Furniture websites often organize products around retailer categories and inventory systems. But people do not always begin with a product name. They may begin with a room, a problem they are solving, or the kind of space they want to create.
I wanted to understand how people naturally organized furniture and home goods—and whether existing e-commerce structures reflected those mental models.

Before sorting, participants described shopping primarily by product type. During the exercise, many organized foundational furniture around rooms. Rugs, mirrors, lighting, and wall art behaved differently, appearing across both room and product groupings.

That contradiction became the center of the project

I reviewed Wayfair, Crate & Barrel, and Burrow to identify common navigation patterns, familiar category language, and points of friction. I then conducted seven card-sorting sessions—six virtual and one in person—supported by pre- and post-session questions.

This let me compare what the market taught users to expect with how participants organized the information for themselves.

I proposed primary paths for Shop by Room and Shop by Product, with flexible décor cross-listed where people expected to find it. The structure preserves familiar retail language while better matching the contexts people use to make decisions.

The result was a recommended information architecture—not a completed redesign. The research supported a dual-path browsing model and clarified where universal décor needed flexible placement.

The next step would be to prototype representative shopping tasks and test whether people can predict where an item belongs, move between room and product paths without losing context, and understand cross-listed décor without confusion.

WooMe reinforced a lesson I return to often: people may describe one habit and demonstrate another. Good research creates room for both—and good design makes the resulting choice easier to understand.
The important decision was not choosing room-first navigation over product-first navigation. It was recognizing that both models were valid in different moments and designing a system that could support them together.