON. The response describes visible attributes of that photo — skin tone and undertone, skin type and condition, hair attributes — with a confidence value per field.

Two details make this CRM-friendly rather than just a nice front-end widget.

It is structured, not free text

Every field comes back in a predictable shape. That means you can map it directly to custom fields, traits or tags in your CRM without parsing prose or asking a human to interpret anything. A field like undertone becomes a filterable property in the same way that "city" or "last order date" already is.

Every field carries a confidence value

Photos vary: lighting, camera, angle. Instead of pretending every answer is equally solid, the API tells you how confident it is per field. In CRM terms, that is a quality flag. You can decide that segments used for paid media require high confidence, while a gentle "products for your skin type" newsletter can run on a lower threshold. You can also trigger a re-scan invitation when confidence is low, instead of quietly building campaigns on weak data.

Skin analysis API for CRM segmentation: the segments you unlock

Once beauty attributes sit next to purchase data, segmentation stops being generic.

Foundation and complexion

Shade and undertone are the classic pain point in beauty e-commerce. With those fields stored per contact, you can build a "foundation shade unknown" segment and invite exactly those people to scan; then a "new shade launched in your range" campaign that only reaches people whose stored tone matches the extension. No more blasting a 40-shade launch to the whole base.

Skin type and condition

A cleanser campaign written for oily skin performs differently when it only goes to contacts whose profile indicates oily skin. The same logic covers routine-building flows: a three-email sequence can branch at step one based on skin type rather than on a self-declared quiz answer the customer half-guessed.

Hair attributes

Haircare brands can segment by the attributes that actually drive product choice, then hold that segment across channels — email, WhatsApp, on-site blocks — instead of rebuilding it each time.

Combined behaviour + profile segments

The most valuable segments mix both worlds: customers with dry skin who bought a cleanser but never a moisturiser; customers whose undertone suggests a shade you now stock; lapsed buyers whose profile matches a new launch. These are the cross-sell lists that behaviour alone cannot produce.

How the data gets in: three realistic entry points

You do not need a new app to collect scans. You need one moment where a selfie makes sense.

Onboarding. After signup, a short "build your beauty profile" step. The scan replaces a long questionnaire and gives the customer something back immediately — a profile and better recommendations.

WhatsApp. In Brazil especially, customers already send photos in chat. A scan inside a WhatsApp flow is native behaviour, and the resulting profile can be written to the CRM contact just like any other event.

Post-purchase. A follow-up message inviting the customer to scan so future recommendations fit better. Enrichment on an audience that already trusts you tends to convert best.

In all three cases, the integration pattern is the same: your backend calls the API, receives the JSON, writes the fields to the contact record, and stores the raw response if you want auditability. Full request and response details are in the API documentation.

Consent, storage and honest expectations

A selfie is personal data. Treat it that way. Ask for explicit consent, explain in plain language what the photo is used for, and decide up front whether you need to keep the image at all — in many cases, storing only the resulting attributes is enough and far simpler to govern. Give customers a way to delete their profile, and keep the scan optional rather than gating checkout behind it.

One boundary matters: the analysis describes visible characteristics in a photo. It is not a diagnosis, it does not detect conditions and it does not replace a dermatologist. Keep your copy on the product side — shade, type, routine fit — and both your legal team and your customers will be more comfortable.

From segment to conversation

Segmentation is only half the value. The other half is what the customer sees next. Once a profile exists, it can also feed a conversational layer: MaIA Beauty Advisor reads the same beauty profile and recommends products from your own catalog on WhatsApp, product pages and chat. The CRM segment sends the message; the advisor handles the follow-up questions the message provokes.

That combination — structured attributes in the database, a catalog-aware advisor in the conversation — is what turns a one-off selfie into a durable personalisation asset.

A pragmatic first sprint

Start narrow. Pick one entry point, map five fields to your CRM, build two segments you can measure against a control group, and run for a month. If the profile-based segment outperforms, expand the entry points. That is a small engineering lift with a clear read on value.

Ready to test it with your own data? Create an account, get your API key and run your first scans today — or talk to our team about mapping the profile into your CRM.