Karda Finance
A personal finance app that people downloaded and never finished setting up. We redesigned the first four minutes, then built the system that kept the rest consistent.
- Client
- Year
- Scope
- Live site
- Visit
The problem
Karda’s acquisition numbers were fine. Their activation numbers were not. Most new accounts stalled at the same place: a bank-connection screen that asked for trust the app had not yet earned, three taps after install.
The wider issue was that eighteen months of shipping fast had left four different button styles and two definitions of “balance” in the same product.
What we did
- Watched eight people install it. Six hesitated at the same sentence. It was a legal disclaimer written by a lawyer and placed by nobody.
- Moved the ask. Connection now comes after the app has shown something useful, so the request arrives with a reason attached.
- Designed the empty states properly — the screens a new user sees most and the ones that had never been drawn.
- Built the component library: every control with its hover, focus, loading, empty and error state specified as tokens engineering could implement directly.
- Rebuilt the identity around the product, not the other way round. The palette was chosen against WCAG AA on the densest table in the app.
Outcome
- Setup completion up from 41% to 68% over the two releases after launch.
- Four button styles reduced to one, with variants — 60% less UI code in the design layer.
- Support tickets mentioning “can’t connect my bank” fell by roughly half.
The system is still theirs to run: we handed it over with a working session and stayed on for design reviews for the following quarter.
