Restructuring a 10,000-page health system
A sprawling hospital network reorganized around how patients actually search for and choose care, not the org chart that built it.
- Role
- Experience Lead
- Engagement
- Digital ecosystem redesign
- Disciplines
- IA · Content Strategy · Design System
- Timeline
- 2023–2024
The challenge
The system had grown by acquisition. Dozens of hospitals, clinics, and service lines had each arrived with their own pages, their own naming, their own idea of what a navigation should be. Patients didn’t experience that history as a story; they experienced it as a maze. Content mirrored internal departments rather than the questions people actually arrive with.
The numbers told the same story. Roughly 10,000 pages, no shared model for how any of them related, and a search for something as common as “knee replacement” surfaced six competing pages with no clear winner. The cost wasn’t cosmetic. People abandoned care journeys, called instead of booking, or quietly left for a competitor whose site simply answered the question.
The approach
An information architecture is a hypothesis about how people think. The work wasn’t to redraw the sitemap. It was to find the model the 10,000 pages were missing, then make every page accountable to it.
Map the territory before redrawing it
Inventoried every page and scored it against traffic, search demand, and clinical priority. Interviewed twenty-plus stakeholders and ran tree tests with real patients. The finding that reframed the project: 60% of pages drew under 1% of traffic, and people navigated by condition and body, not by the department that owned the care.
One model, many doors
Rebuilt the structure around the path patients actually take: condition, then care, then location. Service lines became a connective layer rather than the front door. Defined a page-type system so every page had a job and a shape, and collapsed duplicates into single authoritative pages.
Migrate without breaking trust
Sequenced migration by service line with redirects, a governance model, and CMS training so the structure would hold after launch. A lightweight component system let content teams assemble pages from shared parts instead of reinventing layout every time.
“For the first time, our site reflects how patients actually look for care, not how we happen to be organized internally.”
What changed
Navigation went from department-first to patient-first. A dozen competing entry points collapsed into one clear path to care. Ten thousand pages became roughly thirty-five hundred authoritative ones, each with a defined type and a named owner.
Most importantly, the structure could now defend itself. Governance and a shared component set meant the architecture wouldn’t re-sprawl the day after launch: the failure mode that quietly undoes most redesigns.
The outcome
Twelve months after launch, the model held, and the metrics that matter to a health system moved with it.
- −65%
- Redundant pages removed
- +40%
- Online booking starts (organic)
- 2.4×
- Tree-test task success
What I’d carry forward
- 01
Architecture is a leadership problem, not a sitemap. The hard part was aligning service-line owners around one model; the diagram was the easy part.
- 02
Reduce before you redesign. Cutting 6,500 low-value pages did more for findability than any visual change could have.
- 03
Build the system that maintains the system. Governance and a component kit are what keep an information architecture from re-sprawling the moment it ships.