DoZarplaty

The core product was a microloan service with a network of 100+ physical branches. 48% of applications came through offline channels, with a cycle of one hour to a full day from application to disbursement. The pandemic and forced branch closures put the entire business model at risk.

Redesign the online application flow so customers could get a loan in minutes without visiting a branch. Simultaneously: lift conversion at every step and build a design system.

Instead of a monolithic redesign — sequential iterations with data. Researched every step of the funnel: where we were losing users, why, and what would move conversion first.

In parallel, built a design system with tokens and components for web, mobile web, and native apps — it covered every subsequent product at Devim.

Led a team of 12: product designers, UX researchers, graphic designers.

Conversion started at 37%. Legacy code and accumulated hacks made fast experimentation impossible — we launched a new product on a clean codebase to use as a sandbox.

First wave: removed optional fields, reordered steps, added whitespace → 47%. Second wave: a clear progress bar, rewritten field hints → 64%.

Carried the improvements into a Flutter build of a new product. Due to a shared backend with the web version, some fixes didn't transfer — reached 37% → 60%.

Expensive SMS costs from password recovery — especially for customers who couldn't remember their password on payment day.

First hypothesis: limit passwords to 6 digits — easier to remember. Result: +11% more recovery requests. Worse. The data showed people forget a 6-digit PIN just as easily, just faster.

Second hypothesis: on the same device, ask for a PIN instead of a password. A pattern familiar from banking apps. Result: −64% in password recovery requests.

The final signing step had a surprisingly high drop-off: the user had completed the full journey, passed scoring, been approved — and still didn't sign.

Hypothesis: information overload. Terms, documents, data consent — all at once. The fix was counterintuitive: moved the SMS signing block above the checkboxes so it was visible on the first screen without scrolling. The action was now in focus; the detail-oriented could still scroll down to review everything.

Result: +36% conversion on contract signing.

After the application funnel was optimized, focus shifted to retention: reducing delinquency without changing product logic.

Three sequential experiments in the personal account: — Color accent on the 'Full Repayment' button → +3.5% conversion to repayment. — Status colors for 'Payment due soon' and 'Loan overdue' → −5% monthly delinquency. — Notification system redesign → −30% delinquency.

Small visual changes, no redesign — direct business impact.

Share of online loans: 52% → 98%. The company closed 100+ physical branches without losing volume.

Application conversion: +18% overall. Form: 37% → 64%. Contract signing: +36%. Delinquency: −30%. Password recovery requests: −64%.

Team grew from 3 to 12 designers. Design system covered all Devim products.