Auction Design for Value Maximizers with Budget and Return-on-spend Constraints
Pinyan Lu,Chenyang Xu,Ruilong Zhang
2023-07-10
Abstract:The paper designs revenue-maximizing auction mechanisms for agents who aim to maximize their total obtained values rather than the classical quasi-linear utilities. Several models have been proposed to capture the behaviors of such agents in the literature. In the paper, we consider the model where agents are subject to budget and return-on-spend constraints. The budget constraint of an agent limits the maximum payment she can afford, while the return-on-spend constraint means that the ratio of the total obtained value (return) to the total payment (spend) cannot be lower than the targeted bar set by the agent. The problem was first coined by [Balseiro et al., EC 2022]. In their work, only Bayesian mechanisms were considered. We initiate the study of the problem in the worst-case model and compare the revenue of our mechanisms to an offline optimal solution, the most ambitious benchmark. The paper distinguishes two main auction settings based on the accessibility of agents' information: fully private and partially private. In the fully private setting, an agent's valuation, budget, and target bar are all private. We show that if agents are unit-demand, constant approximation mechanisms can be obtained; while for additive agents, there exists a mechanism that achieves a constant approximation ratio under a large market assumption. The partially private setting is the setting considered in the previous work [Balseiro et al., EC 2022] where only the agents' target bars are private. We show that in this setting, the approximation ratio of the single-item auction can be further improved, and a $\Omega(1/\sqrt{n})$-approximation mechanism can be derived for additive agents.
Computer Science and Game Theory