Iterating on an Overlooked Personalization Touchpoint

Why research mattered

A carousel was meant to collect three things easily: the loan amount, annual income, and the purpose of the loan. With them, Experian could personalize a user's offers. Because Experian already knows a lot about its users, they expect an experience that asks less of them.

Working with analytics, the product team found the carousel wasn't working: many users ignored it, and little information was coming in. The team had three guesses. Users might be only browsing, hesitate to share personal information, or not see why sharing it helped them. The team wanted to know if extra explanation would help.

Research approach

This project was part of Experian's early rollout of its Product Operating Model, where pods of research, design, product, and development worked together to ship ideas faster. As the pod's researcher, I narrowed the designer's many variations over two weeks. Each method answered a different question and set up the next one.

• Method 1: AI eye tracking. What draws attention? A fast first look at which variations stood out.
• Method 2: First-click testing. Would users look in the right place? A check that the loan profile fit how users expect to find things.
• Method 3: Preference test. Which variation do they favor? The remaining few designs, compared head to head.
• Method 4: Moderated usability testing. How do the top two hold up? Live sessions where I could uncover the why.

• Why this order. Unmoderated tools like Attention Insight and Lyssna made recruiting and testing quick, so early decisions didn't have to wait for live sessions. Moderated sessions came last, when only two designs remained and I needed to ask why.
• Participants. 145 Experian users across all methods, from our panel or through a screener about the financial companies they use. I kept those who browse for personal loans at least occasionally, since they were the ones this feature was for.
• Tradeoff. The eye tracking tool predicts attention rather than observing people. I treated its accuracy claims as the vendor's, and used real users in later methods to check what it suggested.

Key findings

1. Reassurance up front mattered more than explanation afterward

Users liked prominent messages that their offers were prequalified, that checking wouldn't affect their credit score, and that they wouldn't get sales calls or emails. Explanations below the fields were often missed or came too late.

2. Most users preferred the simpler profile

Comparing a minimum and a maximum version, most preferred the minimum version's simplicity. Those who preferred the maximum version pointed to its loan-amount input, not its extra explanation.

3. A pre-filled loan amount split users

It felt convenient to some and pushy, even predatory, to others. “Personalized” seemed to mean using, and visibly acknowledging, what Experian already knew along with the user's current needs. A number that ignored their situation felt less like help.

4. Approval odds helped, if users could see them

Approval odds increased confidence, but their prominence mattered. The same information helped less when it was hard to see.

Influence and outcomes

I presented findings after each round, so the team could decide how to proceed. I gathered the unmoderated results on a shared FigJam whiteboard, with user feedback, eye tracking heatmaps, and first-click maps in a single place, and played video clips for the usability findings.

Our product manager had built the maximum version himself in Figma Make, expecting users to appreciate the added information. When testing showed they preferred the minimal version, he accepted the results. Where preferences were closely split, such as pre-filled amounts and input controls, I told the team they could refine those elements or move forward with their preferred version.

• What I recommended: lead with reassurance, keep the profile concise, and let users type and change their loan amount.
• What was implemented: the loan profile replaced the carousel. There have been minor refinements since.
• What was measured: in an A/B test run by analytics for several weeks, the profile was twice as effective as the carousel at getting users to fill out their information. 78,000+ users have since completed it, and it now runs at 100% in production.

Reflection

What I learned. When research moves fast, teams need to keep questions focused, compare alternatives fairly, and separate what users notice, understand, prefer, and actually complete. To a team that wants quick answers, I'd say: agree on the decision the research must inform, and let that be your anchor throughout the research process.

What remained unresolved. Preferences on pre-filled amounts and input controls were closely split, so I left those choices to the team.

Appendix: methods and supporting detail

The full method notes and the rest of the findings, for a deeper read.