👀 Project Overview
Client: Upsy Shopping
Timeline: 2022 – 2023
Role: Product Designer (Lead)
Challenge: Transform an overlooked AI widget into a revenue-generating shopping assistant that customers actually engage with
Tasks:
-Redesign Widget for Desktop and Mobile
-Redesign Admin Dashboard and Analaytics
-Redesign Marketing Website
Impact At A Glance
- 47% increase in widget usage
- 31.8% growth in assisted sales
- 84.6% positive user feedback post-implementation
- 20% boost in website conversion rate
📊 The Business challenge
UPSY had created an AI shopping assistant that wasn’t living up to its potential. As the lead designer, I needed to solve critical issues:
- Identity crisis: Users mistook it for a standard chatbot and ignored it
- Poor discoverability: The widget failed to attract attention despite valuable functionality
- Dashboard confusion: Text-heavy analytics made data inaccessible to clients
- Brand inconsistency: The widget felt like a third-party add-on rather than an integrated tool
🔍 My Discovery Process
Research Methods I used
- Heuristic UX analysis overall product areas; widget, dashboard and website itself.
- Competitor analysis across 12 e-commerce tools
- User interviews with 8 shoppers and 5 store owners
- Analytics review of engagement patterns
- Stakeholder workshops to align business goals
Main Problems with Upsy Widget:
- Poor Visual Hierarchy – Cluttered layout with competing elements and inconsistent styling makes it difficult for users to identify the most important information or actions
- Confusing Interface Elements – Redundant “Show” buttons for each product and ambiguous search functionality with unclear voice input option
- Limited Filtering Capabilities – Basic sorting options (only Alphabetical and Lowest Price) without category, size, color or other essential e-commerce filters
- Inefficient Space Utilization – Too much vertical scrolling required with poor information density and unnecessary repetition of “At hand” status
- Disconnected Visual Identity – Widget appears tacked-on rather than integrated with the main site design, creating a jarring user experience
- Unclear Value Proposition – “UPSY SHOPPING HELPER™” branding doesn’t communicate its purpose or benefits to users, leaving them uncertain about how it enhances their shopping experience.
Main problems with Upsy Dashboard:
- Cluttered Data Visualization – Multiple graphs and metrics are presented without clear hierarchy or relationship, making it difficult to quickly extract meaningful insights
- Inconsistent Design Elements – Tabs, buttons, and data cards vary in style and presentation, creating a visually disjointed experience
- Poor Information Architecture – The multi-step configuration process lacks clear progress indicators and contextual help for complex technical requirements, confirmation steps long and complicated, users need additional help for setting up their account.
- Limited Data Context – Key metrics like conversion rates and sales figures are displayed without benchmarks, trends over time, or explanatory context to understand their significance
- Overwhelming Technical Configuration – The WooCommerce setup screens require technical knowledge (REST API keys) without sufficient guidance, potentially intimidating non-technical users
- Inadequate Mobile Optimization – Dashboard elements appear designed primarily for desktop viewing, with dense information that would be difficult to navigate on smaller screen.
Critical Finding #1
Users were skipping the widget because it visually resembled support chatbots they'd had negative experiences with.
Users were skipping the widget because it visually resembled support chatbots they'd had negative experiences with.
Critical Finding #2
Store owners couldn't justify the tool's cost because they couldn't easily interpret the dashboard metrics.
Store owners couldn't justify the tool's cost because they couldn't easily interpret the dashboard metrics.
Critical Finding #3
The widget's limited recommendation features didn't showcase the AI's true capabilities.
The widget's limited recommendation features didn't showcase the AI's true capabilities.
💡 Design Decisions & Problem Solving
Challenge: Making the Widget Discoverable Without Being Intrusive
My Approach:
I redesigned the entry point to signal "shopping assistant" rather than "support tool" by:
- Creating contextual triggers based on shopping behavior
- Developing a visual language distinct from standard chatbots
- Designing subtle animations that caught attention without annoying users
- Making the widget more interactive & verbal, and giving options to user to close it easily
Challenge: Transforming a Complex Dashboard into an Actionable Tool
My Approach:
The original dashboard was text-heavy and difficult to navigate. I:
- Prioritized metrics that directly connected to revenue
- Created visual hierarchy highlighting actionable insights
- Designed comparison views showing performance improvements
- Simplified the onboarding process from 12 steps to 3
Key Decision Point: I challenged stakeholder assumptions by demonstrating that most clients only used 4 of the 15 metrics regularly. This led to a complete dashboard restructuring focused on these core metrics.
🛠️ The Solution & Implementation
Widget Redesign
Key Features I Designed:
- Product category search with visual browsing
- Daily deals spotlight with carousel interaction
- New arrivals feed with custom relevance algorithm
- Feedback collection with one-click rating
My Process:
- Created rapid wireframe variations (3 distinct approaches)
- Prototyped the most promising direction in high fidelity
- Conducted A/B testing with 3 online webshop over 1000 customers
- Refined based on feedback, focusing on simplicity
Dashboard Transformation
Key Features I Designed:
- At-a-glance conversion metrics
- Visual representation of customer journeys
- Customizable reporting modules
- One-click implementation of AI suggestions
Website Redesign
Key Features I Designed:
- Implemented comprehensive marketing website redesign
- Enhanced product descriptions, views, and content flow structure
- Achieved 20% increase in conversion rate following implementation
- Created more streamlined, user-friendly customer experience
- Generated increased user engagement metrics
- Established ongoing performance monitoring and continuous optimization process
📈 Results & Impact
User Metrics
Widget usage increased by 47% Assisted sales grew by 31.8% 84.6% positive user feedback Website conversion rate improved by 20%
Business Impact
Reduced customer service inquiries by 28%
Increased average order value by 17%
Improved product discovery for long-tail inventory
Enhanced client retention through measurable ROI
Learnings & Next Steps
The most valuable insight from this project was discovering how subtle visual cues dramatically impact user perception of AI tools. For future iterations, I'm exploring:
- Personalization options based on customer behavior patterns
- Enhanced visualization of product recommendations
- Mobile-first approach for the next dashboard version
🔄 My Role & Contribution
As the lead designer on this project, I:
- Owned the end-to-end redesign process
- Facilitated stakeholder alignment through workshops
- Created all wireframes, prototypes and final designs
- Collaborated closely with developers on implementation
- Conducted user testing and incorporated feedback
- Presented solutions directly to client stakeholders
Would you like to discuss your project with me?
Let's talk about your project and see how I can help you to achieve the best results.