1

Auction Based Mechanisms for Electronic Procurement
T. S. Chandrashekar, Y. Narahari, Charles H. Rosa, Devadatta Kulkarni, Jeffrey D. Tew, and Pankaj Dayama

Abstract— This article reviews recent research and current art in the area of auction based mechanisms for electronic procurement. These mechanisms are becoming increasingly relevant in modern day e-procurement systems since they enable a promising way of automating negotiations with suppliers and achieving the ideal goals of procurement efficiency, cost minimization, and agent based deployment. The survey delineates different representative scenarios in e-procurement where auctions can be deployed and describes the conceptual and mathematical aspects of different categories of procurement auctions. We discuss three categories: (1) multi-unit auctions for a single homogeneous type of item; (2) combinatorial procurement auctions where the buyer seeks to procure a bundle of multiple items and the suppliers bid for subsets of the bundle; and (3) multi-attribute auctions where the procurement decisions transcend cost considerations alone, to take into account lead times, logistics costs, and other important attributes. In all three cases, the winner determination problem and the determination of payments turn out to be interesting and challenging combinatorial optimization problems. In our review, we present mathematical formulation of procurement scenarios under each category, bring out the challenge involved in solving the problems, and indicate active research topics. We also present a case study of electronic procurement at General Motors. A Note to Practitioners— Since the burst of the dot.com bubble, many procurement professionals and purchasing managers have begun to question the ability of the Internet to redefine procurement processes within their firms. In this article we set out to show that this would be a misplaced sense of deja vu because the Internet along with a milieu of decision technologies based on Game Theory and Optimization is proving to be a significant tool in the hands of procurement professionals. Sans all the hype, the dot.com phenomena has left behind useful ideas including that of e-platforms for on-line auctions. Building upon this core conceptual construct, familiar to most procurement professionals, we first illustrate the successful implementation of sophisticated auction models by pioneering firms, based on optimization technologies, that meet the requirements of complex business-to-business procurement. We then discuss the exciting field of research this has opened up with a vast potential for immediate and gainful applications. We review the existing stateof-the-art in this field, track its recent developments and classify the models available for different procurement scenarios. We also provide pointers to areas that require further fundamental as well as applied research which calls for the attention of not just academic researchers but also practicing professionals.
T.S. Chandrashekar is a doctoral student and Y. Narahari is a Professor at the Department of Computer Science and Automation, Indian Institute of Science, Bangalore, INDIA and a senior member of IEEE. Email: chandra hari @csa.iisc.ernet.in Charles.H. Rosa and Devadatta Kulkarni are members of the research staff and Jeffrew D Tew is a senior member of the Research Staff at the General Motors Research and Development Center, Warren, MI, USA. Pankaj Dayama is a member of research staff at the GM India Science Lab, Bangalore, India. Email: charles.rosa datta.kulkarni jeffrey.tew pankaj.dayama @gm.com

Index Terms— E-Procurement, negotiations, auctions, multiattribute auctions, combinatorial auctions, volume discount auctions, game theory, mechanism design, NP-hard problems.

ACRONYMS
PR ICT RFQ LP IP MIP MILP GP CA GVA VCG NP FPTAS MAUT MCDA IRDA Purchase Request Information and Communication Technologies Request for Quote Linear Program Integer Program Mixed Integer Program Mixed Integer Linear Program Goal Programming Combinatorial Auction Generalized Vickrey Auction Vickrey-Clarke-Groves (mechanism) Non-Deterministic Polynomial Time Fully Polynomial Time Approximation Scheme Multi-Attribute Utility Theory Multi-Criteria Decision Analysis Iterative Reverse Dutch Auction

HE purchasing function and the associated procurement processes in organizations, large and small, have traditionally been an important area of operations affecting business performance. In many organizations, procurement costs constitute a major part of the total costs. For example, about 60 percent of the cost of a car is attributed to procurement costs. Recent trends in the business environment suggest that the importance of procurement is being both reinforced through the emergence of global supply chains and amplified by the growing incidence of outsourcing in many industrial sectors. Simultaneously, the procurement function itself has undergone much transformation. In one large global firm that the authors have worked with, decentralized, factored purchasing processes have given way to uniform, centralized purchasing practices, with worldwide purchasing decisions being coordinated by a single centralized organization. These changes can, in part, be attributed to the influence that information and communication technologies (ICT’s) have had in reshaping procurement processes both within and between organizations. The procurement process itself may be hierarchically decomposed and a first level decomposition yields four distinct stages: supplier search and analysis, supplier selection stage, automated transactional stage, and supply chain planning and control. While many aspects of procurement, in each of the four stages, have benefited from the application of ICT’s and decision technologies, negotiations, which form a crucial part

T

I. I NTRODUCTION

¡

¡

¡

¡

¡

¡

2

of the supplier selection stage have so far relied on human based processes with little technological support. With the recent advances in ICTs, including the rapid growth of the Internet, this has changed. This vector of change is being further influenced by the application of decision sciences, computational models, and software systems. Till recently, Negotiation Theory has been a topic of research within the economic and social science community. It provides a framework that is used to reach a decision through consensus whenever a person, organization, or any other entity cannot achieve its goals unilaterally [1], [2]. The increased computing and networking power enabled through the Internet has provided an opportunity to automate negotiations that happen at the supplier selection stage. The design of such automated negotiations requires a computer science and information systems perspective to feed back into the negotiation models and procedures. These interdisciplinary studies have contributed to three closely related approaches to automated negotiation systems: (1) the development of negotiation support systems; (2) software agents for negotiation, and (3) economic mechanism design and on-line auction platforms. Each of these approaches addresses the requirements of a wide variety of negotiation scenarios which is captured in [3] as integrative and distributive negotiations. For descriptions of the first two approaches we refer the reader to [4], [5], [6]. Here our focus is on economic mechanism design and auctions for electronic procurement. A. Auctions and Electronic Procurement Auctions, as a sub field of mechanism design, is concerned with the design of the rules of interaction, using the tools of game theory and mechanism design [7], [8], for economic transactions that will, in principle, yield some desired outcome. In the context of negotiations for procurement we require rules governing: (1) bidding for contracts, (2) the issues and attributes that will be considered to determine winner(s) of the contract, (3) determination of winning suppliers, and (4) the payments that will be made. English auctions, Dutch auctions, and sealed bid contracts are well understood, widely used economic mechanisms in the procurement context. Since the rules of interaction in these auctions are well laid out, they have been a natural target for automation, and as a result have formed the core conceptual constructs on which on-line auctions, like those seen in www.ebay.com, www.onsale.com, www.freemarkets.com, etc., have been based. In order to address the software requirements of such on-line auctions firms such as Ariba, Emptoris, Frictionless Commerce, etc. have appeared. These firms have now expanded their software product portfolio to include more general e-procurement tools. See http://www.research.ibm.com/absolute/ for more details on these software vendors. Also, many large corporations have now either used or are in the process of using automated auction-based negotiation methods for their procurement operations. For instance, General Electric has adopted on-line auctions for many of

its procurement operations, procuring more than $6 billion worth of goods and services in on-line auctions in 2000 [9], which led to the Internet Week magazine awarding it the title ”E-Business of the Year 2000”. Many more firms, for example, GlaxoSmithKline [10], Metro [11], Volkswagen [12], and Bechtel [13] have already started using auctions for a significant share of their procurement operations (up to 30 percent in many of them). A study conducted by the Center for Advanced Purchasing Studies [14] shows that more than a third of the firms interviewed by them are procuring goods in excess of $ 100 million through electronic procurement auctions. However, in many cases, the only information that can be gleaned about these implementations is from articles appearing in the popular press. Detailed case studies, apart from a few exceptions [15], [16], [17], [18], [19], [20], [21], [22] have been hard to come by possibly because procurement is considered to be a key source of strategic advantage and hence organizations have been unwilling to put the details of implementation into the public domain. Our concern here is with studying auctions for automated negotiations in complex B2B industrial procurement scenarios since they promise a faster emergence of high quality agreements. B. Contributions of the Paper Motivated by the key role auctions have come to play in automating negotiations in electronic procurement, we provide, in this paper, a review of the issues, mathematical formulations, and the current state of practice in this area. This paper differs from other survey articles that have appeared on related topics in the following ways. This paper is not a survey on auctions in general. There are widely popular books (for example, by Milgrom [23] and Vijay Krishna [24]) and surveys on auctions (for example, [25], [26], [27], [28], [29]) which deal with auctions in a comprehensive way. This paper is not a survey on combinatorial auctions (currently an active area of research). Exclusive surveys on combinatorial auctions include the articles by de Vries and Vohra [30], [31], Pekec and Rothkopf [32], and Narahari and Pankaj Dayama [33]. Cramton, Ausubel, and Steinberg [34] have recently brought out a comprehensive edited volume containing expository and survey articles on varied aspects of combinatorial auctions. This paper is also not a survey on dynamic pricing. There are excellent surveys on dynamic pricing [35], [36]. As is well known, auctions represent a particular mechanism for dynamic pricing. The focus of this paper is on auction based mechanisms which can be used as part of an e-procurement process to improve the efficiency of the process. It is therefore closer in its content to the papers by Elmaghraby and Keskinocak [16], Elmaghraby [37], Bichler, Davenport, Hohner, and Kalagnanam [22], and Bichler et al [36]. While the papers [16], [37] deal with combinatorial auctions for procurement together with a description of a case study, the papers [22], [36] deal with volume discount auctions and multi-attribute auctions. These papers thus specifically deal with certain special categories

¢ ¢ ¢

3

of procurement auctions. Our paper supplements and complements these above papers by taking a bird’s eye view of almost all categories of procurement auctions. The main goal of this paper, therefore, is to provide an umbrella view of the entire area in a self-sufficient way. This is evident from the structure of the rest of the paper which is presented below. C. Outline of the Paper The rest of this paper is organized into seven sections. Section II: Procurement Auctions - Scenarios and Case Summaries: In this section, we describe a range of industrial procurement scenarios followed by summaries of two industrial cases at MARS Inc. [17] and Home Depot [16], [15], where auction based procurement mechanisms have been implemented. Section III: Issues in Electronic Auctions: In this section, we provide a conceptual discussion of the key technical issues involved in the design of auction based mechanisms for electronic procurement. This section is structured as follows: [a] Types of Auctions, [b] Valuation Issues [c] Design of Auctions [d] Properties Desired from an Auction [e] Relevance of Game Theoretic and Incentive Issues [f] Computational Complexity Issues [g] Design Space for Auction Mechanisms [h] Summary. This subsection is intended to present the basic intuition behind auctions, game theoretic modeling, and economic issues. Sections IV, V, and VI: In these three sections, we discuss three major categories of procurement auctions: Procurement of single unit or multiple units of a single item based on a single attribute (Section IV) Procurement of single unit or multiple units of multiple items based on a single attribute (Section V) Procurement of multiple units of a single item based on multiple attributes (Section VI) In each subsection, we begin by discussing the main issues involved followed by a detailed discussion of two or three mechanisms from the literature. We conclude each section with a summarizing discussion on the state-of-the-art and provide a tabular listing of important current papers relevant to the topic. Section VII: A Case Study from General Motors: In this section, we discuss a real world case study of electronic procurement at General Motors. The last four authors of this paper were part of a team that developed an auction/optimization model for this procurement scenario. Section VIII: Summary and Future Work: We summarize the contents of this paper and we conclude with a discussion of some issues that need to be addressed with the current genre of models and directions for future work. II. P ROCUREMENT AUCTIONS : S CENARIOS S UMMARIES

situations: from buying the proverbial pin (very simple) to a plane (very complex). This is followed by summaries of two case studies, drawn from recent publications, to demonstrate the applicability of automated negotiations in practical business situations and to motivate the subsequent discussion of auction based negotiation mechanisms. A. Procurement Scenarios 1) Scenario A: Buying a Cutting Tool: A Purchase Request (PR) to buy a specific cutting tool is generated by a user department, such as production. This request is communicated to a central buying department called Purchasing. The buyer responsible for this category of items acknowledges the request. If the item is in one of the standard price lists of vendors already supplying to the organization and a blanket order already exists then a spot buy order is sent to the supplier. Otherwise a request for quote (RFQ) is prepared by the buyer and sent to a selection of suppliers who respond with quotes. These quotes are analyzed and a sourcing recommendation is made based on the prices that are quoted. This is a very simple buying situation where the decision to buy a single item is made on the basis of a single attribute - price per unit. Now, consider the same situation as above except that multiple units of the item are to be bought. The suppliers are now able to exploit economies of scale and hence provide bids with volume discounts thereby introducing one level of complexity in the buyer’s supplier selection or bid selection problem also sometimes called the winner determination problem. In the scenario above, very often it is observed that the same tool is required by many plants across a geographical region or across all plants worldwide. In such cases the buyer creates a blanket order which can be used by all the plants. The blanket order specifies a single price and the delivery point is the nearest pickup location operated by third party logistics providers on behalf of the manufacturer. Clearly in this case the sourcing decision involves a greater degree of complexity since the total cost of procurement - the per unit cost of the item and the shipment cost, is to be optimized by the buyer.

