Auction with memory for repeated procurement

Abstract

Repetition is one of the key aspects of emerging business procurement auctions. However, very little research has been done about the characteristics of repeated auctions and the design of auction mechanisms suitable for the repeated procurements. Sustainable auction mechanisms for repeated procurement need to enable the auctioneer to gain price advantages by maintaining a sufficient number of bidders over time. For this purpose, we propose the Memory Augmented Auction (MAA) mechanism, which considers not only current bid prices, but also memory scores based on previous bid price competitiveness along with the winning rate. Under the MAA mechanism, beneficial bidders will have a better chance of winning, thus gaining motivation to join the auction market. From the empirical analysis on repeated auction data from iMarketKorea’s (IMK) MRO items, we found a substantial benefit of the MAA. Among the top 25 MRO items which occupies 77% of auctions, 5 items have significantly decreasing pattern in terms of number of bidders per auction over time. In addition, the average of successful auction rate of 11 items is as low as 70%. Thus, 16 out of 25 items that covers 43% of MRO auctions at IMK suffers from the insufficient number of participating bidders particularly after a certain number of iterations. This phenomenon demonstrates the potential benefit of the MAA approach. Furthermore, we analyzed the bidder’s bid price determination model based on the collectable data in the auction market so that we can imitate the bid pricing agents for the experiment. By applying the MAA to the B2B procurement portal site, we aim to demonstrate the decreased purchase cost and increased auction success rate. For this purpose, we implemented the MAA mechanism and conducted a comprehensive performance evaluation by comparing the performance of MAA and Traditional Auction (TA) mechanisms. We confirmed that MAA outperforms for all the goals - purchase cost, number of remaining bi...