Skinwise

The Challenge

Skincare is intensely personal, but most D2C brands sell it like it isn't. A customer lands on a product page, reads a description written for everyone, and has to guess whether it's right for their specific skin type, concerns, and environment. The mismatch shows up in returns, negative reviews, and customers who buy once and don't come back.

Physical retail solved this with consultants trained staff who look at a customer's skin and recommend accordingly. Online, that interaction disappears. Quiz-based tools exist, but they rely on self-reported data: customers often don't know their skin type, can't accurately describe their skin's texture or tone, and give different answers on different days.

The client a D2C skincare brand needed a way to replicate the consultation experience online, at scale, without a human on the other end. The experience had to be fast enough that users wouldn't abandon it, and accurate enough that the recommendations it produced actually built trust.

What We Built

Skinwise is a consumer-facing web application that takes a single photo input and produces a personalized skincare recommendation in real time.

When a user uploads or captures a photo, a computer vision model analyzes the image for visible skin characteristics: overall skin tone and undertone, visible texture (fine lines, pores, uneven texture), signs of oiliness or dryness, areas of redness or hyperpigmentation, and general hydration indicators. This analysis produces a structured skin profile that is specific to what the model can actually see in the image not what the user thinks they know about their skin.

The skin profile is then passed to a recommendation engine that maps the profile against the brand's product catalog. Each product in the catalog is tagged with the skin profiles it is appropriate for, the concerns it addresses, and the order in which it fits into a skincare routine. The system returns a ranked set of product recommendations, organized by routine step (cleanser → toner → treatment → moisturiser → SPF), with a plain-language explanation for each recommendation.

Users can also input specific concerns acne, sensitivity, anti-aging which are layered on top of the image analysis to further personalize the output.

The entire experience from photo upload to full recommendation display runs in under 10 seconds.

Key Capabilities

• Photo upload and real-time computer vision skin analysis
• Multi-attribute skin profiling: tone, texture, oiliness, redness, hydration signals
• Concern-based personalization layer on top of image analysis
• Routine-aware product recommendation (organized by skincare step)
• Plain-language explanation for every recommended product
• Ranked catalog matching with confidence-based ordering
• Embeddable widget format for integration into existing brand e-commerce sites
• Mobile-first UX with native camera capture support

Tech Stack

Outcome

Skinwise delivers a complete skin analysis and personalized product recommendation in under 10 seconds from photo upload a meaningful UX threshold for consumer applications. Users who went through the Skinwise flow showed higher engagement on the recommendation screen compared to users who browsed the product catalog directly, with longer time-on-page and higher add-to-cart interaction on recommended products.