De-risking release with lean experiments
I led end-to-end UX on a new feature enabling buyer-property matching by capturing investment preferences, de-risking delivery with a lean rollout and low-cost experiments, driving 63% of new buyers to create profiles.
Role
Led UX within cross-functional squad
Team
Squad, UX Researcher, UI Designer, Sales, CEO
Challenge
Crowdhouse sold high-value investment properties, some ‘off-market’ and not listed on the online platform. Sales were identifying likely buyers manually, making targeted outreach reliant on individual knowledge and ad hoc notes rather than systemised data.
The goal was to capture buyer investment preferences online, so sales could target potential buyers more systematically.
The longer-term vision was automated property matching, allowing the business to sell more effectively and scalably.
Impact
Smarter, more targeted sales
Structured buyer data enabled sales team to prioritise outreach and prepare for calls with real intent signals.
63%
new users shared buying preferences
Reduced delivery risk
Lean experimentation cycles proved user completion appetite before any significant build investment.
Foundation for automated matching
Profile data laid the groundwork for matching buyers to properties automatically.
A short flow to capture preferences, displayed as part of the onboarding flow, enabled us to systematically store this data as part of a user 'profile'.
More extensive investing preferences could be managed within an easily accessible 'profile' area. This allowed potential buyers to edit and add preferences without overwhelming them with questions during onboarding.
Preference data could be submitted online or captured via scheduled sales calls and input via our sales team. Data was synced between all systems to ensure everyone had the latest information.
We ensured that buyer preferences were captured in a format that made matching to available property metadata possible, allowing us to initially run reports to find top matching buyers for each property and — later — to move forward with automated recommendations.
60+ in-depth datapoints reduced to most valuable sub-set.
Stakeholders had mapped out a huge number of desired buyer datapoints to be added to our CRM system and initially expected the customer-facing platform to mirror this.
Replication would have meant a large development effort and huge form that users were unlikely to complete.
Prioritised categories:
Investment background
Understanding prior investment experience helped sales gauge appropriate risk & property profiles.
Budget & strategy
Allowed us to match sole vs co-ownership models and assess appropriate mortgage packages.
Property preferences
Buyers often have specific needs around location, condition, usage etc that could be dealbreakers if not met.
Considering inputs, tone & motivation.
I had to consider how we might motivate potential buyers to share the information we wanted. Overall we needed input to feel quick, easy and logical.
Simple inputs for easy selection
Single or multiple choice inputs avoided the form feeling laborious to complete and worked well across mobile and desktop, with clear labels for clarity.
Leading with the benefits
To help users understand why completion was worth their time and why each piece of information was valuable.
Conversational copy
For a welcoming, friendly feel, mimicing more traditional conversations with a personal advisor.
Chunked by category
Logical sections gave the flow structure, with optional skip points to avoid users feeling trapped while encouraging completion.
Starting with quick low-effort email 'MVP'.
Rather than us building out a full solution upfront without evidence it would be used, I led the team through an assumption-mapping exercise to identify what needed to be true for this to succeed.
Key assumptions:
We planned to first test our question flow by quickly building it using no-code tool 'Typeform' and sending to a small buyer group via email. Integration with the online platform and data syncing would come later, after we'd built confidence in our solution.
Results & next steps
We proceeded to a full build, with data fully synced between our database and sales' CRM system. Our UI Designer and I had wanted to try a chat-like UI, but had to compromise for build efficiency by making use of existing flow components.
Usability testing & data to guide streamlining.
I worked with our UX Researcher to build a feedback request into our first test question flow, then to plan usability testing sessions to gauge property investors' reactions. We also measured question completion rates.
Though we had tried to make completion quick and easy as possible, early feedback gave us signals that the flow was still too long.
We decided to cut down the flow that appeared as part of buyer onboarding, not wanting to hit users with an exhausting introduction to the platform.
v1: Initial pilot
Even though categories and questions could be skipped, on reflection we felt the flow was too long as part of onboarding, when users want to quickly explore the platform.
v2: Cutting investment experience questions
We decided to separate highest value preference questions — displayed during onboarding — from the rest that could be housed in a central 'buyer profile' area. In discussion with stakeholders, we agreed that the investment background data was less vital for property matching and could be gathered during welcome phone calls.
v3: Further streamlining
We then further reduced the flow to remove separate category introduction pages, which at this point felt redundant.
Using data to guide future feature iterations.
Working with a developer, I identified which datapoints could realistically be tracked and what they would tell us. After agreeing the most valuable metrics as a team, we worked together to get a live metrics dashboard running in our squad area.
We knew that sales planned a welcome call with new platform users within 7 days of registration and found it beneficial to have their preference data on-hand, so we included this as a target to measure.
I then ran bi-weekly review cycles to spot patterns, anomalies and potential areas for improvement (see Proactive feature performance analysis).
I also used the 'Opportunity Solution Tree' framework to connect ongoing improvement ideas to desired outcomes and metrics. This kept us outcome-focussed and aided prioritisation as we progressed beyond initial roll-out.
We were deliberate about the distinction between completion rates and data quality: a high completion rate with inaccurate or vague answers would undermine the value of the feature.
Benefits of squad autonomy
Our squad thrived when given freedom in how to approach a topic and time to focus on it collectively. We chose a smart, experimental approach and our metrics dashboard kept us motivated.
Visual trade-offs
Our quick no-code Typeform pilot used illustrations and had a warmth that our plainer component-based build didn't quite match. Though we later added more illustrations to bring things to life, we never managed to prioritise the slicker, chat-like UI we envisioned at the start and don't know whether this affected completion rates.
Complexities of measuring impact through to sales
In a business built on infrequent, high-value purchases, tracking outcomes through to actual property sales was difficult. We focussed on measuring profile completion but would have liked to measure the impact this data had on speed of sales or buyer satisfaction.
More featured work
Crowdhouse
Addressing data anomalies with quick-win improvements.
I drove more accurate property recommendations through quick design changes that significantly increased granularity of submitted location preferences.
Open Technology Fund
Research-led redesign of a bloated staff portal.
I led user research with customers and internal staff of a complex funding application portal, identifying critical workflow issues, forming a structure and design for a simpler, streamlined platform that eliminated unnecessary tool switching.