In this section we first describe a few practical procurement scenarios, based on the authors’ experience in working with several major procurement organizations. The scenarios progressively illustrate the range of complexity in procurement

¢ ¢ ¢

AND

C ASE

Fig. 1.

Taxonomy of procurement scenarios

4

2) Scenario B: Buying a Family of Cutting Tools: Here is an extension to the previous scenario. Purchase requests may be raised by user departments across the organization for many types of cutting tools. These requests are processed by a single buyer responsible for the procurement of cutting tools. The requests are aggregated and in some cases bundled and an RFQ is generated. The suppliers respond with bids for bundles which provide volume discounts as well as point prices. The buyer has to make a decision of allocating the orders so as to minimize his total cost of procurement as well as restrict the number of suppliers who receive the orders so as to reduce management overhead. The decision problem in this case can be challenging, and in some cases if the number of suppliers is large then the computations involved can overwhelm the buyer. 3) Scenario C: Buying a Machine Tool: While it may be possible to buy cutting tools off-the-shelf by negotiating along only one dimension - price, buying a machine tool for a specific machining requirement depends on many attributes. These may include at least the following: quality of machine tool, lead time to manufacture and deliver, availability of spares, maintenance spares to be held in inventory, etc. It may be possible to attach a monetary value to some or all of the attributes that influence the purchase decision. In some cases, additional features and options could be purchased as addon components to the basic machine tool from some or all of the suppliers. Grappling with such multi-attribute procurement decisions is the raison d’etre of purchasing departments. Figure 1 provides an idea of the range and complexity of different procurement scenarios. B. Auction Based Procurement Mechanisms: Case Summaries Compaq Computer Corp, General Dynamics, Dutch Railway, General Electric, Sears Logistics, and Staples Incorporated are a few organizations that are pioneering users of automated auction technology for procurement [38], [9], [18]. However, as indicated earlier, details of these implementations are unavailable in the public domain. Two exceptions include MARS Inc [17] and Home Depot [15], [16]. These case studies discuss the emerging state of practice in procurement and also capture the issues that arise in the trenches in a fairly detailed manner. 1) Combinatorial and Supply Curve Auctions at MARS Inc: Procurement at MARS Inc, a leading global manufacturer of confectioneries, exhibits characteristics which are not all unique to the situation. The supply pool is often small for each category of materials and contracts are executed with many types of suppliers including private businesses, traded agricultural markets, monopolies, cartels, and governments. Negotiation and tendering are the most extensively used contracting methods and the structure of typical bids in these procurement scenarios include volume discounts and all-ornothing bids. The team at MARS realized that several reasons were responsible for the inefficiency of the procurement process. Firstly, competitive positions could not be fully leveraged for price negotiations, because synergies in supply conditions

were not being fully exploited in the item by item negotiation process. Secondly, disproportionate amount of time was spent determining quantities and prices. Automated auctions, generally seen as mechanisms that make negotiations efficient and promote market competition, were sought to eliminate the limitations in the manual procurement process. But, simple auctions suffer from certain drawbacks too: firstly, they are too reliant on price for market making, making them a brute force way of managing relationships and secondly, they are inappropriate when the firm wants to control business volumes or the number of suppliers. Volume discount and combinatorial auction models provide a way of capturing business requirements such as minimizing the number of winning suppliers, limiting the exposure in business volume with a single supplier, expressing complementarities, etc. But here too, identifying the cost minimizing bid set subject to business constraints could turn out to be a NP-hard optimization problem. The Mars team working with researchers from the IBM designed an auction mechanism that would meet the business requirements and also be computationally tractable. The Auction Design: The MARS-IBM team devised an iterative auction scheme to meet the business requirements and to eliminate the need for soliciting complex bid structures. The scheme allows suppliers to correct their bids using information learned during the auction process and also induces a sense of competition among them. Each iteration proceeds in two stages: first bids are collected and a subset of bids that minimizes the cost of procurement is found. This is used as an input to the second stage where another optimization problem with the procurement cost as a hard constraint is solved. For the combinatorial auction version, the problem of choosing the winning combination of bids from the set of submitted bids, is modeled as a set-covering problem with side constraints and in the volume discount auction, it is modeled as a variation of the multiple choice knapsack problem. The supplier selection mechanism using the MIP framework was deployed using IBM’s eCommerce platform - Websphere Commerce Suite 4.1 as part of MARS Inc’s procurement web site. In order to solve this MIP formulation, a bid evaluation engine was developed as an independent module in C++ using IBM’s OSL as the LP/IP solver. This engine, using heuristic approaches based on customized knapsack cover inequalities and column generation techniques, has proven effective in solving problems with 500 items and up to 5000 bids. Approximate solutions to within one percent of the optimal were obtained within two minutes response time with these heuristics. In order to ensure a steady rate of progress of the auction in practice, bids were collected every two minutes and an allocation decision was made. This decision was communicated back to the bidders to allow them to restructure their bids. The impact of the auction mechanism on procurement processes at MARS Inc has been significant with payback being achieved in less than a year, mostly from better allocation of contracts. The intangible benefits include a fair, faster and more transparent negotiation process that has been vouched for by many of the participating suppliers.

5

2) Combinatorial Auctions for Logistics Services at Home Depot: Home Depot(HD) is a large home improvement retailer in the Americas. Managing its logistics involves coordinating over 7000 suppliers, 1000 stores, 37 distribution centers and numerous carriers. A key component of this logistics effort is the transportation of over 40000 stock keeping units (SKU’s) between different entities in the supply chain. Traditionally, the trucking services were outsourced and the contracting process was completely manual. Home Depot would provide truckers with origin destination zip codes for each pair of locations within its network and the aggregate demand forecasts for the pair. Based on this sparse information, carriers would bid for each origin-destination pair. Home Depot realized the limitations of such a contracting process. Firstly, carriers do not have good visibility into Home Depot’s network; secondly carriers could not bid on combinations of lanes to exploit potential synergies thereby limiting their ability to bid more aggressively on synergistic lanes; finally the manual process was extremely inefficient. To achieve better efficiencies and effectiveness in transportation services, Home Depot partnered with i2 Technologies, to develop and deploy an Internet based flexible auction mechanism that allowed carriers to bid for individual lanes as well as combinations of lanes. Design of the New Bidding Mechanism: The new auction mechanism consisted of three separate stages. The first is a pre-qualification stage where HD provides potential bidders with information on origin destination pairs, the lane details and the forecast demand by truck type. In return, carriers are required to provide pre-qualification information. Based on this information HD eliminates some of the carriers. Qualified carriers are then asked to send in their bids for individual lanes and combinations. The auction design was based on a single shot sealed bid combinatorial auction, in contrast to the iterative auction used by MARS Inc, to prevent carriers from engaging in damaging price wars which was seen as a potential long-term risk in maintaining the quality of service. In the second stage HD eliminates some of bids that it considers are inferior to other bids that have been received. An integer programming formulation based upon the set partitioning problem is solved in order to identify the winning combination of bids. In this stage a subset of the lanes that were originally on auction are allocated. In the third stage, HD used information from previously submitted bids to identify and invite a set of potential carriers who would be likely to submit “acceptable” bids for the unallocated lanes. The optimization engine was used again to award the remaining lanes based on the bids collected in the second round of bidding. Auction Software and Implementation Impact: The auction software consisted of three main modules: (1) Shipper bid support (SBS) module, (2) Carrier bid response (CBR) module and (3) a Bid selection optimization engine. The SBS module helps Home Depot analyze their network and decide which lanes to put out for an auction. The CBR module helps carriers analyze the demand data provide by Home Depot and create bids that reflect their cost structures and existing infrastructure. The Bid selection engine is an optimization module that solves

an underlying integer programming formulation of the auction problem. While the auction mechanism has proved to be a successful innovation, allowing Home Depot to realize better rates, the reactions of the carriers have also been positive because they can now exploit their complementarities in cost structures better through the combinatorial bidding mechanism. For more details on this case, the reader is referred to [16], [15]. III. I SSUES IN E LECTRONIC AUCTIONS As discussed in the previous section, auctions constitute a major class of economic mechanisms studied in microeconomics and game theory [24], [23]. Classical mechanism design literature has delineated several useful properties for mechanisms, that are detailed later, and the challenge in automating these mechanisms is to retain them even as we ensure computational tractability. In this section, we first briefly discuss schemes for classification of auctions and for valuations of outcomes. We then present the properties that are required from an auction along with a discussion of key possibility and impossibility results that inform their design space. We then introduce game theoretic issues followed by computational complexity issues that arise in automated auctions. We do this to build the necessary background for the subsequent discussion on procurement mechanisms. A. Types of Auctions An Auction is a mechanism to allocate a set of goods to a set of bidders on the basis of bids and asks. In a classical auction, the auctioneer wants to allocate a single item to a buyer from among a group of bidders. There are four basic types of auctions described in the literature [25], [26], [28]: Open cry or English auction, Dutch auction, first price sealed bid auction, and second price sealed bid auction (also called the Vickrey auction). An English auction is an iterative auction where the bidders submit monotonically increasing bids. This iterative process continues until a price is reached where there is just a single bidder who is willing to buy. The item is awarded to this buyer at the final bid price. The Dutch auction is the exact reverse of the English auction where the price is monotonically decreased by the auctioneer until there is a buyer who is willing to buy at the currently announced price. In both these auctions one must note that they are iterative in nature and price signals are continuously being fed back to the bidders. The first and second price sealed bid auctions however are single round auctions where bidders submit sealed bids. The winner of the contract is the bidder who submits the highest bid. The payment that he makes in the First Price sealed bid auction is the bid price itself. In the second price sealed bid auction also called the em Vickrey Auction [39], the payment that he makes is the second highest price, and this makes it a truthful auction mechanism. We will have more on this in the sections that follow. Here, we note the revenue equivalence theorem [25], [26] that states that although these auction mechanisms are vastly different from each other, they

6

yield the same expected revenue for the auctioneer when one item is being sold, under the following set of assumptions: 1) The bidders are risk neutral, in the sense that they try to maximize their expected profits and their bids avoid any situation that might give them negative payoffs. 2) The valuations of bidders follow the independent private values model (see the section on valuation issues below) 3) The bidders are symmetric (that is all of them draw their valuations from the same probability distribution) 4) Payments or prices depend only on bids A detailed review of results in Auction Theory is beyond the scope of this paper. The interested reader can refer to [25], [26], [28], and [40] for a game theoretic/microeconomic viewpoint. A computational perspective when auctions are to be automated over the Internet is provided in [41]. Building on these basic types, auctions have evolved rapidly to include multiple resources, business level constraints, and more complex market structures. Kalagnanam and Parkes [29] have suggested a framework for classifying auctions based on six major factors outlined below. (1) Resources: An auction involves a set of resources over which the negotiation is to be conducted. The resource could be a single item or multiple items, with a single or multiple units of each item. Another common consideration is the type of the item: a standard commodity or multi-attribute commodity. In the case of multi-attribute items, the agents might need to specify the non-price attributes and some utility/scoring function to trade-off across these attributes. (2) Market Structure: An auction provides a mechanism for negotiation between buyers and sellers. In forward auctions a single seller sells resources to multiple buyers. In a reverse auctions, a single buyer attempts to source resources from multiple suppliers, as is common in procurement. Auctions with multiple buyers and sellers are called double auctions or exchanges, and these are commonly used for trading securities and financial instruments and increasingly within the supply chain. In this paper, our interest is in reverse auctions which are called procurement auctions in this paper. (3) Preference Structure: The preferences define an agent’s utility for different outcomes. For example, when negotiating over multiple units agents might indicate a decreasing marginal utility for additional units. An agent’s preference structure is important when negotiation happens over attributes for an item, for designing scoring rules used to signal information, etc. (4) Bid Structure: The structure of the bids within the auction defines the flexibility with which agents can express their resource requirements. For a simple single unit, single item commodity, the bids required are simple statements of willingness to pay/accept. However, for a multi-unit identical items setting bids need to specify price and quantity. This introduces the possibility for allowing volume discounts, where a bid defines the price as a function of the quantity. With multiple items, bids may specify all-or-nothing, both with price on a basket of items. In addition, agents might wish to provide several alternative bids but restrict the choice of

