The problem
Castell sells treasury software to companies with a few hundred staff and more bank accounts than anyone planned. Its first product worked, but every customer asked the same question on day one: which of these numbers is the real one?
The balance was spread over four screens and none of them agreed to the cent.
What we built
- One balance, to the cent. Cash across every bank and currency adds up to a single figure, with the accounts that make it up one click away.
- Approvals on a phone. A payment needs two people. Each can approve from wherever they are, and the second sees who went first.
- A forecast that shows its working. Fourteen weeks ahead, with known payments in solid bars and estimates hatched.
If two screens disagree, a finance team trusts neither.
How it went
The hard part was not the design. It was agreeing, with Castell and three of its customers, what a balance means when a payment is approved but has not left. Once that had one answer the screens followed. We built the design system first, so the web app and the phone approval share every component.
