Electronic sourcing with multi-attribute auctions Stefan Strecker University of Karlsruhe Department of Economics and Business Engineering Information Management and Systems Englerstr. 14, D-76131 Karlsruhe, Germany Email: strecker@iw.uni-karlsruhe.de Stefan Seifert University of Karlsruhe Department of Economics and Business Engineering Information Management and Systems Englerstr. 14, D-76131 Karlsruhe, Germany Email: seifert@iw.uni-karlsruhe.de Abstract— Multi-attribute reverse auctions have been proposed as market institutions for “electronic request for quotation” buy- ing processes. The choice of a multi-attribute auction institution for corporate sourcing is a challenge for the auctioneer in terms of utility maximization and allocational efficiency. The authors report on a computer-based laboratory experiment in a sole sourcing scenario of a single, indivisible object and investigate whether a multi-attribute reverse English and a multi-attribute reverse Vickrey auction institution lead to identical outcomes with respect to the buyer’s utility, suppliers’ profits and allocational efficiency. The results show no significant difference in suppliers’ profits. However, the English auction institution leads to both higher allocational efficiency and buyer’s utility and is thus recommendable to corporate buyers. The documented breakdown of the outcome equivalence of the two auction institutions is attributed to bidders deviating from the dominant bidding strategy. The deviations are analyzed and explained by learning effects. I. I NTRODUCTION Corporate sourcing of heterogeneous, differentiated goods and services (henceforth, objects) is typically based on “re- quests for quotations” (RFQ). In an RFQ process, a corporate buyer announces the technical specification of the object in question, lists a number of negotiable attributes and invites potential suppliers to submit multidimensional bids on the negotiable attributes. Subsequently, the buyer evaluates the submitted bids (offers), ranks them according to her preference relation regarding the negotiable attributes and awards the contract to the supplier who has submitted the highest ranked bid. The heterogeneity and complexity of objects regularly requires corporate buyers to consider further attributes in addition to the price of the deal in the buying decision [1], e. g. object quality, lead time, terms of transportation, or warranty as well as internal criteria such as supplier reputation or incumbent switching costs [2]. The rationale underlying these procurement processes is that the buyer seeks to designate the contract to the supplier who offers the best price/performance ratio and not necessarily to the supplier who offers the lowest price, since often the lowest cost seller does not provide the best combination of price and quality [3]. On the other hand, the buyer assumes that (i) the offered objects vary in quality (object heterogeneity) and (ii) that suppliers possess comparative advantages in production costs along the quality dimensions (supplier heterogeneity) [4]. These buying processes apply to a diverse spectrum of inputs, e. g. the sourcing of contract programming [3], [5], truck fleets or coal [6]. In an RFQ-based buying process, bid evaluation and win- ner selection are labor-intensive, time-consuming, and costly which is why procurement departments are seeking to (par- tially) automate the RFQ process [2]. However, negotiations in electronic sourcing are commonly based on reverse auctions in which the price is the unique strategic dimension and all non-price attributes of an object are fixed prior to the negotiation (“price-only auctions”) [1]. Although price-only auctions have been reported to lower explicit and implicit transaction costs for the buyer [7], they potentially lead to inefficient outcomes when differences in quality exist and suppliers exhibit comparative advantages in production costs [4]. An inefficient outcome foregoes gains from trade and leaves potential value unrealized [8]. Recent advances in information and communication tech- nology facilitate the implementation of electronic auctions, which closely resemble the RFQ buying process. These multi- attribute auctions provide means to automate bid submission, bid evaluation and winner selection in terms of an “electronic request for quotation” (eRFQ) [2]. In a multi-attribute reverse auction, the buyer (bid-taker, auctioneer) specifies her value trade-offs among multiple negotiable attributes of an object by defining a scoring rule (value or utility function) based on her preference relation. During the auction process, the buyer solicits bids from invited suppliers (bidders) according to the rules of the bidding procedure. The buyer’s scoring rule is used to evaluate submitted bids and to designate the contract to the bidder providing the highest utility score to the buyer [9]. However, when designing and implementing multi-attribute auctions for the application to corporate sourcing, buyers have essentially to rely on their intuition, since (i) experience and know-how are typically missing for the lack of exemplars and (ii) because the theoretical as well as empirical findings about multi-attribute auctions are rudimentary [4]. In particular, the choice of a multi-attribute auction institution remains a particular challenge for the auctioneer as she is faced by two sometimes conflicting goals: utility maximization and allocational efficiency [10]. Obviously, the auctioneer seeks to maximize her utility. However, a sole focus on maximizing the buyer’s utility may chase suppliers away in the long Proceedings of the 37th Hawaii International Conference on System Sciences - 2004 0-7695-2056-1/04 $17.00 (C) 2004 IEEE 1