Maru - banking, softened.

The brief

Maru's team had spent two years making the money engine correct - budgets, rules, approvals. What shipped to users was correct and unusable: jargon-heavy copy, nine actions competing for every screen, and an empty state that made first-time depositors feel incompetent.

What we did

1. 1We ran a week of interviews with users who had churned after signup. The pattern was consistent: people didn't leave because of fees or features - they left because the app made them feel judged.
2. 2The design system centered on a single principle: one primary action per screen. Every secondary control moved into a drawer or behind the first swipe, and the copy passed a 'grandmother test' - if it needed a footnote, it got rewritten.
3. 3Two whiteboard cycles and one clickable prototype later (about 23 days), the flow was tested on ten real users before a line of production code changed.

The outcome

Activation doubled within a quarter of launch, NPS climbed by 34 points, and the rebuild shipped without a single migration regression - the hand-off doc was 11 pages, and Maru's engineers have since shipped three releases from it without calling us.