Vacation Rental Candidate Ranking
Design an online service that ranks eligible vacation-rental candidates for a traveler request.
Problem statement
For this exercise, assume the source-backed system identity and the explicitly bounded practice contract that follow. Design the candidate-ranking stage of a vacation-rental search system. Upstream retrieval supplies listings that are eligible for the requested place, dates, and guest count.
Your service must order those candidates for a traveler, return a stable paginated result, and collect outcome signals that can improve later model versions. Inventory size, request rate, and latency targets are intentionally open interview inputs; state bounded assumptions before choosing the online scoring path.
What to build
- Rank an upstream set of eligible listings for a search request and traveler context.
- Return deterministic pagination for a ranking version while listings and features continue to change.
- Use listing, query, and traveler features with an explicit freshness and fallback policy.
- Record impressions and outcomes with enough lineage to evaluate a ranking release.
At scaleExercise boundary
- Catalog size, candidate-set size, request rate, and latency targets are intentionally unspecified; choose and defend bounded assumptions.
- The candidate set can be larger than one response page, so pagination must not reshuffle previously returned results.
- Listing availability and feature values can change while a traveler pages through one ranked result.