NerdWallet
Research on the fault line between usefulness and trust — where 65% of people get no offers at all.
I led the research behind the Personal Loan Prequalification revamp — a flow that asks for progressively sensitive information, down to a Social Security number, in exchange for pre-qualified offers. It hadn’t been researched since 2020. I owned the loop end to end: plan, guides, moderation, synthesis, readout.
The declined majority.
The flow trades sensitive information for pre-qualified offers — but roughly two-thirds of people who go through prequalification receive nothing back. That’s the hardest moment in the whole experience: someone hands over their SSN and gets told, in effect, “no.” The design question was as much about that letdown as about the happy path.
Two methods, deliberately paired.
I ran two rounds of moderated, think-aloud usability testing with nine participants across two segments — five “consumers with choice” (good/excellent credit) and four “with fewer choices” (poor/average credit) — framed honestly as directional. Then a preference test on a hundred respondents settled a specific design question with numbers.
Qualitative — 9 participants, 2 segments
All five friction themes surfaced in both credit groups — a strong signal they were structural, not segment-specific.
Quantitative — 100 respondents, 5 designs
A preference test on five candidate “declined-results” screens, to choose the letdown moment with evidence.
Five frictions, in their own words.
“I would never know that it’s a compare button… based off the symbol it’s showing me like $1, like a bill.”
“I’d prefer not to give my social at this point… instead when I’m actually filling out an application.”
“It’s really hard to know who to believe.”
“I’m just looking to see which has the best APR.”
“The signup process in the middle of answering the questions… that’s kind of what brings it down.”
Five candidates for the letdown moment. One won.
Rather than argue the declined-results screen in a room, I put five designs in front of a hundred respondents. Design 2 won with 43% — the next best drew 24% — so the hardest screen in the flow got decided by evidence, not opinion.
Design 1
Design 2 — 43% ✓
Design 3a
Design 4
Five frictions, five moves.
Re-icon Compare
The compare control read as a price, not an action — reposition and re-icon it.
Defer the SSN
Ask for the Social Security number (and account creation) later, at application — not mid-questionnaire.
Elevate APR
The number people actually compare on was buried — lift it in the hierarchy.
Explain the soft pull
Make the data-use and soft-pull language transparent, to earn the trust the flow depends on.
Protect step-by-step
The one thing people liked — the calm, one-question-at-a-time layout — keep it.
Pair the two methods and you get both the why and the which — the story and the number.
The findings fed directly into the team’s 2022 product iterations on prequalification — the qualitative rounds explained why the flow lost people, and the preference test settled which design to ship for the hardest moment in it.
No before/after completion-rate figures were preserved, so I don’t claim a post-ship lift — the honest outcome is a decision-grade body of evidence handed to the product team.