Meridian

Meridian's app had grown feature by feature for six years, and it showed: seven taps to move money, dashboards crowded with data nobody asked for, and an onboarding flow that lost a third of new customers before they finished signing up.

The self-set brief: make it feel like the calmest app on the phone, not the most complicated.

We stripped the interface back to a strict modular grid — one primary action per screen, generous spacing, and a single accent colour reserved for anything that moves money or requires a decision.

• A restrained, near-monochrome palette so meaning carries the colour, not decoration.
• A numeric-first type scale, tuned for balances, dates and transactions at a glance.
• Consistent modules reused across every product surface — no one-off screens.

Six weeks of usability testing, card-sorting the existing information architecture, and a full analytics audit identified the three screens responsible for 80% of onboarding drop-off — the starting point for every decision that followed.

Low-fidelity wireframes were tested with real account data before a single visual decision was made, then rebuilt into a componentised system so every screen shares the same underlying grammar.

A grotesk built for numbers — clear digit shapes, tabular figures for tables and statements, and enough weight contrast to separate balances from metadata at a glance.

A quiet grey-black-white base keeps the interface calm; blue is reserved exclusively for primary actions, balances in motion, and confirmation states.

New flows were prototyped in real fidelity and tested in three rounds with existing customers before build — cutting the onboarding path from eleven screens to four, and the average transfer flow from seven taps to two.

A full design system was built alongside the redesign — every component documented and tokenised, so an in-house team could extend it without the original designer.

Because this is a concept piece there are no live performance figures to report. These were the criteria the work was judged against instead:

• Cut the onboarding path from eleven screens to four without losing a required disclosure.
• Two taps from open to sent money, on the smallest supported screen.
• One primary action per screen, with colour reserved for anything that moves money.
• Every component tokenised, so an in-house team could extend it unaided.