bids. (5) Matching Supply to Demand: A key aspect of an auction is matching supply to demand, also referred to as market clearing, or winner determination. The main choice here is whether to use single-sourcing, in which pairs of buyers and sellers are matched, or multi-sourcing in which multiple suppliers can be matched with a single buyer, or vice-versa. This form of matching influences the complexity of winner determination, and problems range the entire spectrum from simple sorting problems to NP-hard optimization problems. (6) Information Feedback: An auction protocol may be a direct mechanism or an indirect one. In a direct mechanism such as a sealed bid auction, agents submit bids without receiving feedback, such as price signals, from the auction. In an indirect mechanism, such as an ascending-price auction, agents can adjust bids in response to information feedback from the auction. Feedback about the state of the auction is usually characterized by a price signal and a provisional allocation, and provides sufficient information about the bids of winning agents to enable an agent to redefine its bids. In complex settings, such as multi-item auctions with bundled bids, a direct mechanism can require an exponential number of bids to specify an agent’s preference information. With indirect mechanisms the focus of the design is to identify how much preference information is sufficient to achieve the desired economic properties and how to implement informationallyefficient mechanisms. In this paper, our interest is in procurement auctions with a specific attention on three broad types: [a] Single item, single attribute procurement auctions (single unit or multiunit), [b] Multi-item, single attribute procurement auctions, and [c] Single item, multi-attribute procurement auctions. B. Valuation Issues Asymmetry of information is a crucial element in any type of auction [25]. In procurement auctions, it is unreasonable to expect every bidder to possess the same amount of information. Also, there are bound to be differences among the valuations of an item or a set of items. With respect to bidders’ valuations, two extreme models which are possible are: Independent-private-values model and common-value model [25]. In the independent-private-values-model model, each bidder knows precisely how she values the item. She does not know the valuations of other bidders for this item but perceives any other bidder’s valuation as a draw from a probability distribution. Also, she knows that the other bidders regard her own valuation as being drawn from a probability distribution. Formally, bidder is associated with a probability distribution from which she draws her valuation . Bidder alone knows about and all other bidders only know the distribution . The valuations of any pair of bidders are mutually independent. In the common-value model, the item has a single objective value but nobody knows its true value. The bidders, having access to different information sources, have different estiis the item’s true value mates of the item’s valuation. If

£

¥

§

¨

£

¥

§

¥ ¦¤

¥ ¦¤

7

D. Properties Desired from an Auction We now present an intuitive discussion of properties that an auction designer looks for while designing an auction mechanism. For a detailed treatment of these, the reader is referred to [7], [8]. Solution Equilibrium: The solution of a mechanism is in equilibrium, if no agent wishes to change its bid, given the information that it has about other agents. Many types of equilibria can be computed given the assumptions about the preferences of agents (buyers and sellers), rationality, and information availability. They include: Nash equilibrium, Bayesian Nash Equilibrium, weakly dominant strategy equilibrium. Efficiency: A general criterion for evaluating a mechanism is Pareto efficiency, meaning that no agent could improve its allocation without making at least one other agent worse off. Another metric of efficiency is allocative efficiency which is achieved when the total utility of all the winners is maximized. When allocative efficiency is achieved, the resources or items are allocated to the agents who value them most. These two notions are closely related to each other; in fact, when the utility functions take a special form (such as quasi-linear form [7]), Pareto efficiency implies allocative efficiency. Individual Rationality: A mechanism is individually rational if its allocations do not make any agent worse off than had the agent not participated in the mechanism. That is, every

 W" H 4"&&&"F 4"C 1R& IgV(''6gEi4 hB   

R& W 

 W 4"&&& 4 1R& $" H )'(''" F " C 4  

R& W 

 pB

The design of auctions can be viewed as a problem of designing a mechanism that implements a social choice function. Designing a mechanism, in turn, can be viewed as a problem of designing a game with incomplete information having an equilibrium in which the required social choice function is implemented. Consider a set of agents or players with agent having a type set ( ). The type set of an agent represents the set of perceived values of an agent (also called private values). For example, in an auction, the type of an agent may refer to the valuation that the agent has for different items up for auction. Let be the Cartesian product of all the type sets of all the agents (that is is the set of all type profiles of the agents). Let be a set of outcomes. An outcome, in the context of auctions, corresponds to an assignment of auction items and payments to the bidders. A social choice function is a mapping from to which associates an outcome with every type profile. In the context of auctions, a social choice function corresponds to a desirable way of producing outcomes from given type profiles. Let denote the action set of agent , that is is the set of all actions that are available to an agent in a given of an agent is a mapping from situation. A strategy to . That is, a strategy maps each type of an agent to a specific action the agent will choose if it has that type. In an auction, a strategy corresponds to the bid the agent will place based on its observed type. Let be the Cartesian product of all the strategy sets. A mechanism is basically a tuple , where is a mapping from to . That is, maps each strategy profile into an outcome. A given mechanism can always be associated with a game with incomplete information, which is called the game induced by the mechanism. For details, refer to [7], [42]. We say that a mechanism implements a social choice function if there is an equilibrium strategy profile of the game induced by such that for all possible type profiles . That is, a mechanism implements a social choice function if there is an equilibrium of the game induced by the mechanism that yields the same outcomes as for each possible profile of types. Depending on the type of equilibrium we qualify the implementation. Two common types of implementations are dominant strategy implementation and Bayesian Nash implementation, corresponding respectively to weakly dominant strategy equilibrium and Bayesian Nash equilibrium. For definitions of these, the reader is referred to [7]. The weakly dominant strategy equilibrium is an extremely robust solution concept that ensures that the equilibrium strategy of each agent is best whatever the strategy profiles of the rest of the agents.

R& P 

C. Design of Auctions

 eB

 W" H 4"&&&"F 4"C 1R& IgV(''6gEf4  R&  W 

(unobserved), each bidder bidder , has a perceived value which is an independent draw from a probability distribution . All the bidders know the distribution . The above models represent two extremes; real-world procurement auctions will have some features of both.

¥

¥

The Bayesian Nash equilibrium is a weaker solution concept that only guarantees that the equilibrium strategy of each agent is best with respect to the equilibrium strategies of the other agents. A direct revelation mechanism corresponding to a social choice function is a mechanism of the form . That is, the strategy sets are the type sets itself and the outcome rule is the social choice function itself. A social choice function is said to be incentive compatible in dominant strategies (or strategy proof or truthfully implementable in dominant strategies) if the direct implements revelation mechanism in a weakly dominant strategy equilibrium where the equilibrium strategy of each agent is to report its true type. Similarly a social choice function is said to be Bayesian Nash incentive compatible if the direct revelation mechanism implements in a Bayesian Nash equilibrium where the equilibrium strategy of each agent is to report its true type. The revelation principle [7] states that if a function can be implemented in dominant strategies (or Bayesian Nash equilibrium), it can also be truthfully implemented in dominant strategies (or Bayesian Nash equilibrium). The revelation principle enables one to focus attention only on incentive compatible mechanisms. The mechanism design problem is to determine a mechanism that implements a ”good” social choice function. Some desirable properties which are sought from a social choice function and hence from the implementing mechanism (and in the present case, from procurement auctions) are described below [7], [42], [32].

R& W  & W   "&&&" " %H V(''!F EC `  8%H "'&(&'(UF EC dc8%H E$(''()YF Y)'C &" " W `   X H ` A " & b& & "  X F A "  ` a ` ` b ` B `  XH A " & & & " b XF A " ` 8 & Y(('() R& Y)R&a  W    P" H @"&&&" F @" C 1R& QIGV'('6G$UT@ SB   

0"&&&" #"  8)''(765£ 4 ¥ 2 0"&&&" #"  31)(''%$!! 

4

§

4

9

¥

@

9

@

4

©

£

B

£

£

@

P

¥

A

R& P  P " H @ " & & & " F @ " C 1R& QIG'(''6G$ED@  

¥

C XUA P XC EA  

@

£

 ¨ §   ¥ © ¥ @ 4 9

8

E. Relevance of Game Theory and Incentive Issues in Automated Procurement Following the description of the case studies in Section II.B and the properties desired by an auction as given in the section above, it may at first appear that there is a disconnect between theory and practice. In our opinion such an interpretation would be specious and misleading for the following reasons. Firstly, procurement professionals are acutely aware of the information asymmetry that exists in negotiations and as a result the need to bargain hard. The analysis carried

¢

agent gains a non-negative utility by being a participant in the mechanism. Budget Balance: A mechanism in the context of procurement is said to be weakly budget balanced if the sum of monetary transfers between the buyer and the sellers is nonnegative while it is said to be strongly budget balanced if this sum is zero. In other words, budget balance ensures that the mechanism or the auctioneer does not make losses. Incentive Compatibility: A mechanism is incentive compatible if the agents optimize their expected utilities by bidding their true valuations of the goods. Depending on the equilibrium achieved by truthful bidding, an incentive compatible mechanism is qualified as Bayesian Nash incentive compatible or dominant strategy incentive compatible. In the latter case, the mechanism is said to be strategy proof . If a mechanism is strategy proof, each agent’s decision depends only on its local information and it gains no advantage in expending effort to model other agents’ valuations. Solution Stability: The solution of a mechanism is stable, if there is no subset of agents that could have done better, coming to an agreement outside the mechanism. Revenue Maximization or Cost Minimization: In an auction where a seller is auctioning a set of items, the seller would like to maximize total revenue earned. On the other hand, in a procurement auction, the buyer would like to procure at minimum cost. Given the difficulty of finding equilibrium strategies, designing cost minimizing or revenue maximizing auctions is not easy. Low Transaction Costs: The buyer and sellers would like to minimize the costs of participating in auctions. Delay in concluding the auction is also a transaction cost. Fairness: This influences bidders’ willingness to participate in auctions. Winner determination algorithms, especially those based on heuristics, could lead to different sets of winners at different times. Since there could be multiple optimal solutions, different sets of winners could be produced by different algorithms. Bidders who lose even though they could have won with a different algorithm could end up feeling unfairly treated. Failure Freeness: Auction implementations should work as intended under all but the most extreme conditions. Transparency is also important because (1) it simplifies bidders’ understanding of the situation and eases their decision making (2) increases their trust in the auction process by improving their ability to verify that the auction rules have indeed been followed.

out by procurement professionals before such bargaining often implicitly involves game theoretic reasoning. Witness is provided by the plethora of auction formats that procurement professionals employ in practice. A formal game theoretic analysis only helps to place such reasoning on a firm basis. In the absence of sound game theoretic analysis, naive auction strategies may actually result in disastrous consequences. For evidence of this we point the reader to a discussion of spectrum auctions [27]. Secondly, often it is argued that human agents fail to measure up to the strict requirements of the fundamental assumptions of rationality and intelligence that game theory makes. Hence the inapplicability of game theoretic analysis. However if we extrapolate the vector of development in this area of application and assume that at some point in the not too distant future, automated negotiations carried out by software agents will become a reality then a game theoretic analysis would be eminently applicable before such systems are designed and built [43]. In summary, game theoretic considerations are already used implicitly in procurement situations. Therefore, sound use of game theoretic principles (for example, providing appropriate incentives) in designing automated procurement mechanisms will enhance the efficacy of automated procurement. We now turn to issues related to the design of appropriate incentives for procurement mechanisms based on discussions in [25], [32], [44]. Informally, a procurement mechanism describes any process that takes as inputs the bids of the sellers and determines which bidders will be allocated the item(s) and how much payment is received by the winning bidders. A mechanism is incentive compatible if the mechanism is structured in a way that each bidder finds it optimal in some sense to report his valuation truthfully. An incentive compatible mechanism induces truth revelation by the bidders by designing the payoff structure in a way that it is in the best interests of the bidders to bid truthfully. The second price sealed bid auction or the Vickrey auction for a single unit of a single item has been shown to be incentive compatible [39]. The generalized Vickrey auction is an example of an incentive compatible combinatorial auction mechanism [45], [30]. VCG (Vickrey-Clarke-Groves) mechanisms [39], [46], [47], [42] provide a broader class of incentive compatible mechanisms. These mechanisms induce truth revelation by paying a surplus amount to each winning bidder, over and above his actual bid. This surplus which is called the Vickrey Surplus is actually the extent by which the total procurement cost is decreased due to the presence of this bidder (marginal contribution of the bidder to the total cost of procurement). VCG mechanisms have very attractive properties. For example, the GVA mechanism, as already stated, is allocatively efficient, individual rational, weakly budget balanced, and incentive compatible. However these mechanisms are not commonly used because the payment of Vickrey surpluses or Vickrey discounts make them revenue inefficient from the point of view of the buyer. Secondly, the computation of

¢

9

G. Design Space for Auction Mechanisms Ideally, one would like to put in place a mechanism that has all the properties required in an auction mechanism while simultaneously achieving computational tractability. Unfortunately this may not always be possible. A set of results in mechanism design theory dealing with what combinations of properties are possible and what are not possible to be achieved by an economic mechanism such as an auction provide the design points around which mechanisms need to be designed. We indicate below some of the important results germane to the discussion on procurement mechanisms. Impossibility Results: Hurwicz [48] showed that it is impossible to achieve allocative efficiency, weak budget balance, and individual rationality in a Bayesian Nash incentive compatible mechanism. According to Arrow [49], allocative efficiency and strong budget balance cannot be achieved simultaneously in a dominant strategy equilibrium.

