A Three-Question Quiz That Replaces a Salesperson
A mattress and chair catalog that's easy to browse and nearly impossible to choose from without help, the kind of decision customers used to make in a store.
Context
The Sleep Company’s catalog spans a wide range of mattresses and chairs, differing on firmness, material, and body-type fit — the kind of decision customers historically made in a physical store with a salesperson’s help.
The problem
Online, that same decision becomes a wall of product listings and spec sheets. Customers without a strong prior preference either guess, bounce, or order the wrong product and return it later, which costs more than the sale was worth.
The constraint
The tool needed to feel like guided advice, not a lead-gen quiz bolted onto the site for marketing’s sake, and it needed to work for two structurally different product categories, mattresses and chairs, without turning into two entirely separate builds to maintain.
The approach
I built a three-level quiz app where each answer dynamically filters the live product set, rather than mapping to a static, pre-written result page. Sleep style and body type narrow the field first, comfort preference narrows it again, and the final step surfaces the best-matching product from whatever remains. The same underlying app powers both the mattress finder and the chair finder, with the filtering criteria treated as configuration rather than duplicated logic.
The alternative I rejected was a shorter, single-question “quiz” that maps answers directly to a fixed set of result pages. It’s simpler to build and it’s also just a disguised category page — it doesn’t actually narrow anything based on the live catalog, and it breaks the first time inventory changes.
The result
A three-question path from the full catalog to one recommended product, cutting the decision down to the size customers can actually hold in their head, on the two categories where indecision costs the most.
What I’d do differently
I’d log which filter combinations most often lead to “no strong match” earlier than I did. That signal isn’t really a quiz problem, it’s a catalog gap, and it’s merchandising input the quiz surfaces almost for free once you’re paying attention to it.