Closet Genie
Timeline
12 months (ongoing)
Size
6 people
Platform
Native mobile app (iOS-first)
12 months (ongoing)
6 people
Native mobile app (iOS-first)
As the Founding Product Designer, I led product design for Closet Genie, a native iOS fashion app focused on wardrobe-first styling and affordable fashion discovery. To accelerate iteration and validate ideas quickly, I also contributed to frontend implementation, working closely with engineering to translate core UX flows into production-ready interfaces.
Closet Genie brings lightweight AI to budget-conscious young women who want to look put-together without overspending—helping them build outfits from what they already own and find affordable pieces that complement their existing wardrobe. Rather than complex color analysis or body mapping, the app focuses on practical needs: "what can I wear with what I have?" and "where can I find an affordable version of what I saw online?" Gamification through styling challenges, avatar competitions, and community contests keeps the experience engaging rather than purely utilitarian.
Designed the experience to prioritize inspiration-led shopping flows that naturally guide users from browsing to purchase. Optimized recommendations around what looks compelling on the avatar and aligns with trending aesthetics, rather than minimizing overlap between items. Used the user's existing wardrobe as contextual input to suggest complementary additions that expand styling possibilities and encourage continued exploration. Streamlined interactions and navigation to support fast, scrollable discovery and frequent engagement with new products. Emphasized complete-look and avatar previews to clearly communicate how additional pieces elevate the overall styling outcome.
Users decide what to buy by visualizing complete styled looks, not by evaluating fit details, fabric descriptions, or rational comparisons. Stylized avatar-based outfit previews sparked desire by presenting clothing through an aspirational, game-like visual lens, leading users to become highly interested in the recommended items and actively seek ways to purchase them.
Wardrobe-aware, lightweight AI recommendations increased users' interest and purchase intent by framing new items as stylistic expansions of what they already own, rather than constraints or replacements. While users could browse and decide independently, lightweight AI reinforced decisions and encouraged further exploration and purchasing.
Prioritizing clear, guided entry points over open-ended AI input and early interactive elements reduced onboarding friction and helped users reach the core styling and shopping flow more quickly, at the cost of removing free-form input and exploratory interactions in the initial experience.
The first version required users to type their styling request in their own words, and the system generated outfits based on detected keywords. In practice, this created friction because users had to "know what to say," and the lightweight language model performed best only when inputs matched a narrow set of recognizable terms, leading to hesitation, slower onboarding, and inconsistent results.
We replaced free-form input with AI-curated CTA options so users could select an intent instead of composing a prompt. This reduced cognitive load and made the entry step feel more guided and predictable, trading expressiveness for speed and clarity while still enabling the AI to generate outfits without requiring users to write anything.
Trade-off: Less expressiveness in exchange for speed, clarity, and reduced onboarding friction.
In the initial navigation, the avatar entry was visually emphasized to signal it as the core feature, using a prominent, "raised" treatment that stood out from the other tabs. While it drew attention, the strong visual weight disrupted hierarchy and made some users interpret the avatar as the only meaningful starting point, rather than one feature within a broader shopping and styling flow.
We rebalanced the navigation to give equal weight to all primary features—avatar styling, trending, shopping, and closet management—allowing users to enter the app from any angle based on their current intent.
Trade-off: Less immediate feature emphasis in exchange for clearer navigation hierarchy and flexible entry points.
87% of participants successfully found and selected an item matching their style preferences, with average time-to-decision of 47 seconds.
Participants completed full outfit creation with 94% task success, and 73% expressed strong intent to purchase items after virtual try-on.
87% of participants successfully found and selected an item matching their style preferences, with average time-to-decision of 47 seconds.
Saved items generated 2.4x higher click-through to affiliate stores compared to unsaved browsing, with 67% of participants revisiting saved outfits before purchase intent.
Initial wireframes explored a gamified experience featuring customizable avatars, clothing items displayed as collectibles, and a "looks" system for saving outfit combinations. However, discovery interviews (n=20) revealed a critical misalignment: while users accepted avatars as styling inspiration, they prioritized visual previews of complete looks over collection-building mechanics. Participants wanted the app to "just work" visually—showing them styled outfits that sparked desire—rather than engaging with achievement systems or wardrobe management features. This insight prompted a pivot from gamification toward an AI-assisted recommendation flow that keeps the technology invisible while delivering the aspirational, full-look compositions users actually respond to.
Closet Genie launched in early January 2026, with core user flows validated through initial testing. The virtual try-on experience, wardrobe organization system, and outfit recommendation engine are live and in users' hands. Early feedback has shaped key interactions—particularly around the avatar customization flow and the balance between AI suggestions and user control.
Moving forward, we plan to integrate AI-powered facial rendering so users see themselves—not just a generic avatar—wearing outfits, increasing emotional connection and purchase confidence. We'll also expand avatar personalization beyond body measurements to include skin tone matching, hair styling options, and posture/stance preferences for more accurate representation. On the business side, we aim to onboard additional retail partners to broaden our product catalog and create more monetization pathways through commission-based recommendations. Throughout this process, we'll continue to actively gather and incorporate user feedback to refine try-on accuracy, recommendation logic, and the overall experience.