J.
Jess Kim (b. Seoul)  is a NYC based Product Designer. She reads where people stall inside complex tools, then designs to move them. E-commerce trained, fast to ship. An artist at heart.
Get in Touch: jess.young.design@gmail.com
The human behind it
PRODUCT DESIGN


CRAFT
©2026 J.Young EmailLinkedInResume
.zoom { transition: transform .3s; width: 200px; height: 200px; } .zoom:hover { transform: scale(1.1); } .shimmer-text { background: linear-gradient( 90deg, rgba(255,255,255,0.35) 0%, rgba(255,255,255,0.7) 40%, rgba(255,255,255,0.99) 50%, rgba(255,255,255,0.7) 60%, rgba(255,255,255,0.35) 100% ); background-size: 200% auto; -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text; animation: shimmer-pass 13s ease-in-out infinite; animation-delay: 0.5s; } @keyframes shimmer-pass { 0% { background-position: 150% center; } 35% { background-position: -50% center; } 100% { background-position: -50% center; } }
001 AI-Integrated Discovery Optimization
 Redefining Search & Discovery with AI-Vision · +74% ATC Uplift
WCONCEPT


Experience the Future of Discovery
"By integrating AI-driven visual attributes, we transformed a static product list into an interactive discovery engine that understands user intent beyond keywords — achieving a +74% ATC uplift."






Overview: To break through the limitations of keyword-based navigation, we executed a strategic initiative to integrate AI-Vision technology into our search and filtering systems. Representing the design track within a cross-functional Task Force (TFT), I collaborated with department leads to implement semantic search and interactive visual filtering (Virtual Mannequin). This comprehensive overhaul transformed a static product list into an intuitive discovery engine, contributing to a 74% increase in the Add-to-Cart (ATC) rate the team tracked.





 Context:
As Wconcept’s catalog expanded beyond 100K+ SKUs, traditional text-based search became a significant barrier. Data revealed a high frequency of "search-to-abandonment" loops, where users were forced into repetitive manual filtering or deep-paging (navigating to 2nd/3rd pages) because initial results failed to meet their intent.
The mission was to eliminate these discovery dead-ends by leveraging AI that understands "style intent" beyond literal text matches.




 A. Problem Discovery
Identifying Search Inefficiency through Behavioral Audits


The Voice of Users
45% of respondents identified Search and Product Lists as their primary pain points.
High Bounce / Re-query (19%+)
Immediate exit due to low search relevance.
Inefficient Sifting (3~4%)
Excessive reliance on manual filtering tools.
Discovery Failure (12.56%)
Users forced into deep-paging beyond the 1st fold.
The Voice of Users
45% of respondents identified Search and Product Lists as their primary pain points, calling for a fundamental discovery overhaul.
Behavioral Evidence of Discovery Failure
Heatmap analysis reveals a significant 'search-to-exit' loop. Approximately 20% of users immediately re-engaged with the search bar or retreated to category lists after seeing the initial results. Combined with heavy pagination (12.56%), this data proves the legacy engine's inability to surface intent-matched products on the first fold.

a.
Action Synthesized qualitative survey data with post-search behavioral heatmaps to identify friction points.
b.
Impact Validated that 45% of users identified Search and Product Lists as their primary pain points. Heatmaps further confirmed a 19%+ bounce/re-query rate and critical reliance on deep-paging (12.56%), proving the legacy engine's failure to surface relevant items on the first fold.
a.
Action Synthesized qualitative survey data with post-search behavioral heatmaps to identify friction points.
b.
Impact Validated that 45% of users identified Search and Product Lists as their primary pain points. Heatmaps further confirmed a 19%+ bounce/re-query rate and critical reliance on deep-paging (12.56%), proving the legacy engine's failure to surface relevant items on the first fold.





B. Solution: Beyond Keywords
 Transitioning from Manual Sifting to Intent-Based Exploration
B-1. Search
BEFORE
AFTER
B-2. Filter
BEFORE
AFTER - mannequin filter
AFTER - occasion filter (wedding)
AFTER - occasion filter (festival)

a.
Action Shifted from literal keyword matching to AI-Vision attribute analysis, allowing the engine to "see" and categorize products based on visual data (e.g., lace details, silhouettes).