H. Summary In this section, we have discussed important issues in auctions: valuation issues, design of auction mechanisms, properties desired from a procurement auction, relevance of game theoretic and incentive issues in electronic procurement, computational complexity issues, and design space for auction mechanisms. These issues are extremely relevant for electronic procurement auctions. We will be referring to a few important properties (allocative efficiency, cost minimization, and incentive compatibility) while discussing procurement mechanisms in subsequent sections. We will also touch upon complexity issues while discussing different procurement mechanisms. IV. S INGLE I TEM , M ULTI -U NIT, S INGLE ATTRIBUTE P ROCUREMENT M ECHANISMS In this section, we discuss models related to the procurement of a single item with price as the only consideration for

¢

¢

¢

F. Computational Complexity Issues in Auction Design In an economic mechanism where resource allocation is done based on decentralized information, computations are involved at two levels: first, at the agent level and secondly at the mechanism level. The complexity questions involved are briefly indicated below. For a more detailed discussion refer [42], [29]. Complexity at the Agent Level: Strategic Complexity: Should agents model other agents and solve game theoretic problems to compute an optimal strategy? For instance, in a sealed bid procurement contract scenario, sellers will need to not only take their valuation of the contracts into consideration but also the bidding behavior of their competitors. This requires sophisticated bidding capability. Valuation Complexity: How much computation is required to provide preference information within a mechanism? For instance, in a multi item procurement scenario where the items exhibit cost complementarities, estimating a bid for every possible permutation of the bundle of items is costly. Complexity at the Mechanism Level: Winner Determination Complexity: How much computation is expected of the mechanism infrastructure to compute an outcome given the bid information of the agents. Communication Complexity: How much communication is required between agents and the mechanism to compute an outcome. For instance, in an English auction, where individual valuations are revealed progressively in an iterative manner, the communication costs could be high if the auction were conducted in a distributed manner over space and/or time.

¢

Vickrey surpluses and Vickrey discounts involves solving as many problems as the number of winning bidders which in most cases turn out to be NP-hard problems.

Myerson and Satterthwaite [50] showed that no exchange (that is with multiple sellers and multiple buyers) can be efficient, budget balanced, and individual rational at the same time; this holds with or without incentive compatibility. Possibility Results: It was shown by Groves [47] and Clarke [46] that allocatively efficient and strategy proof mechanisms are possible if the utility functions are quasi-linear (that is, of the form utility equals value minus price). Clarke mechanisms are a special class of Groves mechanisms. The generalized Vickrey auction (GVA) [43] is a combinatorial auction version of Clarke’s mechanisms while the Vickrey auction [39] (second price sealed bid auction of a single indivisible item) is a special case of GVA for noncombinatorial auctions. In fact, GVA satisfies four properties simultaneously: allocative efficiency, individual rationality, weak budget balance, and strategy proofness. All the mechanisms above are also commonly referred to as VCG (Vickrey-Clarke-Groves) mechanisms. It was shown by Arrow [49] and d’Aspremont and Gerard-Verat [51] that under quasi-linear preferences, it is possible to have a mechanism (which is called the dAGVA mechanism) that is efficient, Bayesian Nash incentive compatible, strongly budget balanced, and individually rational. It was shown by Myerson [52] that revenue maximization, individual rationality, and incentive compatibility can be achieved simultaneously. For more details on these results, we refer the reader to [42], [7], [53], [24], [29]. The above discussion gives an idea of how the design space for auctions is constrained. Another key factor that constrains the design space further is the computational complexity (discussed in Section III.F) at the agent level and at the mechanism level. The central problem of a designer of procurement auctions faced with a given procurement situation is to come up with the best possible mechanism that is contained in this constrained design space.

¢ ¢ ¢ ¢ ¢ ¢

10

The winner determination here is fairly straightforward if there are no business constraints such as maximum/minimum number of units that are to be purchased from a supplier, minimum/maximum number of winning suppliers, etc. The computational problem is to simply find the supplier who offers the best price and buy the entire quantity from this seller. However, procurement problems in practice rarely are without business constraints. See [55] for a comprehensive tabulation of business constraints that are observed in practice. With inclusion of business constraints, the winner determination problem becomes a mixed integer programming problem (MIP) [17], [56]. In the following subsection, we first present the mathematical formulation of a volume discount auction problem adapted from [54], [56] and then discuss the computational issues that arise in determining the winning bids. 1) Mathematical formulation of volume discount auctions: The volume discount procurement auction is represented in the following way [54], [56]. Table I provides the notation. The buyer needs to procure a quantity of an item. The buyer identifies a list of potential suppliers who can bid in the auction. Each supplier responds with a bid composed of a supply curve. A supply curve from supplier given by a bid consists of a list of price quantity pairs, price quantity pair specifies the price that the supplier charges per unit of the item if the number of units bought from this supplier is within the interval . It is assumed that the quantity intervals for the supply curve are all pairwise disjoint. Also, note that different unit prices are used for different ranges within the overall quantity (that is, if a quantity spans multiple intervals, the unit price for different spanned intervals will be taken as the designated unit prices in the intervals). The MIP formulation is as follows: A decision variable is associated with each pricefor each bid . quantity pair This is a variable taking on the value 1 if we buy some number of units within the price range and 0 otherwise.

B. Volume Discount Auctions In a procurement context when a single buyer and multiple sellers who wish to exploit scale economies are present, a volume discount auction is appropriate. Here suppliers provide bids as a function of the quantity that is being purchased [17], [54]. The winner determination problem for this type of auction mechanism is to select a set of winning bids, where for each bid we select a price and quantity so that the total demand of the buyer is satisfied at minimum cost.

€ 8n i

€ 8n

n m

q

.

Each

 kj

" 8v st gVii (%~'n wu } |n r" } |

j

i

z x

y v¥

 Ux sx 8i" 1v st 8ii (!$n €n i wu €n r" € z y v¥ q 2 Ux sx ~V i } |n{ z "&&&" z s n i wus n r V(''V 6x x )C i" 1vt )C igiy ¥ (" C n  q y v¥ n { 5p n q om a

‡†„ … € sx 8n i$"ƒ1u‚ st 8n i (" 8n w € r € q a €n 8G

z 6x

y v¥

€n i wu €n sx 8i" 8v st 8 ri

l"&&&" $V'('!

‡ ˆ h ‡ ˆ f

¢

¢

¢

In many procurement scenarios for a single unit of an item, the English auction or a sealed bid tender process is generally used to decide the winner of the contract. In the later case the winning rule is generally a first price rule. The second price auction is rarely used because bidders feel that the information may be used to unfair advantage of the buyer. However, in Internet based buying, where anonymity of buyers and sellers and non disclosure of prices may be assured, the second price auction may yet be gainfully employed. While, we have seen from the revenue equivalence theorem, that the expected revenues from all the basic auctions may be the same, there are significantly different implications when the auctions are to be automated. We summarize the differences below. 1) Computational Implications of the four Basic Auction Mechanisms: As indicated in the previous section, complexity can be analyzed at the level of the agent - strategic and valuation complexities, and the mechanism - winner determination and communication complexities. Strategic Complexity: Clearly for an agent participating in a Vickrey auction, which is incentive compatible, the strategic complexity is reduced to just bidding one’s true valuation without any consideration for the strategies that would be followed by other agents. However, agents in the English, Dutch, and first price sealed bid auctions require to base their bids not only on their private valuations but also condition them on the valuations of their competitors. This requires them to process additional information - the number of competitors, the probability distributions of their valuations, etc. which increases their computational overhead. The English auction has certain benefits relative to the Dutch and First Price Sealed auctions. In an English auction, if all bidders bid up to their true valuations, no single bidder can gain by unilaterally deviating from this truthful strategy. Communication Complexity: The communication overhead, and hence the processing overhead, is clearly much higher in multi round mechanisms like the English and Dutch auctions as compared to single shot mechanisms such as the Vickrey and first price sealed bid mechanisms.

‰ ‘'ƒ

A. Procurement Auctions for Single Unit of an Item

 ™˜ ‡ q – “ ‡ e $g) “8ˆ —x 1 ”• 8ˆ q ’ e(d™) “8ˆ —x –1 ”• 8ˆ q ’ ˜ ‡ q “‡ s

…‚ xQQy8Qx)„t ƒ y y v  € s r Q8Q1)wus xyyyxv t

‡ˆ%†  ƒ  ‚

 T€

q r s ¢

decision making under a variety of different business contexts. Specifically, we review: (a) auction mechanisms for a single unit of an item, (b) auctions for multiple units of a single item with volume discount bids, and (c) auction mechanisms for procuring multiple units of a single item for multiple manufacturing points taking logistics costs into account.

TABLE I N OTATION FOR VOLUME DISCOUNT AUCTIONS quantity of item number of suppliers index for the suppliers ( ) supply curve (bid) from supplier number of price-quantity pairs in bid index for price-quantity pairs, unit price the supplier charges if the number of units bought from this supplier is within the interval decision variable that takes value 1 if the buyer buys a quantity in the range a continuous variable that specifies the exact number of units bought

11

A continuous variable is associated with each pricequantity pair, which specifies the exact number of units of the item that are purchased from the bid within this price-quantity pair. The formulation is (see Table I):

proposes a variety of exact and heuristic techniques to solve this class of problems.

subject to:

