Case study / 03
Project Sprout
A trust-first debt planning app. Five-phase tier model meets people where they actually are with their statements, with visible math instead of black-box optimization.
Phase / 01 Discover
Map the workflow.
Most debt tools assume you know your APRs and minimums and are ready to optimize. Most people in debt know neither, and shame-spiral when a form demands them. I walked through the actual flow with people who avoid their statements — the diagnostic isn't "snowball or avalanche?", it's "can you cover minimums this month, and is there anything in savings?" The frame I landed on was a five-tier phase model — Triage, Stability, Payoff, Buffer, Learning — meet the person at the phase they're actually in.
Phase / 02 Build
Ship the system.
React + Vite frontend, FastAPI + Postgres backend, Clerk for auth, Stripe for billing, Anthropic for the (premium-only) AI summaries. The load-bearing UX move was a three-state chip on every numeric field — Known / Rough guess / Later — paired with a transparency block that names every assumption ("APR on Capital One is a rough guess at ~24%"). The planning engine is one pure function: debts, income, expenses, buffer in; tier, today_action, and a calculated/estimated/missing breakdown out. Visible math, not magic.
Phase / 03 Deploy
Earn trust.
Shipped on Render (frontend static + backend web service + cron for email reminders) with Resend for inbound reply parsing so users can check in by email. Initial rollout is to a small private cohort for the v0 redesign — the relief-valve UX ("Take a moment" pill, "Save what I have" buttons, soft landing copy) is the part designed to earn the qualitative feedback that drives the next pass. Deploy here is as much a tone exercise as an infrastructure one — every error message is a chance to make the experience either heavier or lighter.
Phase / 04 Compound
Reuse the learning.
The v0 redesign (Weeks 1–4) was the compounding pass: porting the chip pattern across every input field, adding what-if scenarios (+$50/mo, $500 buffer first, minimums only) directly on the plan view, splitting the budget page into one phase-appropriate question per tier, building the SVG tier path and proportional debt circles on the progress page. Each iteration was driven by the same instinct — the math is fine, the trust is the product.