“I would start with one person, not a system. Pick the designer with the best taste and put them in charge of the shared language: spacing, type, buttons, the handful of patterns that show up everywhere. The others' job becomes matching that, not inventing their own.”
“I’d start with the one thing users actually see across every screen: a single spacing scale and one button component. At Beats Music, I led a unified tools framework because the product was becoming a patchwork of internal apps, and the first thing we locked down was the primitives, not the page templates.”
“I built Arabesq because I watched the same thing happen: three designers shipping three versions of every button. Start with the smallest system that stops the bleeding, a type scale, a spacing scale, and one button variant in each state. The rest of the system grows from those three decisions.”
“I would start with a design system gap analysis: audit every button, spacing value, and type scale already in the product, then lock the smallest set of tokens the whole team can actually agree to use. At Chime and across brands like Dickies and Sundays, I have seen three designers become six once the system stops being theoretical and starts as a constrained kit built from what already exists.”
“I would start with one thing every designer touches daily: buttons and spacing. At Mark43, locking those down first stopped the four different apps feeling faster than any grand system rollout.”
“I would freeze new patterns first, then audit the three most visible flows and turn those into your first real components. At BitGo we saw the fastest consistency gains when we stopped letting designers invent buttons and started from the screens users actually hit every day.”
“I would start with the smallest surface area that causes the most visible damage: buttons, type, and spacing. At Capital One, the Learn & Grow articles looked like four apps stitched together because content creators had a free text editor and no shared rules; once we locked those three things down, the rest of the system followed.”
“I have watched design systems fail when they start as a Figma library nobody updates, so I would not start there. Audit the three most repeated screens, lock in spacing and button rules as code first, then backfill Figma. At ClockShark, that discipline lifted conversion 18% in sixty days because the product finally looked like one thing.”
“At Ordnance Survey I rebuilt OS Maps by first locking down spacing, type, and buttons in a single Figma library, then got the engineers to consume the same tokens in code. Start with the smallest set of decisions your three designers actually disagree on, make them impossible to override, and only then add more components.”