The coefficient

In the formulation provided by [54], it is assumed that types of items are being bought, which is a more generalized problem. However even with a single item, the MIP formulation is NP-hard. Additional side constraints such as a limit on the number of winning suppliers and quantity constraints at the level of the supplier increase the complexity of the decision problem. 2) Computational Issues: The model is developed to address the winner determination problem with the implicit assumption that agents participating in the auction have a bidding strategy in place, perhaps with the use of game theoretic analysis. This being the context, we restrict our attention to computational issues that arise at the mechanism level and neglect analysis of any computational issues at the agent (bidder) level. The MIP formulation in the previous section is a variation of the multiple choice knapsack problem which is -Hard. Although Knapsack Problems, from a theoretical point of view, are almost intractable as they belong to the family of -hard problems, several of the problems may be solved to optimality in fractions of a second [57] and can be considered as the simplest among -Hard problems in combinatorial optimization. This is because the knapsack problem exhibits special structural properties which makes it easy to solve. Dynamic programming techniques, branch-andbound algorithms and polynomial time approximation schemes have been proposed to solve knapsack problems. For a detailed exposition of several of these methods refer to [57]. Kameshwaran [58] shows that multi-unit procurement with volume discount bids leads to piecewise linear knapsack problems and

n 8'('6›ˆS1('(6™˜G—„ i8n Q”1v st 8n “’x • €  wu € i … p&&&"  š" l&&&"  j –



C t  8vt ‘t ii©… x x ‘t ii t n  € ~$n ‘ wus n s n y g¥  q Ž ¨ C V€ €n $u‘ n $V&('("6›7Q"Sl8''&'(6¦…œ—Ÿ 8n ˆ p" && š – "& &"  j – „ € &§np$"'&''("6›7Q"S$"V'&('6¦…jœ–¥6'¤£¢i8  && š – l & &"  2 " „ ¡ €n   C 8€ C Vn

is a constant and computed a priori as: (5)

n‰

€n ‘€ 8uY$n 

& l"&&&"  j – S8'''(6›…œ • 8I €n

i  wu €n i€n Ÿž 8v st 8iY8I  8Eˆ €n

q



  C 8€ C Vn

} g| Ž  Ž

 € € u8n 8n ˆ

€ $n ˆ q

  C 8€ C )n

} g| Ž  Ž



} g| Ž  Ž

q

 C 8€

} ~| Ž



Œ‹ ‚…Š

q

y g¥



€ € sx 8n i … $n ˆ 

¢

C. Single Item, Multi-Unit Procurement to Minimize Total Cost of Procurement (Supply Chain Auction) In the previous subsections, we examined auction mechanisms in automating supplier selection situations which coincide with a fairly operational view of procurement. When the supplier selection process within a supply chain setting is viewed, at a higher level of granularity, through a strategic lens, a procurement manager is concerned with a larger set of issues: What is the impact of a sourcing decision on the total cost of procurement? How does one structure the sourcing pool so that the total cost of procurement is minimized? A partial list of important cost components that contribute to the total cost of procurement could include: price per unit, logistics cost, inventory holding costs, and lead time costs. Classical auction literature, however, has focused on price and ignored the impact of other costs on the sourcing decision. A first step in incorporating these other cost components while making a strategic sourcing decision is taken by Chen, Janakiraman, Roundy, and Zhang [59]. They provide the design of single item, multi-unit auction mechanisms that achieve overall supply chain efficiency while taking into account production and transportation costs. The concern is to bring together the business requirements within a global supply chain, economic/game theoretic desiderata, and computational issues in building the decision model. We call this the supply chain auction. The procurement problem addressed is as follows: A buyer has requirements, called consumption quantities, for a certain component at a set of geographically diverse locations. It is assumed that the buyer has private valuations of consumption quantities at the demand locations, which forms the consumption vector and that she will act strategically to maximize her utility. Multiple suppliers, each owning a set of production facilities, are available to satisfy this demand. Every supplier has a production cost, which can be described by a convex function of the quantities that he produces at his production locations. It is also assumed that each supplier is a rational, self interested player who is trying to maximize his payoff (payment received minus the production cost). In addition the buyer knows the per unit transportation cost along each link in the supply network and she pays for all shipments along the links. A key requirement in such resource allocation problems is to ensure that the allocation is efficient, in the sense that contracts are allocated to suppliers who provide the best value. To do this the buyer needs to know the production costs of the suppliers, truthfully. So, efficient allocation and incentive compatibility are two of the key desiderata. There are three key mechanisms for multi unit auctions: Pay-as-you-bid, uniformprice, and VCG auctions. From auction literature, it is known that VCG based mechanisms are both incentive compatible and efficient. The auction models presented below for supply chain procurement essentially invoke the VCG structure.

(1)

(2) (3)

(4)

12

TABLE II N OTATION FOR AUCTION T, AUCTION R
AND

(10)

(6)

subject to:

(7) (8) (9)

A Vickrey based payment rule, belonging to the more general truth-inducing VCG family described in [60], is used. The rule, in words, essentially is:



Ñ {

X på



Ñ {



aåÑ 

æ

1) Model Formulation: Chen, Janakiraman, Roundy, and Zhang [59] present three separate auction mechanisms - Auction T, Auction R, and Auction S. The first two mechanisms incorporate transportation costs explicitly whereas in the third model, only production costs are considered for the allocation decision and transportation decisions are made subsequently. This provides a means to compare the two approaches and the implications for total cost of procurement strategies. Before we present the models, the notation used is introduced in Table II. 2) Auction T: In this auction mechanism, the buyer submits a fixed consumption vector q to the auctioneer. Supplier k submits to the auctioneer a bid function for supplying units, for which he incurs a production cost , . The supplier may or may not see the consumption vector. The auctioneer decides the quantities awarded to each supplier, and the amounts transported from suppliers’ production centers to buyer’s demand locations by solving the following winner determination problem.

While it is true that the payment rule induces rational suppliers to bid their true costs, = , irrespective of other suppliers’ bids, it also results in adverse effects for the buyer. In situations where the production capacity is tight, and the production costs are sharply convex, the contribution that a supplier makes to the system can be significantly high resulting in disproportionately high payments even when production is at a low cost. Alternately, as pointed out by the authors, when suppliers’ capacities are asymmetric, removing a supplier to solve the optimization problem may result in computing an infinite value for the contribution that is made, which is an undesirable consequence. To rectify these problems, the authors propose Auction R. 3) Auction R: In this auction mechanism, the winner determination (optimization) problem solved by the auctioneer retains the essential structure of Auction T except that the payment rule is modified. This modification results in much lower payments, compared to Auction T, by the buyer while simultaneously ensuring that the sellers bid their true costs. In essence, incentive compatibility is retained. The key idea introduced to achieve this result is as follows: , which is a Here, the buyer submits a utility function proxy for the true consumption utility , in place of a consumption vector as in Auction T. The auctioneer solves a modified version of the winner determination (optimization) problem using a new payment rule. This rule follows the VCG payment structure but also incorporates the utility function in computing the payments. The utility function essentially serves the role of a reservation price function for each unit of the item that the buyer might acquire. While in most auctions the payment made to a supplier is based on bids from all suppliers, in Auction R, part of the payment is determined by the proxy utility function acting as a reservation price function. This prevents unusually large payments. There is however a price to pay for this improvement. In order to submit an optimal utility function , the buyer needs to know the suppliers’ production cost functions. However this may not hold in reality. So the buyer will have to expend effort to at least get a probabilistic belief of the cost functions. This causes uncertainty in both payments and consumption quantities. Numerical examples to illustrate this effect are provided in [59].



Þ

ß 1"

„ à n

Þ¦b Ýܙ Á 2 Û Ú

Á

 !n

 n

Á â b

Á b

n ‘

n “À  

Ñ a

Ð 8"

 !n

Ø Ö Ñ Ô Ù×cÕÑ

 G¨ Ð Ñ a n

Ñ {

Á b

n ~À

Ð … „ä

Ð  Ýã n Ñ n a G¨ Ñ â { ƒá

j

  eÓÒ



Ñ {

Ð

Ò ©¡

Ñ

number of suppliers index for suppliers, total number of supplier production facilities index for production facilities, total number of buyer locations index for buyer locations, set of production facilities owned by supplier k index for supplier that owns production facility n consumption at demand center m, production quantity at production facility n vector of size quantity shipped from the production facility to demand center cost of transporting unit quantity from production facility to demand center , total quantity shipped to demand center by supplier production cost function for supplier ( ) bidding function from supplier ( )

j

j

j

AUCTION S

Bonus payment on + Bid of account of the value vendor that the vendor adds to the system by participating in the auction. That is, if is the optimal value of the objective function for a given q; and we restrict to ensure sufficient supply capacity. If is an optimal solution, and is the optimal value of the objective function with the additional constraint that the ), the supplier k does not participate in the auction ( buyer will pay supplier :

Payment to vendor

=



Ñ a

å

¡ n

Á

 Un

Á b

& "&&&"  0Í p"&&&"  Ë" „ h8'''(6¦h8Î$V(''!™Ì8%Ÿ Å H È

‚ QQ8Q)„t ¬ xyyyxv ª 1QQ8)wt « xyyyxv n ‘ ± ° ¯ c$® ¬ Å H ~Å H È  Un s Á a n gÀ s s |Ž

Í p"&&&"  Ë É ÎV'('!™Ì8" Å Ê Å H È

Í "&&&"  0" H  Ï$V'('6›S1I’Ê Å H È

r QQ8Q)„us xyyyxv t Æ ÇC  Å C  H Ď ¬

«

³ ª³ 

 Un 

¬

Á a

¸· ¯ ­ ´ } EU­ ¶

¼ ™¾ ½ ¸ ½ ¼ } ¼ ¾ ½ ™Y½ ¸ f¼ }

n “À

 C Vn

C  Å



|Ž C H

Ž

Ď

«

Œ‹ ‚…Š

» Y ² ºU ¹ ¯h ¯­µ ¯­´ ² ­ f ¯ 8® ­ Ys  ª¬ »  ²º  ¿ à } Ä w Ãn Á

r ª ‚ s

«

13

(11)

subject to:

(12)

(13)

(14)

D. Summary and Current Art In this section we reviewed models proposed to handle the procurement of both single unit and multiple units of a single item, with a single attribute - price as the criterion for decision making. In the first model, we illustrated the game theoretic considerations that influence the bidding behavior of agents. In the second model, in the absence of game theoretic requirements, we pointed out the computational complexity of the winner determination problem in a more complex sourcing situation. The third model brought together the game theoretic as well as computational issues that complicate the sourcing problem. This was done to illustrate the fact that in the absence of proper mechanism design, the buyer could end up leaving money on the table. In Section II, we have already presented the case summary of MARS inc. [17] which shows the successful application of single item, multi-unit procurement auction with volume discount bids from suppliers in a real-world setting. The computational challenges involved in the winner determination problem there are described in [56], [54]. Kameshwaran [58], in his recent work, has shown that single item, single attribute, multi-unit procurement with volume discount bids leads to piecewise linear knapsack problems to be solved. In his work, Kameshwaran has developed several algorithms (exact, heuristic-based, and fully polynomial time approximation schemes) for solving such knapsack problems. Kothari, Parkes, and Suri [68] consider single-item, single attribute, multi-unit procurement auctions where the bidders use marginal-decreasing, piecewise constant functions to bid for goods. The objective is to minimize cost for the buyer. It is shown that the winner determination problem is a generalization of the classical 0/1 knapsack problem, and hence NP-hard. Computing VCG payments also is addressed. The

If is the optimal objective value, the production vector, and the optimal objective value without supplier , the buyer’s payment to supplier k is given by the rule:

(15)

We can observe that this payment rule still retains the VCG structure, and hence the suppliers will submit their true cost functions. Subsequently the transportation costs are determined by solving the following optimization problem: Transportation Problem:

(16)

subject to:

(17)

(18)

, is given by . The numerical experiments clearly show that Auction S minimizes total production costs but leads to higher supply chain costs than Auctions T and R.

 í

ç ì



The buyers total outflow,

 Y  l



4) Auction S: In order to compare the cost savings, if any, that could be achieved by taking an integrated view of sourcing decisions, the authors formulate Auction S. Here the buyer’s sourcing decision is based only upon the bids submitted by the sellers, and the transportation costs are determined subsequently. The buyer submits a consumption vector q as in Auction T. The auctioneer solves two optimization problems, one to determine allocation and payments and the other to optimize transportation costs, separately. Winner Determination Problem:

5) Computational issues: The crucial contribution of this model has been to take an integrated view of the sourcing problem by combining the pricing and the transportation decisions. The mechanism incorporates game theoretic considerations in supply chain formations when products are sourced globally for demand points that are distributed geographically. Since the mechanism is incentive compatible for the suppliers, the complexity of evolving an optimal bid strategy is simply reduced to reporting the actual production costs, thereby eliminating the need for complex modeling of competitors’ behavior. At the mechanism level however, optimization problems need to be solved to decide the allocations and payments. Further, since the mechanism is not incentive compatible for the buyer, the buying agent needs to solve an optimization problem to decide an optimal bidding strategy. Such an optimization problem is known to be challenging and hence a numerical approach may be warranted. Also, from the analysis presented in this paper, it is clear that in designing auction mechanisms for complex supply chains, focus on truth revealing auction mechanisms is to be augmented by pragmatic considerations of limiting the total payout of the buyers. This can be achieved by using the idea of a reservation price suitably expressed as a utility function.

