MITLIBRARIES DUPL 3 9080 02246 0544 a LIBRARIES ( I tx '^ofTB^ ^3 DEWET iD28 .M414 MIT Sloan School of Management WorkingPaper4226-01 December 2001 AN INVERSE-OPTIMIZATION-BASED AUCTION MECHANISM TO SUPPORT A MULTI-ATTRIBUTE RFQ PROCESS LawrenceM.Wein,DamianR. Beil ©2001byLawrenceM.Wein,DamianR.Beil.Allrightsreserved. Shortsectionsoftext,nottoexceedtwoparagraphs,maybequotedwithoutexplicit permissionprovidedthatfullcreditincluding©noticeisgiventothesource. Thispaperalsocanbedownloadedwithoutchargefromthe SocialScienceResearchNetworkElectronicPaperCollection: http://papers.ssm.com/paper.taf?abstract_id=295066 MASSACHu.cYTSINSTITUTE OF"^CHMOLOGY JL'M 3 2002 LIBRARIES AN INVERSE-OPTIMIZATION-BASED AUCTION MECHANISM TO SUPPORT A MULTI-ATTRIBUTE RFQ PROCESS Daniian R. Beil and Lawrence M. Wein Operations Research Center, MIT, Cambridge, MA 02142, [email protected] Sloan School ofManagement, MIT, Cambridge, MA 02142, [email protected] Abstract Weconsider a manufacturer who uses areverse, or procurement, auction to determine whichsuppUer will be awarded acontract. Each bidconsistsofapriceand asetofnon-price attributes (e.g., quality, lead time). The manufacturer is assumed to know the parametric form of the suppliers' cost functions (in terms of the non-price attributes), but has no prior information on the parameter values. We construct a multi-round open-ascending auction mechanism, where the manufacturer announces aslightly different scoring rule (i.e., afunction that ranks the bids in terms ofthe price and non-price attributes) in each round. Via inverse optimization, the manufacturer uses the bids from the first several rounds to learn the suppliers' cost functions, and then in the final round chooses a scoring rule that attempts to maximize his own utility. Under the assumption that suppliers submit their myopic best-response bids in the last round, and do not distort their bids in the earlier rounds (i.e., theychoosetheirminimum-cost bid to achieve anygivenscore), our mechanism indeed maximizes the manufacturer's utility within the open-ascending format. We also discuss several enhancements that improve the robustness ofour mechanism with respect to the model's informational and behavioral assumptions. December 6, 2001 INTRODUCTION 1. Although the market for onhne business-to-business auctions is enormous (estimated at $746B in 2004 by Kafka et al. 2000), the price-only auctions that dominate the current eCommerce landscape severely hinder the range ofproducts that can be auctioned over the Internet. In particular, within the industrial procurement setting, many low-cost standard- ized items are being transacted by current online procurement (or reverse) auctions, while high-value complex items are still being procured via the traditional Request for Quotes (RFQ) process. An RFQ process allows thesale to bedetermined by avarietyofattributes, involving not only price, but quality, lead time, contract terms, supplier reputation, and incumbent switching costs. It also lets the manufacturer reveal his preferences and permits the suppliers to compete on their own specialized dimensions. Consequently, eMarketplaces arecurrentlybeingdeveloped topartiallyautomatetheRFQ process; i.e, tocreateaneRFQ process (see Kafka et al. for examples). This paper was stimulated by about eight hours ofdiscussions (during the fall of2000 and winterof2001) with theChiefTechnologyOfficer (CTO) ofFrictionless Commerce, who was seeking help with designing a multi-attribute eRFQ mechanism. The CTO described the company's multi-attribute procurement software (we are not at liberty to discuss its details) and the perceived needs and preferences oftheir customers (i.e., the manufacturers whoowntheirsoftware) and thesuppliercompanies (i.e., thepotential bidders) withrespect tovariousaspectsofbothtraditionalandelectronicRFQprocesses. Thisinformationhelped to guide our eRFQ design and our assumptions about supplier behavior. After receiving a rough first draft of this paper, the CTO shared its ideas with several customers, and their impressions are briefly summarized in §4. The appropriate mathematical context for this setting is the multi-attribute, or mul- tidimensional, auction; consequently, we often refer to the manufacturer as the auctioneer