Overview
With the findings agreed, the question was how to move forward. I mapped the routes with their trade-offs, then briefed a feasibility test so the decision could rest on real engineering findings rather than assumptions.
Two directions
After a steering committee session, two viable directions emerged, and a third was set aside early.
Targeted restructuring
Give the existing tokens a proper semantic structure, and migrate components onto it.
Incremental rebuild
Start from a clean semantic foundation, and rebuild components against it over time.
Incremental fixes
Any meaningful improvement needed the token architecture reworked regardless.
Testing a real component
Engineers on Web, iOS and Android tested one component with complex token dependencies against each route. Over two to three time-boxed weeks, they estimated effort and noted what would carry over and what would break, and the findings were played back as a decision matrix.
What each option solves
The two directions were refined into four scoped routes, so leadership could weigh effort against what each one fixed.
Effort against return
Plotting the routes side by side made the trade-off clear: the lower-effort routes fixed the tokens but left most of the return on the table, while the rebuild routes carried more effort for a much bigger long-term gain.
Rollout plan
The decision
I recommended rebuilding components on the new foundation (Route 3), as the only route that closed the gap between Figma and production.
The team chose Route 2, the fastest path to a consistent foundation with the lowest engineering effort: migrate every platform onto the new tokens and styles now, and address component complexity in a later phase. I proposed stress-testing the new foundation with the design teams once migration was complete.
Migration onto the new token architecture was underway as I completed my contract, with component work set as the next phase.
Looking back
The system sat between teams that needed different things from it. Product designers wanted speed and room to explore, platform engineers needed something they could build and maintain across Web, iOS and Android, and leadership needed a plan that delivered within the timelines and budget.
The work was balancing those constraints rather than solving for one of them: auditing what existed, listening to each team, defining a token architecture everyone could share, and laying out routes with honest trade-offs so the decision could be made on evidence.