Ñ {

j ç Ð  Ñ n a G¨  IÐ ç ç Á Ñ { & "&&&"  0Í p"&&&"  Ë" „ h8'''(6¦h8Î$V(''!™Ì8%Ÿ Å H È

 çn n À   ¢3 G¨ ç êé n ç Ð … Ð  Á b Ñ { Ñ a n Ñ è a uá

Í p"&&&"  Ë É ÎV'('!™Ì8" Å Ê Å H È

Í p"&&&"  Ë É ÎV'('!™Ì8" Å Ê Å H È

Í "&&&"  0  Ï8'''(6¦h8" çH Ý Å H È

Í "&&&"  0  Ï$V'('6›S1" H Ê Å H È

Å H È ÝÅ H ë

 Un

Æ  Å ç H È îÅ H C  |Å ¶ C  H ¶   { ç n C )n ¶ ë Ñ á  Ä ç ì Ñ a

Á b

n “À

Æ C  Å C H

 C )n |Ž

 Ž

Ď

Œ‹ v˜Š Œ‹ v˜Š C  Å C  Å C H C H |Ž |Ž Ď Ď

14

Scenarios in Single Item, Single Attribute Procurement Basic Auction Types Volume Discount Auction Supply Chain Auction

Math.Programming Models

Game Theoretic Analysis [5], [23], [52], [61] [41], [64] [66], [17], [67], [68] [59], [69] TABLE III

[58], [65], [65]

Case Studies and Surveys [62], [14], [10], [11] [12], [13], [3], [4] [23], [38], [40] [17]

Other Relevant References [25], [26], [27], [63] [29], [28], [24]

C LASSIFICATION OF S INGLE I TEM , S INGLE ATTRIBUTE P ROCUREMENT AUCTIONS

authors provide a fully polynomial time approximation scheme (FPTAS) for the generalized knapsack problem. This leads to an FPTAS algorithm for allocation in the auction which is approximately strategy proof and approximately efficient [68]. It is also shown that VCG payments for the auctions can be computed in worst-case time, where is the running time to compute a solution to the allocation problem. Dang and Jennings [67] consider multi-unit auctions where the bids are piece-wise linear curves. Algorithms are provided for solving the winner determination problem. In the case of multi-unit, single-item auctions, the complexity of the clearing where is the number of bidders algorithm is and is an upper bound on the number of segments of the piecewise linear pricing functions. The clearing algorithm therefore has exponential complexity in the number of bids. To summarize, the two main issues in single item, multiunit, single attribute auctions are (1) to reduce the complexity of the winner determination problem and (2) to make the mechanism strategy proof. Table III provides an appropriate taxonomy for this type of auctions and gives a listing of relevant papers under the various categories. In the next section, we review mechanisms that enable the procurement of multiple items with a single attribute (namely price) as the criterion for decision making. V. M ULTI -I TEM , S INGLE ATTRIBUTE P ROCUREMENT M ECHANISMS In the previous sections, we reviewed mechanisms that support single item procurement. In one of the mechanisms, we discussed the inclusion of logistics costs as an additional element influencing the procurement decision. Clearly such supply chain criteria are of crucial importance to procurement managers responsible for large global supply chains. Typically purchase requests within such large purchasing organizations provide opportunities to exploit complementarities in logistics costs and often in production costs too. For instance, in one of the negotiation scenarios depicted in Section II.A, a family of cutting tools may be procured by a series of sequential reverse auctions, one for each of the items. This has at least two consequences: Firstly, the price that a supplier may be willing to offer may depend in complicated ways on what other items he wins, and since there is uncertainty associated with this, the suppliers have no incentive to bid aggressively. Secondly, the suppliers do not have an opportunity to express their unique complementarities in production costs and hence the buyer cannot exploit the economies of scope associated

with bundling the demands. In such cases it may be beneficial to allow the suppliers to bid on combinations of items rather than on single items. Such auctions are called combinatorial auctions. Simply defined, a combinatorial auction (CA) is a mechanism where bidders can submit bids on combinations of items. The winner determination problem is to select a winning set of bids such that each item to be bought is included in at least one of the selected bids, and the total cost of procurement is minimized. In this section we first present the crucial issues that arise when CA’s are applied to procurement settings. We then review approaches to modeling multi-item procurement in supply chain settings which take business and capacity level constraints into consideration. A. Issues in Multi-Item Procurement Important issues in combinatorial auctions are well surveyed in [30], [32], [16], [37], [33]. In the multi-item procurement case, when suppliers are allowed to respond with combinatorial bids, the winner determination problem becomes a weighted set covering problem, which is known to be NP-hard. In addition, since the bidders can submit bids on combinations of items, the representation of the bid is also a crucial implementation issue. Bidding language issues are discussed in [70], [71]. As opposed to single item procurement scenarios, in multiitem procurement scenarios, the range of issues to be considered is vastly expanded because practical business level issues and constraints need to be included in the decision model. These could include exclusion constraints (for example, item A cannot be procured from supplier X), aggregation constraints (for example, at least two and at most five suppliers need to be selected for goods of category B), and exposure constraints (for example, not more than 25% of the procurement value should be assigned to any one supplier) [72], [17]. By including these constraints in the decision model, a sensitivity analysis of side constraints may also prove to be a useful tactical input to procurement planners. B. A Single Round Combinatorial Procurement Mechanism As indicated above, the reverse combinatorial auction problem is a set covering problem, which is NP-hard. Additional side constraints make a fundamental impact on the problem. Even finding a feasible solution when exposure constraints are in place is NP-hard. The basic formulation of the problem and the additional side constraints are presented below and solution

ð

 D0

ó òñ ð !vha 0

ï

 Y  @ 0 ï  H  

@

15

C. An Iterative Reverse Dutch Auction for Combinatorial Procurement In the previous section, we discussed a purely computational view of the multi-item procurement problem without regard to economic desiderata. It is however important for a variety of reasons to lend credence to economic issues in the modeling and analysis of multi-item procurement. Firstly, we would like to ensure that procurement contracts are allocated to those that value them most. In the absence of economic analysis, the game is open for suppliers to indulge in strategic bidding behavior in an effort to extract the maximum possible surplus utility from the negotiation. This is especially true when single shot mechanisms are used. Secondly, even if multi-round combinatorial auctions are used where allocation and pricing information is disclosed at the end of each round, it is not clear to suppliers as to how they need to reformulate their bids. The two issues together may prompt wasteful usage of resources by each supplier in trying to outsmart the buyer and other suppliers. One way to amend the situation is to design an incentive compatible mechanism, that is. a mechanism which provides incentives for truthful bidding thereby making it the dominant strategy for all participants in the mechanism. Generalized Vickrey Auctions (GVA) generalize the single item second price auction (Vickrey auction) proposed in [39]. This is an incentive compatible mechanism for combinatorial auctions, which can be applied to the procurement context [73]. However, one issue that needs to considered is that the GVA requires an optimal solution to the allocation problem which is NP-hard. In the absence of an optimal solution to the allocation problem, the resulting mechanism, whose allocation is obtained through some approximation scheme, is no longer incentive compatible [60]. However, iterative algorithms for

(19)

(20)

(21)

(22)

(23)

(24)

¢

techniques are discussed thereafter. The following formulation is from [54], [17]. Table IV provides the notation. 1) Mathematical Formulation: For each item type there is a demand . Each supplier is allowed up to bids indexed by . Associated with each bid is a zero-one vector , where if will supply the entire lot corresponding to item , and zero otherwise. Associated with each bid is price at which the bidder is willing to supply the combination of items in the bid. A mixed integer programming (MIP) formulation can be written as follows:

H Å @ ¥

j n È  ¤¢Å s n $œH Å s n å å" £¡ ¥ 

£¡ ¤¢Å @

¢

total number of items to be procured index for items, number of units of item demanded number of suppliers index for suppliers, maximum number of bids allowed for any supplier index for a bid, bid of supplier 0-1 variable which takes value 1 iff will supply the entire lot corresponding to item price associated with bid minimum quantity that should be allocated to supplier maximum quantity that can be allocated to supplier decision variable that takes value 1 iff is allocated indicator variable that takes value 1 iff supplier is allocated any lot minimum number of winners required maximum number of winners allowed

j

N OTATION FOR COMBINATORIAL PROCUREMENT

and are the minimum and maximum quantities that can be allocated to any supplier ; Constraints (21) and (22) restrict the total allocation to any supplier to . is an indicator variable that lie within takes the value 1 if supplier is allocated any lot. and are respectively the minimum and maximum number of winners required for the allocation and constraint (24) restricts the winners to be within that range. 2) Solution Approach and Discussion: Although the decision model is NP-hard, integer programming techniques are known to be effective in solving problems with 500 items and up to 5000 bids [17]. For the purpose of this paper, the authors use IBM’s OSL (Optimization Solutions and Library) to solve the IP formulation. From the experiments that were conducted, the following conclusions could be derived: Firstly, varying the aggregation constraint seems to have a large impact on the computation time. This is because, as we commented earlier, the problem of finding a feasible solution itself is NP-complete. Secondly, the min-max exposure constraints also have a significant effect on the computational time. In some sense, the exposure constraint could itself be a proxy for the aggregation constraint and hence the behavior appears similar.

£¡ ¤¢Å s n å

H Ås n å ¥

TABLE IV

€ $n þ £ € €8om d 8n ý n ¥ € 8nim l ¡ hîj

 í£

s

s

p"&&&"  š $V(''!¦7–

ô

l"&&&"  j $'(''!™˜G–

C 8€ C  ¥ • n È H Å s „å € l$"V&'&(&'"6›˜jG–V€8nI ü 8n ý n ¥ |Ž Ď

"&&&"  £ – 8V('(6›¦  f$  8n ý  €n €

‡ ˆ €

l"&&&"  j 8V('(6›…œ–

‡ ˆ €

l"&&&"  j 8V('(6™˜G–

• £¡ ¤¢Å @ n È

  C 8€ C )n €8nI(€8n þ Œ‹ ‚…Š |Ž Ž

€ 8n m

s ‚ 1Q8Q)„t ƒ xyyyxv r Qy8yQ1)wus x yxv t ô ª 111Q)„t ô xyyyxv ‡ ˆ € ¥ ¥  C )n ¥  Ž ü

l8"'&'&'&("6›…œ– j 2 " „ € U(%Yw¡ 8n    C 8€ C )n ÿ 6A &

8"V&(&'&("6df£ 8n ý  € ¥ š

n U¤¢Å s ¢å • $  ü $n ý €n € È£¡ n ¥ ¥

n È f$   €n

2 " „ U'¤£¢¡ n È

• H Å @ ¥

¥

s

|Ž Ž

 C 8€

ƒ

|Ž

C 8€ C  ¥ |Ž Ď

úù )¦¯ û ­˜ ¯ û ´ ‡ — f ú  )ùT¯ “' ­ ˜ ¯ “'øø  ‡ ˆ ÷ ‡ ˆ˜ ö ‡ ˆT€ ƒ ‚ s r ˜õ ô p "&&&" $V'('6

ª

16

The notation for the Iterative Reverse Dutch Auction (IRDA) algorithm is provided in Table V. The IRDA Algorithm: As opposed to a naive approach, the IRDA algorithm relies on choosing an increment in each round such that it is based on the size of the bundle to be procured. 1) Suppose the buyer’s initial willingness to pay for the entire bundle is zero i.e. . Therefore the willingness to pay for each item is also zero i.e. . Since no sellers are likely to be interested to bid at this price, therefore

F þ

 F þ

X F H ~m 

C §

@  F ¦(gm

V § @



å

F om

 ¢ÿ

C§§ @  

@

W  !§ C

W

The payment made by the buyer for the subset in iteration is . The average price paid by the buyer for each item procured is . Iterations in which no items are procured are ignored. The reserve price of the seller for any bundle in iteration is . GVA with reserve prices [64] used in each iteration to solve the allocation and payment problems efficiently. The bundle procured in round is denoted by . Therefore the integer programming formulation of the GVA problem with reserve prices in iteration becomes

C im

C Iþ fm  Vfm å X H C   C  W  '§ ™Gþ V  C W

(25)

2) Increment the average willingness to pay for each item by to . This actually means that the buyer’s willingness to pay for the bundle is changed to . We assume that the increment in every iteration is constant. The reserve price of any bundle for the sellers becomes . 3) Solve the allocation problem if there are any bids i.e. for iteration solve Eq. 26 and calculate . This is again a combinatorial optimization problem. But this is much smaller than the complete problem. 4) Allocate the subsets to the winners. Remove the allocated items from the set to be procured and increment the average willingness to pay for each item to , i.e. the maximum willingness of the buyer to pay for the remaining bundle is . The new reserve price of any bundle of items for the sellers is .

