Why design systems fail in year two

Year one of a design system is a honeymoon: launch energy, executive attention, adoption charts pointing up. Year two is where systems die — quietly, from three operating decisions nobody made.

Decision one: who owns it when the champion leaves

Almost every failed system we've audited traces back to a single departed champion. The system was a person, not an institution. The fix is boring and essential: named ownership with budget, a succession answer, and system health metrics reviewed at the same cadence as uptime. If nobody's performance review depends on the system, the system has no owner.

In our Northlight engagement, the governance charter was written before the first component — owners, decision rights, escalation paths. When their design director left in month nine, the system didn't notice.

Decision two: how hard it is to contribute

Systems die of constipation. If a product team needs a component variant and the path is "file a ticket and wait six weeks," they'll fork it locally — and every fork is a small secession. Healthy systems make contribution cheaper than divergence: clear intake, a working-draft tier, and published response-time commitments.

Measure contribution velocity like a product metric, because it is one. Northlight's target: any squad gets a reviewed pattern decision in five working days. Adoption held at 92% because leaving the system was never worth it.

Decision three: what version two costs

Nobody plans for the system's own evolution. Then a rebrand, a platform shift, or an accessibility mandate arrives, and teams discover that changing a token breaks forty screens with no migration path. Semantic versioning, deprecation windows, and migration tooling aren't engineering pedantry — they're the difference between a system that ages and one that calcifies.

Budget year two before you celebrate year one. A system with no evolution plan isn't an asset. It's deferred demolition.

Every local fork is a small secession. Systems die of constipation before they die of neglect.

---

The takeaway

• Institutionalize ownership: budget, metrics, succession — not a champion

• Make contribution cheaper than divergence, and publish the SLA

• Version the system like software; plan migrations before you need them