Summary
Problem
My role
I joined the project after the initial ideation and stakeholder interviews, once the core concept had been defined. My role was to:
- refine and iterate on the detailed use cases,
- conduct usability testing on high-fidelity designs,
- and incorporate feedback to develop final designs for handover.
I also supported the development process and saw the project through to the launch of the MVP.
Impact
After just 6 weeks, the impact was clear:
Process
Gathering ideas
Through stakeholder interviews and cross-functional workshops, a winning concept emerged: a series of targeted questions that would return a tailored list of recommended car models.
What others are doing
A competitive review was conducted to understand industry trends, identify best practices, and uncover opportunities to differentiate our approach.
Laying the foundation - the AI model
The goal was to generate objective recommendations based on multiple data sources, ensuring results were tailored to each user’s needs—not just another way to filter search results.
To do that, it was key to determine what factors mattered most to car buyers so that the right questions could be asked. While the AI team worked through these complexities, design team shifted focus to other projects, waiting for the right moment to pick things up again.
Preparing for the build
I joined the project when design work resumed after a pause, while the AI model was being developed. My first steps were to:
- review existing concepts, identify gaps, and refine the design approach;
- identify key stakeholder groups using RACI framework;
- align with key teams to ensure consistency and stakeholder buy-in;
- iterate on detailed use cases with a focus on user needs;
- collaborate closely with the Product Manager to keep the team moving towards the same goal.
Through this process, I’ve identified key unknowns that needed te be addressed in designs:
- Is the entry point into the flow clear and intuitive for users?
- Does the number of questions feel right to users—should it be fewer or more?
- How many recommendations should we show at the end?
- What key information do users need to see about each recommended car?
- What page layout best supports users in taking the next step?
To answer these, we turned to user research. I set up and conducted 6 moderated user interviews and ran a short survey to quantify results.
Addressing user feedback
Discoverability issues
During user interviews, we found that the homepage entry point to the new feature was often overlooked, overshadowed by the primary search options.
However, making it too prominent risked pushing critical content further down the page and disrupting key revenue-generating ad placements.
Solution
I explored various visual treatments that would make the entry point noticeable without taking up too much space. We settled on a text-based button with a ‘New’ badge—subtle yet effective. The plan was to monitor engagement post-MVP and refine it based on actual usage.
Recommendations: car models or cars for sale?
Some users were confused about the nature of the recommendations—thinking they were actual listings of cars for sale rather than general model suggestions.
Solution
I designed a new card format inspired by carsales’ Research pages rather than the search results page to reinforce that these were model recommendations, not live listings.
I also explored adding a hero section with a short explanation and supporting imagery—similar to Research pages—to build familiarity and trust. While this feature didn’t make it into the MVP, it remains in the backlog for future iterations.
More is better
- User interviews didn’t reveal a clear preference for the number of recommendations. To resolve this (and a few other questions), I ran a survey. The majority of respondents preferred seeing nine recommendations, so we went with that.
- To differentiate the top matches, I designed a split-section layout—highlighting the top three recommendations at the top and grouping the rest under ‘Further recommendations.’
- Why nine? The AI model determined this limit. While we could show fewer, we couldn’t exceed nine for the MVP.
Other challenges and constraints
Rethinking the CTA
Solution
The build
To support team alignment, I kept communication open throughout—sharing design updates and providing rationale for any changes, actively participating in team rituals, and exploring better ways to collaborate in Figma with the BA and developers.
Leveraging the design system was key to accelerating development.
I worked closely with the design system team, ensuring we used existing components wherever possible and contributed back to the UI library when needed.
Key learnings
Balancing user and business needs
In a complex ecosystem like carsales, achieving balance between user needs and business objectives is essential. Design solutions often require thoughtful compromise, recognising that what benefits the user experience can impact other business areas. A balanced approach allows for real-world testing and evidence collection, enabling informed, iterative improvements over time.
Early team alignment
Conducting a thorough inception phase with cross-functional workshops established a shared vision and aligned the team on project scope and goals, resulting in accelerated and effective delivery.
Focused collaboration
Open communication, focused meetings, and continuous tracking ensured alignment across roles, helping us stay on track and adapt quickly. Regular reassessment allowed us to address issues early and keep progress steady toward our shared goals.