„ ~7þ  V

l"&&&"  j –" I Q @ – " „   j" $V'('!¦˜GQ˜TUG'%Êä%$”@ È  3 2  7Hç C )n I @ TQRG–S ¥ m å • ¤$3@ ä@ n §  j"    È  Ž Ž  l8"(&'&'&("6¦…jœ– "˜ISRœ– C ˆ'§ @ ƒ%$3@ ä@ 6§  Q @ ¨¥   j"  n    È  3 2 7Hç l"&&&"  j – 8V('(6›…œÏ • %$3@  j"

A

A

A

A

„  V ÊäE¨m

6 5 7ô s „

r QQ11)„ƒs xyyyxv t

 V§  V Ê'Ê¢E4@ X ¨ 

IP¡ä£ –흃%j3@  È "



s

å

‰ û

A

5 8 @9û

û

A

5 6 û

V ¦m

 È Ž  G C )n Hç



Ž

‰ €

¥

Ž

» —7û º ´ sx » ûº  E » ‰ ‘© €º ‰ ø÷ ‰E » ‰ û (DB ºC » ‰‰ û º €

& & )ÿ !A

5 û r s

F

ô

A

combinatorial auction problems have been proposed by [74] and [75] which try to resolve the tension between economic and computational efficiencies. In a similar vein, Biswas and Narahari [73] have proposed an iterative reverse Dutch auction scheme for combinatorial procurement auctions which we discuss next. 1) Integer Programming Formulation for the Reverse Dutch Auction: The basic idea behind this approach is to formulate the procurement problem as a weighted set covering problem and use the Dutch auction format to develop an iterative solution scheme. This approach reduces the computational complexity by breaking up one large GVA into several smaller GVA’s. The iterative algorithm itself is motivated by [76], and involves a two step process: (1) progressively increasing the average reserve price of the items in each round, and (2) allocating items for which bids satisfy the reserve price. Classical (forward) Dutch auctions are decreasing auctions which have been typically used for multi-unit homogeneous items. In the single unit Dutch auction the auctioneer begins at a high price and incrementally lowers it until some bidder signals acceptance. Similarly in the multi-unit case the price is incrementally reduced till all the items are sold or the seller’s reserve price is reached. In the reverse auction for procurement the buyer has a procurement budget and tries to procure a bundle of items at minimal cost not exceeding the procurement budget. She starts with a low initial willingness to pay (say equal to zero) and keeps on increasing the willingness to pay until the total bundle is procured or the budget limit is reached. This total budget cannot be divided linearly into budget for each item because of the complementarities involved. The iterative mechanism proposed in [73], [77] consists of ( is the multiple bidding rounds denoted by , maximum willingness initial round). The buyer sets to pay for the remaining bundle to be procured in round . The pricing of items is not linear, therefore the cost of the allocated bundles cannot be divided into price of individual items. Therefore we compute , the average willingness of the buyer to pay for each item in round .

TABLE V N OTATION FOR ITERATIVE REVERSE D UTCH AUCTION ALGORITHM set of items to be procured index for an item to be procured, a subset of items, total number of suppliers index for suppliers, iteration number bundle remaining to be procured in iteration set procured in iteration buying price of the set in iteration average buying price of each item in iteration average price set by the auctioneer in iteration maximum price set by the auctioneer for the bundle in iteration valuation of set to seller indicator variable that takes value 1 iff bundle is allocated to agent increment in buyer’s willingness to pay

(26)

ÿ

¥ ¦m ¥ 1¦m å „ 3ÿ )(''!'¤Êäÿ   "&&&" " „  ¥ @
@

¥ 1@

 ¨m  ʨ¥ m m  å 3§þ  ¥    © "  ¥   ¥ 

ÿ

%j$”@ È ä@ n § "   

¥ þ

ÿ

ÿ

342ç C )n  X Œ ‹ Ó ¥ @  …¨ Ž ‚˜Š

& 0ç ( & Ã $ "Ã )'ç %#!

Ž



 ¥§

 ¥ @ ¨ X 

¨¥ C ˆ'§ @  

ÿ ÿ

17

TABLE VI N OTATION FOR PROCUREMENT UNDER CAPACITY CONSTRAINTS

E. Summary and Current Art In this section we reviewed mechanisms designed to support multi-item procurement scenarios. In the first part, we discussed a purely computational approach to the procurement decision problem to illustrate the computational problems that arise. In the next section we introduced economic concerns and discussed the limitations to achieving both computation and economic efficiencies. Finally we discussed the implications of capacity constraints at suppliers on decisions made through combinatorial auction models. Also, we deliberated on one approach to providing the suppliers with a bid suggestion device to help in reformulating bids in a complex combinatorial environment. Currently, combinatorial auctions constitute a very active research area. There are several surveys that have appeared on this topic, for example, see [30], [31], [32], [16], [33], [42]. There is also a recent comprehensive book by Cramton, Shoham, and Steinberg [34]. There have been numerous efforts in solving the winner determination problems arising in complex combinatorial auctions. The reader is referred to the papers [45], [88], [90], [80], [78], [79], [81], [94], [67]. Another topic of active research is design of truthful combinatorial auctions. Please refer to [83], [87], [60], [31] for more details. Design of iterative combinatorial auctions [84], [85], [42], [86], [93], [77], [91], [92] is one more direction in which a fair amount of research activity is in progress. Multiattribute combinatorial auctions, however, have received little attention due to the intrinsic difficulties involved. Table VII provides an appropriate taxonomy of combinatorial procurement auctions and gives a listing of relevant papers under each category.

D. An Iterative Procurement Mechanism for Capacity Constrained Environments In some strategic sourcing settings, where suppliers are looked upon as extensions of the enterprise, it is imperative to engage in prudent capacity management. Failure to do so could result in production line disruptions and high overtime costs. So when making multi-item procurement decisions, even when the product attributes do not show complementarities in costs, but share resources it is necessary to factor in capacity planning at least at the aggregate levels. One recent effort to incorporate this aspect into auction based procurement mechanisms has been due to Gallien and Wein[69]. The procurement scenario that they analyze is as follows (see Table VI for notation): The buyer wants to procure units of component . There are suppliers willing to supply some or all of these quantities subject to overall capacity constraints. Overall capacity constraints of resource (supplier) is and the usage can be described by a linear model where is the amount of resource required for component . An auctioneer acts as an intermediary between the buyer and the seller thereby decoupling the information between the two, much like what happens on http://www.freemarkets.com/. 1) The Auction Mechanism: The mechanism is designed as a multi round auction where in each round the supplier submits a bid to sell any quantity between 0 and of component . Further non-reneging and minimum bid

 £  ÿ é Xn 

 1ÿ n Q n6§ž)1ÿ V‰ ¥  ¥ ¶ D11ÿ  g‚  … n ƒ    n  ¥ ¥  $¥  H j

£

¥

n 6§

ÿ

This approach to the combinatorial procurement problem, solves the problem of capturing cost complementarities. However, in typical strategic procurement settings, capacity planning at suppliers is a crucial task meriting detailed attention if supply lines are not to be disrupted. This planning of capacity and allocation of contracts can be significantly difficult when multiple contracts are to be established simultaneously. We discuss this issue in the next subsection.

At the end of each round, the potential allocation is displayed to the supplier which can be used to better the bid in the following round. The supplier can choose to follow a myopic best response (MBR) strategy embedded within a software based bid suggestion device. To use this feature, the supplier is required to submit his actual production costs for component to the auctioneer. The next bid to be submitted by supplier is such that he maximizes his potential payoff in round by carrying out a suitable computation [69].

 1ÿ n  Á

j

5) Go to step and repeat until the buyer can procure the entire bundle or the upper limit i.e. the total procurement budget is reached . In any iteration the following condition should be satisfied:

8'&''&(6›T£ – 8"V('&(6›…œ– ݃1ÿ I " & "  l && "  j „   n  ¥  C )n Í "&&& Ï8'''("6›¦£ – Éê   n ¥ ¥ Ž  C  Í58"V(&'("6›…œ– (r • 1ÿ n  n7ý ¥ l & & j n  ÿ !A &  $¥ ¥ Ď

 C  C Vn 1ÿ n  1ÿ nV‰ ¥ Œ‹ v˜Š Ď Ž  $¥  8¥

number of components (items) to be procured index for components, number of units of component to be procured number of suppliers index for resources (suppliers), capacity of supplier amount of resource required for component index that represents auction rounds bid submitted in round by supplier for any quantity between 0 and decision variable for bid

decrement rules are in place. In each round of the auction a potential allocation is obtained by solving the linear program given below.

s u y s u s    y Ä s xwxwxwC vs ¥ s € RsxwxwxwC vtn 11ÿ  ¥ n   ˜1ÿ äÁ 

j

¥

j

É

 @ X ¨ ¥ 

ô

V  ¥  Q¨m ՝ ¥ å b  7Hh pb Œ  Š d %gò  ñ db ò b  ó q i  hf e c ¨ C ˆ¥Ž

r x1Qy8Qx)v„us y y t ô ª Q8Q1)wt ô xyyyxv l ÿ s ÿ ˜ ˜ ® »Aº  `

2 "&&&"  ¡ ©$V'('6§›c£

A

s

s

¥

n ý

nr

£

£  1ÿ n ‰ ¥

j

a

»Aº˜  f »Aº˜  ` ˜  öA

˜ r® Y s

ª ô

(27)

(28)

(29)

¥j

É

18

Scenarios in Multi-Item Procurement Combinatorial Auction

Math.Programming Models [78], [79], [80], [81], [83], [84], [85], [42], [86] [66], [65], [58]

Volume Discount Auction

Game Theoretic Analysis [45], [22], [16], [68], [82] [73], [77], [87], [88], [90], [91], [92], [67] [93], [77], [91], [92] [66], [65]

Case Studies, Surveys, References [17], [15], [37], [18], [19] [20], [21], [89], [31] [32], [33], [34], [22] [17]

TABLE VII C LASSIFICATION OF M ULTI -I TEM P ROCUREMENT AUCTIONS

VI. S INGLE I TEM , M ULTI -ATTRIBUTE P ROCUREMENT M ECHANISMS The procurement mechanisms of the two previous sections involve a single attribute based on price. In practice, these mechanisms address only a limited band of the negotiation spectrum, whereas sourcing decisions involve multiple criteria - both quantitative and qualitative [95]. Benyoucef, Ding, and Xie [55] present an exhaustive, hierarchical list of criteria that are used in evaluating and selecting suppliers. Incorporating these criteria into an automated negotiation tool to support sourcing decisions has been the holy grail of purchasing professionals, industrial engineers, and computer scientists. Many software solution vendors now support multi attribute reverse auctions in their e-sourcing solutions. Weighted, multiparameter, multi-line item request-for-quotes (RFQ’s) and reverse auction capabilities are provided by Ariba, freemarkets, and Procuri; i2 Technologies goes one step further and claims to support all these within an optimization module that allows business level constraints to be incorporated [38]. Since these are commercial products, it is not possible to independently verify the mechanics of the automated sourcing solutions. However, we conjecture that they use one or more of the known approaches to multiple criteria decision analysis (MCDA) such as additive value models, analytic hierarchy process (AHP), lexicographic ordering, multi-attribute utility theory (MAUT), simple multi-attribute rating technique (SMART), and traditional weight assessment. For a detailed treatment of these approaches, we refer the reader to [96]. In this section, our discussion focuses on single item, multi-attribute procurement problems to investigate (1) the problem scenarios addressed, (2) solution approaches, and (3) computational and game theoretic issues. A. Issues in Multi-Attribute Procurement Simply defined, multi-attribute procurement refers to the decision process related to the determination of a contract by considering a variety of attributes involving not just price but also aspects such as quality, delivery time, contract terms, warranties, after sales service, etc. In practice, multi-attribute procurement scenarios come in various hues and shades. To aid in analysis, we group these problems into three distinct categories which we believe adequately reflect the issues to be

¢ ¢ ¢

In the next section, we review models that address procurement scenarios with one more degree of complexity - sourcing decisions based upon multiple attributes.

