Computational feedback mechanisms for iterative multiunit multiattribute auctions

Traditionally firms aiming to achieve competition among suppliers have used sealed bidding procedures in their procurement processes. However, the advances in computational technologies now allow companies to use different and more complex auction mechanisms for their sourcing needs. The multiattribute auction is a mechanism that allows negotiation over multiple characteristics of a contract including price as well as non-price attributes. Expected gains include faster negotiation, higher market transparency, and greater allocative efficiency. In this paper, we study the problem of improving information exchange in such auctions. We develop a model of a procurement auction in which the sales item is defined by multiple attributes. We consider the case of partial preference revelation where the buyer provides some feedback to the bidders to aid in their bid formulation but does not disclose her utility function in its entirety. We propose a feedback technique that is based on bid ranks and conduct laboratory experiments to explore the impact of such feedback on bidder strategies and performance. Based on our analysis of obtained experimental data, we propose additional advanced feedback metrics.