Introduced an Interactive Virtual Mannequin filter, enabling users to filter 100K+ SKUs by simply touching garment areas instead of using technical fashion terms.
eg. design filters with newly introduced mannequin filter for some of the categories
TOPS
PANTS
SKIRTS
JACKETS & COATS

b.
Impact Successfully lowered the cognitive barrier for discovery and eliminated navigation dead-ends, ensuring a seamless transition from search query to relevant product exposure.
a.
Action Shifted from literal keyword matching to AI-Vision attribute analysis, allowing the engine to "see" and categorize products based on visual data (e.g., lace details, silhouettes).

Introduced an Interactive Virtual Mannequin filter, enabling users to filter 100K+ SKUs by simply touching garment areas instead of using technical fashion terms.

TOPS
JACKETS & COATS
PANTS
SKIRTS
b.
Impact Successfully lowered the cognitive barrier for discovery and eliminated navigation dead-ends, ensuring a seamless transition from search query to relevant product exposure.





C. Recommendation: Personalized Discovery Hub
Transforming Product Detail Pages into Tailored PLPs
BEFORE
AFTER

a.
Action Deployed a multi-layered recommendation feed at the bottom of the PDP, transitioning from a passive tab-system to an active discovery flow (Brand Affinity, Aesthetic Similarity, Cross-Category Coordination).

Implemented sticky anchor navigation (Details, Reviews, Recommended) and a "Back to Top" shortcut to maintain usability across the extended page depth.
b.
Impact Maximized product exposure by aligning recommendations with the user's immediate aesthetic interest, effectively creating a "Personalized PLP within a PDP" that significantly increased high-intent engagement.
a.
Action Deployed a multi-layered recommendation feed at the bottom of the PDP, transitioning from a passive tab-system to an active discovery flow (Brand Affinity, Aesthetic Similarity, Cross-Category Coordination).

Implemented sticky anchor navigation (Details, Reviews, Recommended) and a "Back to Top" shortcut to maintain usability across the extended page depth.
b.
Impact Maximized product exposure by aligning recommendations with the user's immediate aesthetic interest, effectively creating a "Personalized PLP within a PDP" that significantly increased high-intent engagement.





D. Performance & Validation
Proving Success through High-Intent Conversion Metrics
74%Increase in ATC Rate Achieved a significant uplift in the Add-to-Cart rate by reducing discovery friction through visual-first navigation.
32%Growth in search conversion Improved the relevance of search results by decoding semantic intent and visual attributes.
2.4xHigher Product Exposure Expanded the variety of products encountered by users through a multi-layered, AI-driven recommendation logic.
Scalable
DiscoveryAlgorithmic Architecture
 Shifted from manual product curation to an automated, AI-vision powered framework that scales with 100K+ SKUs.
74% Increase in ATC Rate Achieved a significant uplift in the Add-to-Cart rate by reducing discovery friction through visual-first navigation.
32% Growth in search conversion Improved the relevance of search results by decoding semantic intent and visual attributes.
2.4x Higher Product Exposure Expanded the variety of products encountered by users through a multi-layered, AI-driven recommendation logic.
Scalable
Discovery Algorithmic Architecture Shifted from manual product curation to an automated, AI-vision powered framework that scales with 100K+ SKUs.

*Measurement: metrics were tracked by the internal team after the AI-Vision integration; as the designer, I owned the discovery model, filter, and PDP flow, not the AI engineering or the measurement setup. The 74% reflects the team's reported lift.

a.
Action Monitored key conversion indicators and interaction depth following the AI-vision integration.
b.
Result The team reported a 74% relative uplift in the ATC rate, and a 32% growth in search conversion, with product exposure up 2.4x. Together these indicate that an intent-based discovery engine is a meaningful driver for large-scale e-commerce growth.
a.
Action Monitored key conversion indicators and interaction depth following the AI-vision integration.
b.
Result The team reported a 74% relative uplift in the ATC rate, and a 32% growth in search conversion, with product exposure up 2.4x. Together these indicate that an intent-based discovery engine is a meaningful driver for large-scale e-commerce growth.