How would you architect a white-label content system for multiple brand voices?

WHAT THIS TESTS: This question evaluates your ability to separate content from presentation in a multi-tenant system. The interviewer wants to see if you can design a single backend that serves distinct brand voices without forking application logic or creating operational sprawl. Specifically, they are looking for familiarity with headless CMS patterns, tenant resolution strategies, and component-level abstraction.

A GOOD ANSWER COVERS: First, propose a headless CMS as the single source of truth where each tenant owns isolated content spaces or variant fields for the same UI keys. Second, explain tenant resolution at the edge or application layer using domain, subdomain, or a custom header to identify the incoming client. Third, describe a content delivery pipeline that fetches the correct copy set at build time for static sites or at request time for dynamic apps, caching by tenant to keep latency low. Fourth, insist that UI components remain brand-agnostic, consuming tokenized keys and receiving injected strings rather than embedding text internally. Fifth, mention governance such as role-based access in the CMS so client editors cannot see or mutate another tenant's copy.

COMMON WRONG ANSWERS: A major red flag is suggesting separate frontend codebases or per-tenant app deployments, which duplicates engineering effort and destroys scalability. Another is hard-coding copy in components or JSON files inside the repository, forcing a release for every text tweak. Proposing a monolithic traditional CMS tied to a single presentation layer also fails because it couples brand voice to code. Finally, overlooking cache isolation can cause one tenant's copy to leak to another, creating brand safety risks.

LIKELY FOLLOW-UPS: The interviewer may ask how you would handle A-B testing copy for a single tenant without affecting others, how to support right-to-left languages or locale variants under the same tenant, or how you would migrate an existing hard-coded app into this system without a big bang rewrite. They might also probe failure modes, such as what happens when the CMS is down and the edge cache is cold.

ONE CONCRETE EXAMPLE: Imagine a global restaurant chain using one headless CMS instance. Each country site is a tenant sharing the same menu component code. A visitor hitting uk.chain.com triggers a resolver that maps the domain to the UK tenant ID. The app requests the Menu section from the CMS filtered by that ID, receiving localized prices and a formal brand voice. Meanwhile, a playful sub-brand at play.chain.com resolves to a different tenant ID and pulls casual copy into the identical component structure, all without a separate deployment.

Source: thebcms.com