Design System Foundation
Shared UI standards, sustained four-plus years without dedicated resourcing — still in use after I left.
Foundation
Company
Remember
Timing
2022 – Jun 2026
Role
Senior Product Designer
Team
Design, Engineering
Contribution
System foundation ownership
Overview
When I joined, there was no design system. The team was mid-migration from Sketch to Figma, and designers worked in separate product crews with little reason to align. Scattered references taxed every experiment cycle. Having shipped a group-wide design system at my previous agency, I knew the fix, and that a shared guide of what already existed had to come first.
Problem
Screens built before the system carried their own local versions of common patterns. The same component lived in multiple variants; designers re-decided solved problems; engineers received contradicting handoffs. As an early member of the job-posting product, I argued the cheapest time to organize was now, while screens were few. Every month of waiting raised the price.
What this shows: the drag was real; solved problems being re-decided file by file.
Definition of Success
No conversion metric, so I defined success before starting: teams reuse the shared standards in live projects without being required to. Voluntary adoption, not coverage percentage, was the test.
My Role
Legacy screen audit
Component standards
Detailed guides & edge cases
Handoff loop with Engineering
Design Decision
Bottom-up from evidence, not top-down from a spec. Standards that solve a problem a team already has get used; standards that anticipate problems get ignored. The deepest design work was component structure: buttons carry a dedicated states layer, tinted white or black for hover and pressed, so one rule survives every color variant.
Considered and rejected: starting with a full design system. Nothing had even been collected yet; a guide of what already existed had to come first, or the system would have codified guesses instead of patterns.
1
Collect legacy screen resources
2
Map patterns and repetition
3
Merge, then systematize the rest
4
Write guides, reinforce missed edge cases
5
Handoff loop: intent shared, structure adjusted to implementation
6
Build → joint review → release
Result
2022
I left
2026+
4+ years, no dedicated resourcing: built on time designers and engineers chose to invest. Still in use after my departure.
Work nobody values doesn't survive four years on voluntary effort. In use across four B2C and two B2B product areas, by six designers and more than twenty engineers.
1
Single source of truth: scattered references consolidated
2
Reflected in Storybook by engineers investing their own time
3
Consistent design–engineering handoff
Shipped without quantitative instrumentation: no numbers retrofitted onto it.
Trade-offs & Guardrails
Foundation work competes with metric-visible feature work. Here it was structural: it had to fit inside volunteered time. Each consolidated component was usable on its own, so the system delivered value incrementally instead of waiting on a finished whole.
What This Shows
It started small. I wanted a shared guide so my colleagues and I could work with less friction, and so the product could stay consistent. Then I wanted the same shared language with the engineers we build with, to lower each other's workload. Because we all felt that need, people kept putting in their own time over the years, and the system kept getting better.
Next
Tokenization was the mapped next stage: direction defined, not yet executed when I left. Former colleagues confirm the system is still in use, and as AI-assisted workflows spread, a system machines can read cleanly has become more important, not less.
