Maintenance Bill

The maintenance bill is the real ongoing cost of a design system that almost nobody calculates before they start building one. It covers every hour spent syncing Figma variables to code, updating components after Radix or Tailwind updates, rewriting docs that went stale in a quarter, and chasing visual drift in the actual product. The concept exists because the entire industry narrative sells build once reuse forever while quietly ignoring the daily tax required to stop the thing from rotting. Teams treat it like a side quest. It is a full time product with its own roadmap and customers.

It is not the exciting six week Figma sprint where one designer publishes 80 components and calls it done. That part feels like progress and looks great in a roadmap slide. It is not the one time cost of choosing a color palette or setting up a token file either. The maintenance bill is the invisible grind. It shows up as Friday afternoon Slack messages about why the button looks different in production. It is the Storybook instance that breaks on a dependency bump and sits broken for three weeks before anyone notices.

Most teams confuse the initial payoff with perpetual savings. They imagine linear growth in reuse hours. The actual math is that maintenance hours compound with every component variant and every framework update. The lines cross faster than anyone admits. By the time they notice the team is back to copying Tailwind classes the bill has already eaten their velocity.

Look at a typical 12 person startup in 2023. Senior designer spends six weeks building a component library in Figma. They set up 15 tokens and write a Notion page no engineer reads. Six months later 40 percent of the components are outdated. Tokens in Figma do not match the codebase. Themes have diverged. Engineers ship three slightly different loading states because fighting the system takes longer than rebuilding. That is the bill showing up as lost time and visual debt. Contrast it with Stripe. They run a dedicated design infra team that treats the system like its own product with a changelog and version releases. The bill exists but it is paid by people whose full time job is paying it.

Shopify Polaris and GitHub Primer follow the same pattern. Entire orgs or public repos with real governance. The maintenance bill is justified because the customer base runs into the thousands both internally and externally. Atlassian poured years into their system. Most teams will never approach that scope or headcount. Their bill would be pure tax.

Factor the maintenance bill into your decision before you touch Figma. If you have one product and fewer than 50 engineers the bill will almost always outweigh the payoff. Six solid tokens plus Shadcn primitives plus one designer with taste wins every time. The system tax stays flat while the AI payoff cratered in 2024. A junior with Cursor can generate consistent UI in minutes. Your 200 component system still demands the same upkeep.

Use a real design system only when three or more hard signals appear. Multiple products sharing one identity. Headcount in the hundreds. A staffed design ops role. External consumers. Or a regulated domain where compliance must be encoded once. Miss those conditions and the bill turns your maturity flex into expensive cosplay. The tradeoff is clear. Pay in dedicated headcount or pay in drift, inconsistency, and engineer frustration.

The cleanest products in 2026 run the opposite playbook. Linear keeps a tiny token set. Vercel ships heavy Tailwind with their own primitives. Cal.com publishes primitives publicly. None carry a fat maintenance bill. They look more coherent than companies ten times their size because they refused to build theater.

Track the bill explicitly. Run a quarterly audit. Count hours spent on system work versus hours saved in product shipping. Most teams discover the optimal size is far smaller than they first assumed. The honest accounting changes everything.

The maintenance bill always comes due. Pay it with dedicated staff or shrink your system until the bill disappears. There is no third option.