Applying Generative AI: Creating an Emotionally Resonant Experience to Boost E-commerce Retention
Product Design, Jun-Aug 2023. Protected under NDA.
Yahoo Shopping wanted to improve user retention. The team had already done a lot to create a smooth shopping experience, but that’s become more of a basic expectation these days. With so many external distractions and temptations pulling users away, we started thinking—could generative AI help us keep users coming back?
Outcome
We came up with a playful and continuous personalized recommendation system powered by generative AI, shipped at the end of 2023.
-> Jump to: The Design Outcome
-> Let's Check: The Online Version
Contribution
I was the only designer in the project (from the department of User-Centered Design), working closely with the PM and two engineers from the Data team.
-> Jump to: My Approaches
AI generated scenes
The Initial Ideas
The data team provides two ideas to me in the beginning. I was questioning if those ideas could boost up the retention rate effectively, even they sounds interesting though.
Personality Quiz / Fun Facts Page: Provided a playful experience but lacked clear ties to actual shopping objectives of users.
Chatbot: Proposed to help users find products instantly. Faster checkouts can end sessions sooner—opposite of the retention goal.
One of the data team's early ideas was inspired by Spotify Wrapped.
Insights to Action Items
I dug into user trend reports, and spent time chatting with friends, colleagues, and internal users. Even without a research budget and under tight time constraints, I still uncover some insights that opened up design opportunities for this project.
First, define the task by discovering the topic, then identify the target audience to personalize features for them.
I can only show the process notes and drafts from my research and thinking, due to confidentiality agreements.
One key idea emerged
My responsibility as the only designer in the team
The Design
The design needed to center around personalized shopping experiences that felt immersive—like stepping into a different world.
Telling the story for immersive experience
The goal was to build something engaging enough to pull users in, spark curiosity, and keep the conversation going, ultimately driving retention.
The trigger element
The strategy is based on the research: TWO design opportunities were identified, which led to THREE design strategies, and resulted in NINE designs using gamification.
The whole picture of the design
Link various activities to 'sharing', such as landing counts and passport stamps.
Earn stamps based on your actions. Your passport changes appearance in different phases.
Use 'tags' and 'user behavior' to reveal numerous secrets between users and planets.
We use the Leaderboard and Countdown timer to keep the social buzz alive for this shopping scene.
What you see here is just some samples of my design.
Get in touch if you want to know more.
How To Come Up With The Idea?
Inspiration From Workshops
I facilitated two kinds of workshops to brainstorm ideas from the different perspectives- data and life. The results did not become the final solution, instead, they inspired me a lot by creating the final one!
Brainstorm Storylines
I tried countless combination of Feature-set, such as product data, user data, behavioral data, etc. In parallel, I crafting the concept from information architecture and content structure to make it be a attractive storytelling and shopping immersion.
Craft concept more and more (PT=product tags; UR=user tags)
From Concept To Proposal
I proposed several concepts to the team, then narrowed them down to the three all together. The common between the three is about:
Link to the user: The user persona will be labeled and visualized as part of the product concept. It is where the "Emotionally Resonant" comes from.
Recommend relevant goods: Products will be recommended based on user relevance, as the main purpose of users on Yahoo! Shopping is to browse and make purchases.
This new design concept was inspired by a Sci-Fi Interactive Design class I took back then.
All of three concepts offer users an unreal shopping scenario, where the 'tiger' representing the platform invites users into a new scene – a world built by Generative AI.
Concept Proposals
I set up the criteria for voting that align with the research insights, like attractive, sustainable, feasible, to push the progress of the project without concerned.
Proposal comparison
Deepen UX with Data & AI Model
After defining the needs and features, I moved on to designing product scenarios and outputs, such as I listed the required assets and marked whether they were AI-generated or manually made beforehand. More approaches are below.
My AIUX Steps
Data Flow & Machanism
Each user's retention rate affects their generated results, since the product builds personalized shopping scenarios from accumulated data. I explored different data scenarios to help the team design a smart calculation system that keeps the experience smooth and enjoyable.
Data Visualization for discussion: Retention rate for each person matters the generated result.
Experiments for Prompting
Even with the same goal, different prompts often led to unexpected and fun results. To keep outputs consistent and safe, I worked with engineers—sharing sample prompts and results to guide the AI training and large-scale generation.
Experiments for prompting: from raw data, labeling, to generative AI
The Outcome
The Design Shipped in The Late 2023 🥳
The shipped version was slightly adjusted due to time and cost constraints, but it stayed true to the core ideas from our team.
The Shipped Version
Learn & Think Further
Input & Output System
The current setup only outputs to users, but collecting user input could help train the model more effectively.AI Product Operations
LLMs advanced faster than expected. I plan to align design early with tech updates to reduce future operational effort and cost.AI Product ROI
At the time, costs (e.g. API usage, operations) and rewards weren’t fully measured. Potential gains include higher retention, better conversion, and increased revenue for Yahoo! Shopping.
I love how AI transforms certainty into creative uncertainty. It gives designers more room to explore, shape, and craft experiences—beyond just focusing on visuals.
Jump to