I've built five design systems. Three of them shouldn't exist. Two of those cost their companies more than they saved. I want to talk about why — and what I'd do instead, given fourteen years and a wiser liver.
Systems before customers
The failure mode I see most often is teams building a system before they have the customers to justify one. The product has three screens and four users; the team has a Storybook with 120 primitives. That's not a system, that's an altar.
“A design system is a deployment strategy. If you're not deploying faster, it's a museum.”
The version that actually works
- Start with tokens, not components. Colors, space, radius, type. Fit on one page.
- Ship primitives when they appear twice. Not thrice, not the first time.
- Documentation is a design deliverable. Not an afterthought.
- Governance should be a weekly hour, not a council.
Built this way, a design system is a margin — a little extra speed on every feature — not a product unto itself. Which is exactly what it should be.