Airbnb: Search Ranking — Balancing Hosts and Guests
How Airbnb designed a search ranking algorithm that had to satisfy two completely different customers at once — guests looking for a stay, and hosts wanting bookings.
The challenge
Unlike a typical e-commerce search where ranking simply optimises for the buyer, Airbnb's marketplace had to balance two sides simultaneously. Ranking purely by price or rating might surface great listings for guests but starve new or lower-rated hosts of any bookings, causing them to leave the platform entirely. Airbnb needed a ranking system sophisticated enough to grow both sides of the marketplace at once, not just optimise conversion for guests.
The strategy
Airbnb built a machine learning ranking model that considered dozens of signals beyond price and rating — including host responsiveness, booking lead time, guest search history, and even factors specific to marketplace health like whether a listing was new and needed initial exposure to gather reviews. The strategy explicitly traded some short-term conversion optimisation for long-term marketplace liquidity.
Want the full story, outcome and quiz?
Read how Airbnb executed it, the results, key lessons and test yourself with a quiz. Free on CaseLearn.
Try the full case free →Key lessons (preview)
- Two-sided marketplaces require ranking algorithms that explicitly account for both sides' long-term health, not just immediate conversion.
- Deliberate exploration (showing non-optimal results sometimes) can be necessary infrastructure for marketplace fairness.
- New supply-side participants need a mechanism to get initial traction, or a marketplace calcifies around early winners.