Hailey Hsu
Product Designer
Reshaping Shopping Experience
Helping users shop more efficiently across the web
Overview
Realry is a smart shopping platform that helps users find the best deals by aggregating prices across the web. With 1.2 million monthly active users, the platform acts as a price comparison engine that saves shoppers time and money.
My focus was improving the navigation and overall user experience to increase conversion. In this case study, I'll walk through the problem space, my research process, the design solutions I created, and the impact on business metrics and user behavior.
The Problems
Through discovery research and analytics audit, I identified four critical UX issues causing user drop-off:
Landing Page Layout
Users were abandoning the platform before engaging with products due to unclear value proposition and poor information hierarchy.
- • Unclear hero section: AI feature dominated but lacked clear description
- • Misplaced content: AI blog appeared before products
- • Buried products: Core shopping experience hidden below the fold
Unintuitive Search
Users were dropping off during search due to forced AI activation, disrupting their shopping flow and causing confusion.
- • AI and search bar combined into one interface
- • Search keywords triggered AI conversation instead of results
- • Search bar disappeared after exiting AI
Category Navigation
Users couldn't find products through browsing due to confusing, non-standard category structure that led to dead ends.
- • Categories appeared as horizontal buttons (looked like tags)
- • Sub-category pages looked like product listings
- • "Back" button returned to homepage, not parent category
Price Comparison Unclear
The main price comparison feature was buried and invisible to users, undermining the platform's core value.
- • "Do not purchase" recommendation blocked conversions
- • Related products pulled users out of conversion path
- • Price insights hidden, requiring user exploration
Constraints
Timeline Pressure
6 weeks to redesign the entire platform, driven by an upcoming investment pitch deadline.
Multi-Platform
Needed to design both web and iOS app simultaneously. Device distribution was split 50/50.
Design System
Had to create reusable components from the start to ensure consistency across web and mobile.
Design Process
Discovery & Audit
Analytics review
Competitive Analysis
Honey, ShopBack
Design Exploration
Daily syncs w/ CEO
Design System
IBM Carbon base
Handoff
Dev collaboration
A/B Testing
Iteration
Solutions
Redesigned Homepage Hierarchy
Moved AI feature from hero to floating animation (fixed bottom-right) to maximize partner exposure and grab attention without blocking content.
Created themed collections (Popular Items, Not Sure Where to Start, Trending Now, Top Deals) vs. single "Top Sales" list—diversifying entry points for different user intents.
Enhanced Brand Visibility
Replaced single-product brand displays with bento box layout showcasing multiple brands with bigger space simultaneously.
Decoupled Search from AI
Decoupled search input from AI chat interface: users can now search for products without triggering unwanted AI conversations.
Created persistent search with dropdown suggestions that include product previews and store recommendations—turning a friction point into a revenue opportunity.
Persistent Menu with Clear Hierarchy
Replaced ambiguous horizontal buttons with a persistent top nav menu that opens an overlay, making it visually clear users are in a navigation menu, not clicking tags.
Added Stores Directory alongside product categories—giving users two ways to shop (by product type or by preferred store).
Transparent Price Insights
Elevated price comparison visibility with clear "Best Price" badges and list all the available options.
Reframed recommendations from prescriptive commands ("don't purchase") to informative insights ("Price higher than average" / "Price lower than average"), empowering users to decide rather than blocking them.
Results
A/B testing over 3 weeks
What I Learned
Efficiency > Engagement
While click-out rate went up, average session duration actually dropped by 15%. For an aggregator, less time on site means we helped users find what they want faster. This is a success, not a failure.
Avoid Over-Engineering
I spent a lot of time designing a complex Advanced Filter (by material, sleeve length, etc.). However, heatmaps showed less than 4% of users touched it. Most users only used "Sort by Price." Next time, I'll validate feature necessity with an MVP first.
Solve the "Dead End"
The new design drove more traffic to popular items (high click-out rate). However, we noticed a high "Bounce Back" rate—users clicking out but coming back within 5 seconds. High clicks mean nothing if items are out-of-stock. We need to sync real-time inventory to manage user expectations.