Governance to keep platforms from diverging

WHAT THIS TESTS The interviewer wants an organizational design that prevents drift without strangling teams. The sweet spot is a federated or hybrid model, not pure central control nor a free-for-all.

A GOOD ANSWER COVERS Propose a federated governance model. A central design-system team (the stewards) owns the canonical specification, the shared tokens, and the quality and accessibility standards, and curates the roadmap. Each platform, iOS, Android, and web, has designated contributors and reviewers who bring platform expertise and represent their constraints. Define a clear change process: a proposal or RFC stating the problem and rationale, an initial design review against the system's guidelines, implementation across all three platforms ideally in lockstep so none lags, cross-platform code and design review, sign-off from the stewards plus each platform rep, then a coordinated versioned release with a changelog. Document contribution guidelines so teams can propose changes rather than fork. Track adoption to catch divergence early.

COMMON WRONG ANSWERS A single central team dictating with no platform voice, which breeds resentment and shadow forks. No process at all, letting each platform implement its own interpretation and drift apart. Approving and shipping a change on one platform before the others, creating temporary inconsistency.

LIKELY FOLLOW-UPS How do you handle a platform that genuinely cannot implement a change the same way? Who breaks a tie between platform reps? How do you keep releases coordinated across three codebases?

ONE CONCRETE EXAMPLE A team wants to change the core Button's focus behavior. They file an RFC; the stewards and platform reps review it against accessibility guidelines; web, iOS, and Android each implement the change on a branch; a cross-platform review confirms parity; the stewards and all three reps approve; then the system ships a coordinated minor release with a changelog so every platform updates together and none diverges.