Beyond Relevance: A Demand Balancer Model for Rental Platforms with Single-Unit Inventory (WSDM 2025)

Overview

This entry summarizes a WSDM 2025 paper: "Beyond Relevance: A Demand Balancer Model for Rental Platforms with Single-Unit Inventory."

• Publication: ACM International Conference on Web Search and Data Mining (WSDM 2025)
• Link: https://dl.acm.org/doi/abs/10.1145/3701551.3706126
• Research area: Information retrieval / vertical search / marketplace ranking

Problem Setting

Rental platforms differ from typical e-commerce search in a crucial way: each listing usually has single-unit inventory. A rental apartment, vacation home, or unique item can only be transacted once for a given period. When a ranking system optimizes purely for relevance (e.g., predicted click or booking probability), it tends to concentrate user demand on a small number of highly attractive listings. Consequences include:

• Popular listings sell out immediately, degrading user experience for late-arriving users;
• Long-tail inventory receives little exposure and fails to transact;
• Marketplace health and supply-side satisfaction deteriorate.

The paper therefore argues for going beyond relevance and proposes a demand balancer model that explicitly accounts for how search rankings shape the distribution of demand across single-unit inventory.

Why This Matters for Search and Recommendation

This work sits in the broader trend of moving ranking objectives from pure user-engagement signals toward marketplace-level objectives:

• Two-sided optimization: balancing user satisfaction with supply-side transaction success;
• Inventory-aware ranking: treating available inventory and its depletion over time as a first-class constraint;
• Position and exposure budgeting: allocating limited impressions across listings to maximize overall market outcomes rather than a single session's relevance.

Similar considerations arise in ads allocation, flight/hotel search, ticketing, and any vertical with scarcity constraints.

Reader Takeaways

• Researchers: single-unit inventory is a distinctive benchmark setting that breaks standard relevance-only assumptions; compare demand-balancing objectives against relevance baselines under inventory depletion simulations.
• Engineers: consider how inventory state, listing availability, and demand forecasts can be fed into the ranking layer, and how to evaluate such systems with online A/B tests measuring both conversion and inventory utilization.
• Product managers: pure relevance optimization can create availability failures that users experience as broken search; demand balancing trades slight short-term relevance for long-term marketplace liquidity.

Reference

• Original paper: *Beyond Relevance: A Demand Balancer Model for Rental Platforms with Single-Unit Inventory*, WSDM 2025. DOI page: https://dl.acm.org/doi/abs/10.1145/3701551.3706126

> Note: Detailed method description and quantitative results should be verified against the original PDF; this entry is based on the paper title and public metadata.