Design token pipeline serving Web, iOS, and Android

WHAT THIS TESTS: This question evaluates your ability to architect a cross-platform design system pipeline that treats design decisions as data. The interviewer wants to see that you understand how to prevent drift between platforms, how to structure tokens for scalability, and how to automate distribution without losing governance. Senior candidates are expected to discuss layering, source-of-truth immutability, and build-time transforms rather than runtime hacks.

A GOOD ANSWER COVERS: First, choose a W3C design token specification compliant JSON file as the single immutable source of truth, not a design tool or CSS. Second, organize tokens into three layers: option tokens that define raw values like color palettes, decision tokens that map those values to semantic meanings like brand-primary, and component tokens that bind styles to specific UI elements like button-background. Third, select a build tool such as Style Dictionary, Tokens Studio, or a custom Node pipeline to read the JSON and generate platform-specific outputs like CSS variables for Web, Swift constants for iOS, and XML or Compose objects for Android. Fourth, address governance by keeping option tokens private to product teams to reduce file size and support non-breaking changes, while exposing only decision and component tokens. Fifth, define the CI/CD flow, including whether the pipeline is fully automated or requires manual approval for token changes to prevent unintended visual regressions.

COMMON WRONG ANSWERS: Treating Figma, Sketch, or CSS variables as the source of truth is a red flag because design tools are not version-controlled artifacts and CSS is platform-specific. Proposing runtime API fetching for tokens in mobile apps shows a misunderstanding of build-time optimization and offline reliability. Ignoring token layering and dumping all values into a flat file signals poor scalability. Failing to mention transforms or acting as if hand-copying values into each repo is acceptable will immediately downgrade the answer.

LIKELY FOLLOW-UPS: How do you handle token versioning and breaking changes across platform releases? What is your strategy when a platform needs a token value that does not exist in the global set? How do you keep designers and engineers in sync when the token spec evolves? When would you choose a fully automated pipeline versus one with manual approval gates?

ONE CONCRETE EXAMPLE: Imagine a z-index token set. The source JSON defines option tokens for raw numbers and decision tokens for layers like navigation, toast, and modal. The build pipeline transforms these into a CSS custom property for Web, a Swift enum for iOS, and an Android resources XML integer for Android. When design needs to add a new spinner layer, they update the JSON, the pull request triggers the pipeline, and all three platforms receive the new value in their native formats without manual copy-paste.

Source: martinfowler.com