Faceted Search — Finding the Right Doctor

The faculty database on hopkinsmedicine.org is one of the site's most valuable features. It is structured with multiple facets per speciality, but the site was not presenting that functionality to users — meaning a powerful search tool was effectively invisible. The goal was to find the most effective way to surface those facets and test two competing design approaches before committing to either one.

The Problem Users needed to narrow a large database of physicians by speciality, location, and other attributes. The question was not whether to offer faceted search — the decision to add it had already been made — but which design approach would actually help users find what they needed fastest.

Key Performance Indicators Success would result in high scores on perceived ease of use, high pass rates on counter-balance task testing, and measurably improved results in round two testing after design iteration.

How I Tested It Two designs were tested against each other: open facets and closed facets, on both mobile and desktop. Testing used a counter-balance method to remove bias, with two testers per session — one administering, one observing — and two distinct rounds designed to iterate on findings from the first.

Round one tested the core question: open vs. closed. Round two presented only the winning design with targeted improvements based on what round one revealed.

What We Found Open and closed facets performed similarly in task completion, but users showed a strong preference for closed facets. Desktop versions were easier to use than mobile due to better affordance. Users consistently tried to scan results before engaging facets — meaning search results themselves affected how much users trusted the tool.

Round two improvements included changing "narrow your results" to "refine your results," improving the affordance of the refine call to action on mobile, and visually differentiating the refine title from the interactive facets themselves.

Outcome Both mobile and desktop presentations saw significantly improved results in round two. Average ease of use scores improved across the board. The iterative two-round structure proved its value — the first round surfaced the real problems, and the second round solved them.