Work 03 of 03 Product Strategy · Product Design · Experimentation

Property Finder: Rebuilding search, app to web

Search was the product. It just wasn't acting like it.

Client: Property Finder
Role: Product StrategyProduct DesignExperimentation
Sector: Real Estate & PropTech
Engagement: Embedded product leadership
Delivered: Delivered by Rai's founders, while at Property Finder · 2025–2026

The brief

How might we make the results page better at helping people decide, without touching the ranked order advertisers pay for?

The problem.

01Discovery

What we observed

Every session runs through search results, and the surface had grown cluttered. The hierarchy was weak, too many signals competed for attention, and the details people needed to compare sat below the fold where most never scrolled. People couldn't scan, couldn't compare, couldn't quite trust what they were looking at.

The root cause

The space being sold to advertisers is the same space a buyer needs in order to decide. Years of adding placements and signals had crowded out the information that actually drives a decision.

Problem statement

How might we rebuild the results page around deciding, without reordering the ranked list advertisers pay for?

What we learned.

02Research

The page was costing both sides of the marketplace.

So what

The friction wasn't just costing leads. A results page that converts poorly is worth less to the agents paying to be on it, which made the fix a commercial case, not just a design one.

Price and title are what people scan for first.

So what

The hierarchy had to lead with them, put trust and quality cues on the card itself, and cut everything that wasn't helping someone decide.

One pattern had to carry every listing type.

So what

Consistent interaction logic across listing types means the page scales instead of fragmenting, and a new type slots into the existing pattern instead of arriving with its own.

The call.

03Direction

Leave the monetisation layer alone

Work around the placements and improve what sits between them. The safe move, and it would have left the actual problem in place.

B Chosen

Cut the low-performing placements

Give the space to what helps people decide, and make the case commercially: we could show the conversion uplift was worth more than the placements it cost.

Why we chose it

The maths made the argument. A page built for decisions converts better, and the uplift was worth more than the space we removed.

What shipped.

04Build

01 Feature

Real hierarchy on every card

Price and title lead. The trust and quality cues that tell you whether a listing is worth opening sit on the card itself, rather than one tap further in. Everything that wasn't helping someone decide came off.

SolvesCards where the deciding information sat below the fold or a tap away.

02 Feature

One pattern, every listing type

The same interaction logic runs across every listing type, so a new type slots into the existing pattern instead of arriving with its own.

SolvesA page fragmenting as inventory types multiplied.

03 Feature

Measured before scaled

A controlled A/B test on app first, where session volume returns an answer fastest, then carried to web with the same logic, adapted to how people search on desktop.

SolvesScaling a redesign on opinion rather than evidence.

The impact.

05Outcome

The redesign was A/B validated on app and again on web, two platforms where people search differently.

+1.8pp conversion rate lift, A/B validated on app and web
~500k additional leads a year
2 platforms, proven on both: app first, then web

Reflection

The space being sold and the space for deciding were the same space. The work was making one page do both.

Product

Property Finder: Making the map the search engine

Rethought how buyers browse listings on the map. Map-session leads rose from 5% to 32%.

View case study View all work