Appari is a gen AI try-on app that helps shoppers visualise clothing on themselves before buying. I led the end-to-end design of the MVP from user research and strategy through to a live product with 250+ early users and two retail partners signed for pilot testing.
The goal was to reduce purchase uncertainty, lower return rates, and give users confidence in their online fashion choices through a simple, mobile-first experience.
Role
UX Visionary
Link
appari.ai
Medium
Mobile App & Website
Tools
Figma, VS Code, ChatGPT, Claude, Vue.Js, Tailwind
Shopping for clothes online is convenient until it isn't. Appari set out to solve common, frustrating problems shoppers face:
Issues
This leads to customer frustration, lower purchase confidence, and costly return rates for retailers.
The challenge: How might we help users see how clothing will look on them before they buy - simply, quickly, and visually?
The goal was to build an MVP that would allow users to upload a selfie and try on clothing virtually using AI. It needed to be intuitive, frictionless, and visually compelling — without requiring complex onboarding or body scanning.
I conducted research with 340+ participants through interviews and surveys, uncovering key insights:
"I usually buy two sizes, try them on, and return one. It's wasteful"
"If I could see how it fits before buying, I'd shop online more often"
"I want to try the dress before ordering"
I analysed existing solutions including Zyler, True Fit, and various AR try-on tools that had been made.
Key gaps identified:
Design a mobile-first, cross-platform solution that works with any retailer.
I prioritised features for the initial launch based on user needs and technical feasibility:
Must-have (Phase 1):
Primary flow:
To create the most accurate digital twin.
a - Scan face & body with camera or upload images
b - Input body measurements for accuracy
Discussion with potential users:
- Too long winded, too many steps. Needs to be simpler
a - Option to input link or screenshot/image of the items.
I added images/screenshots as clothes can be captured from influencers and images friends share.
b - Clothing is remodelled on the user's digital twin.
Feedback:
This is simple and straight forward.
I decided for the initial concept, we should see if people actually can integrate this type of tool into their shopping flow as it's a new way of doing this. I decided to simply the AI model to just a full body selfie as Gen AI is able to work with that.
a - Simple onboarding with Google login
b - Single full body image with instructions for best output.
c - Users can take a photo or upload something from their camera roll
c - Removed url paste for simplicity, added instructions for garments so the gen AI can be more accurate
d - A preview of the outfit is shown so user can change it if they selected the wrong image
e - Garment type is put in so more accurate results can be achived
f - Final output screen
Finding: Users were hesitant to upload personal photos immediately
Solution: Added demo models and outfits to explore first
Finding: AI image generation did not perform to good standard
Solution: Added 'regenerate' option without charging an additional credit
Finding: On first use, people didn't have clothes to try
Solution: Added a gallery of clothing to test with
Finding: Credits were too expensive
Solution: Reduced pricing and increased number of free credits
Goal: People can look at their saved clothing for when they make buying decisions, write notes and share them with others.
Goal: User is able to do basic account tasks and look at their outfit credits, buy more and refer friends.
Based on user feedback and research, the next design priorities include:
Long-term vision: Create an experience where people can visualise themselves in any outfit from any retailer, building confidence and reducing the friction of online fashion shopping.