Rebuilding Discovery

Buying is a months-long decision. People arrive at every stage of it. Some already know the community they want; others are comparing cities, climates, prices, amenities, and hundreds of communities across several states.

Roughly 1,700 pages carried that traffic. Geographic destination pages were the primary bridge between broad exploration and the detailed community information that made 55places valuable — one for every state, area, and city the company covered. They were among the site’s most important organic-search landing pages, and more than one in five sessions touched one. That structure was already in place when I joined 55places in the fall of 2024 — an information architecture closer to a publishing site than a real estate marketplace. A user might start on a state page, choose an area, navigate to a city, scroll through editorial content, then open communities one at a time.

A foundational study, run end to end. The problem came out of research conducted across September and October 2024 — the first large-scale qualitative user study the company had run, and one I designed, managed, and executed myself. It was scoped to three questions: what users want to see and do on 55places, where navigation and usability break down, and what pushes people out of the ecosystem mid-search. Fifteen one-hour semi-structured interviews — eight current users, three past, four prospective; nine on desktop, six on mobile — each dropping the participant onto a live page and following their real search behavior before testing concepts against it. The study spanned five page types and five hypothesis areas across the site; the destination-page findings are the slice that drove this redesign.

The tension it surfaced. Users saw 55places as a trusted, uniquely valuable resource for researching 55+ communities, but when they were ready to narrow their options, the experience felt far more manual than Zillow and Realtor.com, the products setting their expectations. Four core problems stood out:

• Search required too much manual work. Users began with specific criteria — location, home type, price, pet policies, accessibility, proximity to healthcare — but couldn't apply them across the experience. They opened communities one by one, only to discover an option didn't meet a key requirement.
• Geography was critical, but the map was hard to use. Location was one of the biggest decision factors, yet the map sat apart from the results and felt unresponsive. Some users left for Google Maps to understand proximity — “I would just Google Map it… this way I know how far it is to the airport.”
• The page structure interrupted exploration. On desktop, a lead form appeared after the first two results — many participants, new and returning alike, read it as the end of the list and assumed only two communities existed. On mobile, filters were easy to miss and long content delayed the results.
• The product stopped at communities. Users could research communities but couldn't browse the homes within a destination, pushing them to Zillow or Realtor.com to continue.

“This just seems like you have to pick a place and then look at it, and then pick another place and look at it… it is kind of daunting, especially for places like Arizona and Florida where there’s like a gazillion different communities.”— past user, desktop

Reorganized around the real task. I built the experience around what users were actually trying to do: evaluate the available options in a location. Search and inventory became the primary experience; educational and market content stayed available, but no longer stood between users and the results they came to explore. Four decisions did most of the work.

1. Communities and homes in one journey. For the first time on 55places, users could browse both communities and individual homes from a destination. A prominent Communities–Homes toggle supported two mental models — communities for comparing lifestyle and fit, homes for evaluating active inventory.

2. A persistent, synchronized map. The map became an interactive part of the search surface rather than a separate destination. On desktop, users could choose list, split, or full-map views, with cards and markers synced so interacting with one clarified the other. The map also revealed detail progressively by zoom — named community markers at broad views, then homes and price markers up close — reducing clutter while keeping detail when it mattered.

3. Filtering at the point of decision. A persistent filtering system let users refine communities or homes without leaving the page. On mobile, filters opened in a dedicated full-screen flow with an explicit result count and apply action.

4. A lead form that no longer ended the list. The form that had convinced people there were only two communities became an inline card, mixed into the results every six cards and opening into a modal when tapped. It stopped reading as the bottom of the page, and as a form it became both easier to find and easier to use.

Working under real constraints. The redesign touched search, geographic data, inventory, content systems, responsive layouts, lead generation, analytics, and SEO. When frontend capacity tightened, I worked with product and engineering to protect the core discovery journey — deferring or simplifying lower-priority content modules so the team could concentrate on maps, cards, filters, views, and responsive behavior.

Validating the model. In January 2025 I ran four moderated prototype sessions on Lookback — three desktop, one mobile — with product and executive stakeholders observing live and feeding questions into the room. Participants consistently described the new experience as more intuitive, more visually appealing, and easier to navigate than the existing site, and the sessions surfaced a detailed punch list I worked through before build. Not every request became a change — several were worth more as documented decisions than as features.

What shipped. The redesigned experience launched across desktop and mobile on April 28, 2025. For the first time, users could explore communities and active homes together, refine without leaving the page, compare results against a responsive map, save options, and move fluidly between broad geographic exploration and specific inventory.

How it was measured. The data team compared equal 15-day periods before and after launch for sessions that included a destination-page visit. Because the after period had more traffic, the most defensible comparisons normalize by session volume:

What it unlocked. The old model bound one page to one location — each destination a separate, largely static document, and a ceiling on what search could ever become. Rebuilding discovery as a dynamic, filterable surface removed that constraint: a results page no longer has to correspond to a place. It can be composed from any combination of geography, inventory type, price, and amenities, which is what makes personalized, intent-specific results pages possible at all. 55places can now assemble a page around the query itself — whether the visitor arrives through on-site search, a search engine, or an LLM-driven assistant. That is no longer hypothetical: the same surface now powers dynamic Explore pages generated from arbitrary search queries, with geography matched to the map and list, keywords mapped onto the filter set and pre-applied, and the Communities–Homes toggle set to whichever the query implied. The same structure that made results easier for people to explore made them far more legible to crawlers and answer engines, strengthening organic reach alongside the in-session gains above.

The structural change underneath the redesign — hundreds of fixed pages replaced by one page composed from the query

What I took from it. This was my third project after joining 55places, and the first time I argued for replacing a core model rather than improving the pages built on top of it. What won the argument wasn’t a critique of the old structure — it was showing what the business could do once a results page no longer had to be a place.