Contextual Procurement Auctions with Bandit Learning
For platforms running repeated procurement auctions with unknown product values, this work provides the first mechanisms with provable regret guarantees and incentive properties.
The paper studies repeated contextual procurement auctions with bandit feedback, proposing mechanisms that achieve sublinear regret in welfare loss relative to full-information efficient allocation, with a tradeoff between regret and incentive error.
We study repeated contextual procurement auctions in which the platform must learn context-dependent product values from bandit feedback. We give an exactly truthful explore-then-commit mechanism with $\widetilde O((ng)^{1/3}T^{2/3})$ regret. We also give a frozen-payment UCB mechanism with a regret-incentive tradeoff: the near-UCB tuning attains \(\widetilde O(\sqrt{ngT})\) welfare regret, while for fixed \(n,g\) its total incentive error is \(\widetilde O(T^{3/4})\); the balanced tuning gives \(\widetilde O(T^{2/3})\) on both scales. Regret is measured as welfare loss relative to the full-information efficient allocation. We prove a matching lower bound for the frozen-payment regret-incentive tradeoff.