Single-Sample Bilateral Trade with a Broker

A team of researchers including MohammadTaghi Hajiaghayi, Gary Peng, and Suho Shin has published a groundbreaking paper titled 'Single-Sample Bilateral Trade with a Broker,' introducing a novel framework for brokerage in markets with extreme information scarcity. The work, presented at WWW'26, addresses the fundamental challenge of how a broker can facilitate efficient trade between a buyer and seller when only a single sample is available from each party's valuation distribution. This model extends previous work on bilateral trade without brokers by incorporating the three-sided interaction where a broker mediates transactions, providing crucial insights for designing robust marketplace mechanisms under severe data limitations.

The research demonstrates that remarkably simple mechanisms can achieve constant-factor approximations to first-best gains-from-trade (GFT) and social welfare (SW) under standard economic assumptions like monotone-hazard-rate distributions. The team analyzed two key settings: identical valuation distributions and stochastically ordered distributions, showing that the approximation factors remain strong even with the broker's presence. Notably, these results stand in stark contrast to previous impossibility results under strategic brokers with full distributional knowledge, highlighting how minimal data can paradoxically enable more robust brokerage. The findings have immediate implications for designing data-efficient mechanisms in online marketplaces and decentralized trading platforms where intermediaries must operate with limited information while maintaining incentive compatibility for all participants.