Mechanism Design for Daily Deals

Daily deals are very popular in today’s e-commerce. In this work, we study the problem of mechanism design for a daily deal website to maximize its revenue and obtain the following results. (1) For the Bayesian setting, we first design a revenue-optimal incentive-compatible (IC) mechanism with pseudo-polynomial time complexity. Considering the high computational complexity of the mechanism, we then develop a greedy mechanism that is much more computationally efficient yet maintains a constant competitive ratio regarding the Bayesian optimal revenue in expectation. (2) For the prior-independent setting, we first propose a randomized IC mechanism with a pseudo-polynomial time complexity that can achieve a constant competitive ratio. Then, by leveraging the greedy mechanism designed for the Bayesian setting, we come up with a new mechanism that can achieve a good tradeoff between computational efficiency and competitive ratio. After that, we discuss the robustness issue regarding the two mechanisms (i.e., they both use the trick of random partition and may perform badly for the worst-case partition) and propose an effective way to guarantee a constant competitive ratio even for the worst-case partition.