A booking flow rebuilt around a step we were not allowed to change
Twelve weeks, four people, and a payment provider that could not be touched during peak season. Completed bookings up 22% on flat traffic — and a seat-selection interaction we spent three weeks on that nobody uses.
Meridian Foods
What the twelve weeks actually contained
Three points, and the third is the one that should tell you most about how we work.
- The constraint: we could not touch the payment providerMigrating it would have put the client through a PCI re-certification in the same quarter as their peak season. So the redesign had to work around a checkout step we were not allowed to change, and roughly a third of the design work went into making that step feel like it belonged.
- What actually movedCompleted bookings up 22% year on year, on flat traffic. The larger number is that support tickets about the booking flow fell by two-thirds, which is what the client noticed first.
- What we would do differentlyWe spent three weeks on a seat-selection interaction that testing later showed nobody used. It shipped anyway because removing it was politically awkward by then, and it is still there.
Three more
Four projects in total rather than eleven. A portfolio that scrolls past the fourth is optimising for a reader who left at the second.
Lantern Trust · 2025 · Identity and site
A charity that needed to look smaller, not bigger
Lantern had grown to forty staff and their brand had grown with it — into something that read like a government department. Donations from individuals had been falling for three years while institutional grants rose, and the board had attributed it to the market. The work was mostly subtraction: a warmer palette, photography of four named people instead of nine hundred beneficiaries, and a donation flow that stopped asking for a title.
- The constraint: no rebrand budget, so the wordmark stayed exactly as it was
- Individual donations up 31% in the first year
- What we got wrong: the first photography round was too casual and had to be reshot
Kestrel Software · 2025 · Product design
Making an admin panel that three people use feel worth building
An internal tool with a total user base of three, which is the kind of project that normally gets a template and a shrug. It was worth doing properly because those three people spent six hours a day in it, and the existing version was costing roughly ninety minutes of that to navigation alone. We rebuilt around keyboard-first navigation and a single command palette.
- The constraint: no new backend endpoints, at all, for six months
- Measured task time down 40% across the four commonest jobs
- What we got wrong: we shipped without a mouse-only path and had to add one
Harbourside Retail · 2024 · Design system
The system that replaced four systems and nearly became a fifth
Four teams, four component libraries, and a genuine question about whether consolidating them was worth a quarter of everyone’s time. We argued it was not, initially, and were overruled. Six months later the honest assessment is that it was worth it for three of the four teams and a waste for the fourth, whose product is different enough that they now maintain a documented fork.
- The constraint: no team could stop shipping during the migration
- 3 of 4 teams migrated fully; the fourth runs a documented fork
- What we got wrong: we should have scoped the fourth team out on day one
Start a brief
Two sentences is enough to begin. We reply within two days either with questions or with the name of a studio we think fits better — we turn down roughly half of what comes in, usually on timing rather than on fit.
Portfolio Index 02
One project per screen, four in total, each naming the constraint it worked around — nobody reads four case studies, and the constraint is the only sentence a competitor cannot copy.