Design an Airbnb Feature Powered by an External Partner API

You are a product manager at Airbnb and have been asked to add a new feature that Airbnb's users would find useful. The one firm requirement is that the feature must integrate with an external API, meaning a service run by a company other than Airbnb. Propose the feature: who it serves, which problem it solves, why you picked it over other ideas, how the outside integration fits in, and how you would judge whether it worked. ```hint Problem first, API second Pick the user and the moment in the trip or hosting journey that hurts, then ask what outside data or capability Airbnb does not have in-house that would fix it. Starting from an appealing API usually produces a feature in search of a problem. ``` ### Clarifying Questions - Should the feature serve guests, hosts or both, and is there a stage (searching, booking, the stay itself, after checkout) the team cares about most? - What should "useful" mean here: more bookings, better stays and higher ratings, more host supply, or stronger trust and safety? - Are there limits on outside providers, such as what guest or listing data may be shared with them, cost per API call, or countries where a provider is unavailable? ### What a Strong Answer Covers - A goal tied to Airbnb's two-sided marketplace and a deliberate choice of user segment and journey stage, with the reason for it - Specific, prioritized pain points identified before any API is chosen - Several candidate features, each tied to an external capability Airbnb lacks, compared on user impact, effort and risk, ending in one clear pick - Integration design: what data goes to and comes back from the provider, privacy, cost at Airbnb's scale, coverage by country, and behavior when the provider fails - Success metrics, guardrails and a rollout or experiment plan ### Follow-up Questions - After launch, the provider sharply raises its price per call. How do you respond, and what would you have built in up front to limit that risk? - How would you handle countries where your chosen provider's data is thin or the provider is not available at all? - What experiment would you run before a full launch, and what result would make you shut the feature down?