![[2602.14358] High Precision Audience Expansion via Extreme Classification in a Two-Sided Marketplace](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.14358] High Precision Audience Expansion via Extreme Classification in a Two-Sided Marketplace
Summary
This paper discusses a novel approach to audience expansion in a two-sided marketplace, focusing on high precision retrieval methods for Airbnb listings using extreme classification techniques.
Why It Matters
The research addresses the challenges of efficiently matching diverse guest expectations with a vast inventory of listings. By optimizing the retrieval process, the methodology could significantly enhance user experience and booking rates in platforms like Airbnb, impacting the overall effectiveness of marketplace algorithms.
Key Takeaways
- Introduces a new retrieval methodology for Airbnb listings.
- Utilizes extreme classification to enhance precision in audience targeting.
- Addresses the unique challenges of location-based search in two-sided marketplaces.
- Demonstrates the impact of rearchitecting search mechanisms on booking efficiency.
- Highlights the importance of filtering inventory effectively before ranking.
Computer Science > Information Retrieval arXiv:2602.14358 (cs) [Submitted on 16 Feb 2026] Title:High Precision Audience Expansion via Extreme Classification in a Two-Sided Marketplace Authors:Dillon Davis, Huiji Gao, Thomas Legrand, Juan Manuel Caicedo Carvajal, Malay Haldar, Kedar Bellare, Moutupsi Paul, Soumyadip Banerjee, Liwei He, Stephanie Moyerman, Sanjeev Katariya View a PDF of the paper titled High Precision Audience Expansion via Extreme Classification in a Two-Sided Marketplace, by Dillon Davis and 10 other authors View PDF HTML (experimental) Abstract:Airbnb search must balance a worldwide, highly varied supply of homes with guests whose location, amenity, style, and price expectations differ widely. Meeting those expectations hinges on an efficient retrieval stage that surfaces only the listings a guest might realistically book, before resource intensive ranking models are applied to determine the best results. Unlike many recommendation engines, our system faces a distinctive challenge, location retrieval, that sits upstream of ranking and determines which geographic areas are queried in order to filter inventory to a candidate set. The preexisting approach employs a deep bayesian bandit based system to predict a rectangular retrieval bounds area that can be used for filtering. The purpose of this paper is to demonstrate the methodology, challenges, and impact of rearchitecting search to retrieve from the subset of most bookable high precision rectangular map ...
