Pottery Barn Kids:
Smart Shopping Tool with Machine Learning
PROJECT:
Williams Sonoma, Inc, Pottery Barn Kids, TEK Systems
ROLE:
UX Designer working in a team of two designers, collaborating with design management, project management, and development team.
ABOUT:
The smart shopping tool utilizes algorithms that were developed in-house. This tool is designed to gain insights on shopper preference and give users highly targeted product recommendations. The tool helps shoppers who are browsing across product categories early in funnel .
BRIEF:
Create a whole room shopping experience for clients. Serve users personalized product recommendations by utilizing inspirational brand imagery.
Willams Sonoma, Inc. is known for inspirational brand imagery. This project aimed to help users identify preferences for styles, colors, and product categories by asking users to rate their favorite lifestyle images.
RESEARCH:
We used analytics and qualitative research to look for opportunities to increase “units per transaction”. The business identified living room shoppers as a possible target for a smart shopping tool. Using the data we collected we identified nursery shoppers as a potential target for full room shopping.
Nursery room shopping data:
Crib shoppers are much more likely to purchase additional products, eg. a crib liner + bedding
Customers shopping for glider/rockers are more likely to be shopping for a nursery room.
Registry is great target for a product discovery tool. Registrants are usually early in the shopping funnel.
DESIGN:
WSI machine learning algorithms follow the following logic:
Browse, Browse “you may also like” display the top items that users have browsed in order of most browsed.
Browse, Buy “customers ultimately bought” display the top items that people who browsed this item ended up buying in order of most purchased.
Buy, Browse “customers also viewed” display the items a person who purchased this product also looked at in order of most viewed.
Buy, Buy “you may also need” display the items that people also bought.
INTERACTION:
I created multiple interaction models, experimenting with layouts and on page behaviors. I also explored image styles, dismissal patterns and css micro-interactions. A woking prototype was created in InVision and Framer to experiment with interaction speed and to scroll animations.
LAYOUT:
I worked on multiple layouts with different hypothesis: Larger lifestyle images, with secondary product silo images, smaller scrolling carousels, and image trays.
COLLABORATION WITH DEV:
I worked closely with a remote development team to pilot this project in Zeplin. We wanted to have a handoff-free experience that streamlined communication and made the specification process effortless. We met bi-weekly to validate designs, make UX recommendations, and to view progress.
DEPLOYMENT AND IMPACT
Interest and continued investment in online design tools.
Cross brand interest in use as a template.