DESIGN
THE PATH TO APPROVAL
Rebuilding a trust-broken renter questionnaire — and the AI-native way it got built.
My Role
Product Design Lead
Team
1 Product Manager
10+ Engineers
Tools
Figma
AI design-to-code
FullStory · ListenLab
Adobe Creative Suite
Employer
TheGuarantors
Overview
TheGuarantors steps in as a renter's guarantor so they can qualify for an apartment. The questionnaire is how they apply — the front door to core revenue. I led its end-to-end redesign on a new modular architecture: mobile-first, persona-aware, and built with an AI-native workflow. Launched 2026.
Design Challenge
Renters didn't trust the old flow, struggled on mobile, and hit dead-ends without US credit. Rebuild it to feel secure, effortless, and inclusive — without losing conversion.
Business Goal
Migrate every renter onto the new architecture and turn it into growth — higher completion, more purchases — protecting the ~$200M GWP this flow gates.
*Due to NDA, certain details and visuals of this project cannot be fully disclosed here. If you’d like to explore the complete case study, including in-depth insights and outcomes, please reach out to Wen.
Year
2025-2026
THE STAKES
One questionnaire gated almost every renter — and the business behind them.
Nearly every applicant had to pass through this one flow to get approved — most of them on a phone. That made it the company's highest-leverage surface: the gateway to its core, renter-paid revenue, where a little friction meant real money and lost trust.
99%
of renters went through this flow
55%+
applied on mobile devices
~$200M
in annual guarantee volume (GWP)
THE PROBLEM
Three barriers stood between renters and approval
The questionnaire wasn't failing on one dimension — it broke down on trust, on mobile, and on who it was built for.
From 20 in-depth interviews + a 100-person AI-moderated A/B test (results in 3 days).
Honestly, I thought it was spam — I had no idea how you even got my information.
TRUST · a first-time renter
Renters couldn't tell it came from their own landlord.
I was filling everything out on my phone, and it just wasn't that easy.
MOBILE · a mobile user
55%+ applied on a phone, but it was built desktop-first.
I don't have a credit score here in this country, so it was really hard to get through.
FIT · an international newcomer
The flow assumed a credit-established American.
THE SOLUTION
Mobile-first, one question per page — a flow that adapts to each renter
The old flow stacked every question on a single scroll. I rebuilt it into a guided, mobile-first experience that adapts to each renter.
❌ BRFORE - OLD QUESTIONNAIRE
One "ENDLESS" scroll 😩
Every question crammed onto one endless scroll — dense declarative text, "+" toggles, tiny targets, clunky native inputs. It read like a legal form, not a guided flow.
✅ AFTER - REDESIGN
One question at a time 👍
One question per page, tap over typing, a visible progress stepper, and a flow that reshapes itself to who's applying.
Trustworthy
Dynamic
One question/page
Tap over typing
✅ AFTER - REDESIGN
Trust, up front.
I rebuilt the header around a mobile-first pattern: on scroll, it collapses from our tagline to your property's address — showing the application comes from your own landlord, not a stranger. That visible connection is what kills the "spam" fear.
✅ AFTER - REDESIGN
Less typing!
Wherever an answer could be a tap, I made it one — country pickers, employment status, and more. Less keyboard, fewer errors, and a faster flow on mobile.
✅ AFTER - REDESIGN
A payoff, not a dead wait 🎉
Loading into a celebratory approval!
✨THE HEADLINE: VIBE DESIGN → VIBE CODING
I taught our design system to AI —
~90% production-ready UI from a prompt
01
Design System
Our components & patterns
02
Encoded as AI skills
The system, taught to AI
03
Prompt
Describe the page
04
Design-tuned UI
Web + mobile front-end
~90%
New pages generate near-complete from our design system — minimal cleanup, sometimes a single prompt. So when the funnel pointed to checkout as the next problem, I didn't sketch — I mocked up the whole checkout redesign straight from prompts (see example).
"It's really cool — you're taking my job… just kidding. This actually makes my job much easier."
— An engineer, on seeing the workflow
Honest caveat: the output is a strong front-end reference, not production-ready — and the real work is adoption, which I lead as change management.
THE IMPACT
Conversion jumped 30% → 38% — then I followed the funnel to the next bet
30% → 38%
Questionnaire conversion (+8pp)
Real post-launch data (Fullstory)
2×
Trust score
ListenLab Concept Test
40%
Faster completion
Usability testing
The honest twist
More people finished the questionnaire — but revenue barely moved. FullStory sessions showed why: renters stall at checkout, hesitating at the full price. The redesign didn't fail — it relocated the problem. My next bet: a checkout redesign aimed straight at that hesitation.
WHAT'S NEXT - PROTOTYPED NOT SHIPPED
The checkout fix — mocked up straight from prompts
Same AI workflow, pointed at the bottleneck the data exposed: make the price legible, the value obvious, the commitment feel safe. Prototyped in a fraction of the usual time.