considered from a computational / automation point of view. The groups and their characteristics are: Multiple attributes are known a priori and are uncorrelated; individual attributes have point values. The suppliers provide point bids where each bid has a single price and each attribute has a single value, either provided by the supplier or computed by the buyer. Multiple attributes are known a priori and are uncorrelated; each attribute can take an individual value from a domain of possible values for the attribute. The suppliers provide configurable bids which specify multiple values and price markups for each attribute. The multiple attributes are not known a priori and they are uncorrelated. The suppliers may provide point bids or configurable bids. This is not unlike many negotiation situations where the buyer has only a rough idea of her requirements but relies on suppliers to educate him about attributes relevant to the procurement decision. A further level of complexity is involved when the attributes are correlated in each of the above cases. Also, each of the above scenarios can occur in combinations. In the first scenario above, if we assume that there exists some decision rule which, as a function of all the multiple attributes, ranks the bids, then the winner determination problem is relatively straightforward. In the latter two cases, however, the options are combinatorial in nature and hence the winner determination problem will have exponential complexity. This raises the issue of compactness of information or bid representation. We now focus on methods to develop the decision rule for ranking of bids. Traditional approaches to developing the decision rule have relied on either a hierarchical elimination process or on weighted average techniques. Since determining the hierarchy or the choice of weights is not always straightforward, especially in the face of a large number of attributes, many sophisticated approaches, including the use of optimization tools, have been proposed which we review next. 1) Elimination Methods: In this method, at each level, we eliminate from the bid list, bids that do not satisfy the selection rule. The selection rule may be a conjunctive rule or a lexicographic rule. In either case, a hierarchy of the attributes needs to be established in the order of their importance, which is fuzzy for most decision makers. To overcome this handicap, optimization based approaches have been devised. We briefly describe these techniques below: Lexicographic Ordering: Here, on the first level, we select the most significant criterion and we compare bids based on

19

in bid

proaches that have been proposed to develop multi attribute procurement mechanisms. B. Additive Value Model Based on Ordinal Ranking of Alternatives A straightforward approach to multi-attribute procurement is to assume that the utilities of each of the attributes are additive in nature and hence a virtual currency, like the stock market index can be developed. Formally, we have a vector of relevant attributes of a bid. We index the attributes by and the set of bids by . See Table VIII for notation. A vector is specified, where is the level of the attribute in bid . In the simple case of an additive utility function , each attribute is evaluated through a utility function and the overall utility is the sum of all weighted utilities. This produces a virtual currency to be used in mechanisms of the type proposed for single attributes. The crucial issue however is the following: How are the weights or the utility function decided? Naive approaches include the elicitation of buyers preferences through the design of smart web forms [1]. More recently techniques based on decision analysis have been proposed by [97] and [98] which we describe next. The WORA technique has been developed as part of an application framework to provide buyer side decision support for e-sourcing [97]. It is based on the realization that the buyer can make deductions about the superiority of bids through simple pairwise comparisons. By making many such comparisons, not entirely exhaustive, it is possible to build a larger information set to elicit intelligently the scoring function. This is done in a two step process. In the first step, sample ordinal rankings of bids are provided by the buyer. They are of the form . These rankings are checked for intransitive preferences and dominance violations. These sample rankings are then transformed into constraints to a linear program, which then generates estimates for the decision maker’s weights. The LP is formulated as follows: Let be a subset of bids that are ranked in the order . The score of each bid is computed as: (30)

n 

¥

 ’F m ‘C m 

i

£

 €

¥ €m  ''&  ¦m •F m ”C m && “   2¥ m"&&& m m $€$'(''" F $" C £  m

ô  ¦€

Set of Bids, index for supplier bids ( number of attributes index for attributes, vector of attributes level of attribute in bid vector of attributes of bid utility function for attribute score of Bid weight for attribute

« 1Q11)„t ô xyyyxv r 8QQ8)Sžs x y y¡ y x v t † xyyy „ ‡€ Q1Q1x …€ t € j œ– j s ô ô  €

n æ  Á 6n ¥ b1¥æ Á b n(‰ £  n "&&& n  H 1'''(" C   cn Á 2  8'('(" C YÎ m m"&&& m

C   n  æ – ¥ fœ@  n  ¥ 8¥ ¥ H Ž

n fm

˜ ‰  û » ˜ ² º ˜ ˆ  ² ˜ f

€ q s

«

ô

“ ¦m

n œ@

this. If a bid satisfies this criterion much better than the other suppliers, then it is chosen, otherwise the bids are compared with respect to the second criterion, and so on. Satisficing: In this technique we set minimum levels for every attribute except one, which is the target attribute, such as price. We select bids which satisfy the minimum levels and choose the one with the optimal value of the target. It is also possible to iterate through the minimum levels. Analytic Hierarchy Process (AHP): This is an analytical tool, supported by simple techniques, that enables people to explicitly rank tangible and intangible factors against one another for the purpose of resolving conflict or for setting priorities. The process involves structuring a problem from a primary objective (e.g. selecting the best bid) to secondary levels of objectives (e.g. performance objectives, quality needs, etc.). Once these hierarchies have been established, a pairwise comparison matrix of each element within each level is constructed. Goal Programming: This is a way to handle multiple objectives in what would otherwise be a LP. The basic concept is to set ”aspiration levels” (targets) for each objective and prioritize them. One would then optimize the highest priority objective with respect to the original (”hard”) constraints. Next, a constraint is added saying that the first objective function’s value must be at least as good as what was achieved or the aspiration level, whichever is worse. Now the second objective is optimized, turned into a constraint, etc. 2) Weighted Average Methods: The weighted average methods essentially rely on the introduction of a virtual currency which expresses the overall utility of a bid to the buyer. The computation of the virtual currency is based upon the utility function specified by the buyer and bids that achieve the best overall utility are declared winners. It is obviously of interest to the buyer to specify the best possible utility function. We briefly describe some of these techniques below: Traditional weighted average technique: Purchasing professionals have traditionally relied upon assigning weights to individual attributes and using a simple additive rule to derive the virtual currency. This is not dissimilar to approaches in micro-economic theory for developing utility functions. This has been further refined by the swing weighting approach described in [97]. The application of this approach can be severely restricting when the number of attributes to be considered is large. Weight Determination Based on Ordinal Ranking of Alternatives (WORA): This approach recognizes the pitfalls in using the weighted average technique and hence relies on linear programming techniques to compute the optimal weights for each of the attributes. We detail this approach in the next section. Inverse Optimization Methods: This technique is an improvement over WORA, in the sense that it does not rely on a single central agency (the buyer) to provide inputs to compute the optimal decision rule. Rather it uses a novel and intelligent technique based on inverse optimization to gather information distributed among various agents (sellers) in the marketplace to develop the optimal scoring (decision) rule. In the following sections, we emphasize two different ap-

TABLE VIII N OTATION FOR O RDINAL RANKING OF A LTERNATIVES

)

20

TABLE IX N OTATION FOR I NVERSE O PTIMIZATION T ECHNIQUES

)

of buyer

. . .

C. Additive Value Model with Inverse Optimization Techniques In some procurement scenarios where the establishment of a contract depends upon multiple attributes (price, quality, delivery time, features and options, etc), a Request-for-Quote (RFQ) process is preferred. In this process, the buyer announces a scoring rule in terms of the bid price and the various attributes to be considered. It is not uncommon for the buyer to change the scoring rule to reflect any new information that has been gleaned during the RFQ/negotiation process. This



q



q

A feasible solution to this LP can be found and results in a set of weights to be used in the scoring rule. The authors also provide some numerical experiments to show the efficacy of the technique. These weights are then used within a standard single attribute like procurement mechanism with the virtual currency, obtained by combining price and all other attributes, substituting for price. While this technique is an improvement over ad-hoc weight assignment models, it still relies on a single buying agent to indicate an ordinal ranking in order to come up with an optimal scoring function. This approach does nothing to exploit the cost complementarities that production systems of the suppliers may exhibit in providing certain attribute levels. In such cases it may be beneficial for the buyer to understand the nature of the suppliers’ cost functions in terms of the nonprice attributes. Understandably, it is not likely to be in the interest of suppliers to part with this information. In the next section we discuss one approach which tries to overcome this problem.

q

£

for each

q

q

"&&&"  V'('!¦ip

Maximize subject to:

¥



þ … †3 o n jV'('" fC d"  § C  H¥ ¶  " & & & o ¥  ¥ $¥ ¥  þ …› n V&('(EC "  § C  ¥ ¶ ¢ " &&" H ¥  á n¤n ''('á UC ¥ n  ¥ " & &¥ & " ¥` ¥ b `

where the weights are unknown and is the number of the attribute, and these satisfy the bid rankings. The LP is:

 ¥ n r ƒn¤n "V&'&(&'" C n "  n r ¥ C  ¥ ¶ H ¥ E` ¥  ¥

¥ E`

¥



 V(''!™im " & & & " q 0"&&&"  1V('(6™”£ q  q þ

 H 1'('(" C 8Dþ "&&& "



l

¢ ¢ ¢ ¢

ô ˜ xyyyxv v ™† 11Q8)St —

ô

ô

˜ v ”†

number of suppliers index for suppliers ( ) number of attributes index for attributes, number of cost parameters price specified in a bid index for rounds of auction ( magnitude of non-price attribute magnitude of parameter for attribute of supplier magnitude of parameter affecting attribute magnitude of parameter affecting attribute in round of the auction unparameterized scoring rule in round

idea is formalized in the eRFQ mechanism in [98] which is designed to address a procurement scenario with the following characteristics and assumptions (see Table IX for notation): A single item with multiple attributes is to be procured. ; k indexes the The attributes are indexed by suppliers; and indexes the rounds of the auction, where is the number of cost (and utility) parameters. Each bid submitted by a supplier is of the form where denotes the price and is the magnitude of non-price attributes which are continuous, nonnegative variables. Supplier k’s cost function , is additive across attributes and is increasing, convex and twice continuously differentiable in . The auctioneer is assumed to know the form of the suppliers’ cost function but not the actual parameter values , is the true utility function of the buyer and the scoring rule is , in round , which is increasing and concave in . Further, to guarantee existence of solutions to the optimization problem, described later, a set of technical requirements are imposed upon the cost, scoring and valuation functions. The eRFQ mechanism is based upon an open ascending rounds. The first auction format consisting of rounds enable learning of the parameters, through Inverse Optimization, and the last round is with an optimized scoring rule. Ahuja and Orlin [99] describe Inverse Optimization in the following manner: ”A typical optimization problem is a forward problem because it identifies the values of observable parameters (optimal decision variables), given the values of the model parameters (cost coefficients, right-hand side vector, and the constraint matrix). An inverse optimization problem consists of inferring the values of the model parameters (cost coefficients, right-hand side vector, and the constraint matrix), given the values of observable parameters (optimal decision variables)”. In each round of the auction, the buyer announces a scoring rule in response to which suppliers submit bids. Activity rules and transition rules are imposed to move from one round to the other. In each round the buyer ranks the bids according to the latest scoring rule and announces the rankings without revealing the bidders or the actual bids. The winner determination, at the end of rounds, is based simply upon an English auction like rule, with the bidder providing the highest utility at the end of rounds being offered the contract at his bid price. The key questions that arise are: (1) what is the likely bidding behavior of the suppliers; (2) how does the buyer estimate the suppliers’ cost functions; and (3) how is the optimal scoring rule determined after learning the cost functions. Bidding Behavior of Suppliers: The supplier is assumed to follow a myopic best response bidding behavior which is in line with the approaches used in [100], and [69]. Here, the supplier chooses his bid such that he maximizes his current profit with the assumption that other suppliers do not change

£

« QQ11)„t ô xyyyxv r QQ8Q)„ƒs xyyyxv t „ s ÷ ÷ ÷

„ Ÿ " !¦ – C  ¥ ¶ H ¥ – ¥ ¥ @ lݝ C ˆ4@ ¨¥

“ ÝoG@ @  F @ F ݝ C @

ô

¥ –

—

»˜ fº˜ k ije ˜ ˜ h e g e f ˜ ˜ d

r s †

« ô

— f ÷

21

their bids. The bid is chosen by solving the following NonLinear Optimization problem:

(31)

subject to

(32)

In the final round however, the constraint would be:

(33)

While the bidding behavior does not include a game theoretic analysis of competing suppliers, it still requires bidders to be sophisticated enough to solve optimization problems. This is striking a middle ground with respect to the bidders’ rationality. Estimating cost functions of the suppliers: By virtue of the auction design, the auctioneer at the end of rounds has for each attribute and each supplier , equations. These equations obtained by solving the first order conditions for , (34)

In the model described above, assumptions about (1) undistorted bidding behavior by suppliers and (2) knowledge of the form of the suppliers’ cost functions may be untenable in practice. The authors propose several extensions to the basic model in order to make it more robust. First, in developing the scoring rule for the last round, the authors envisage three problems: (a) a difficult mathematical programming problem needs to be solved, (b) the scoring rule may turn out to be too complex, and (c) the scoring rule may force the losing supplier to submit bids with negligible values of non-price attributes. To overcome these problems the authors propose to (1) changing the method of finding the best competitor, (2) providing the scoring rules in graphical form, and (3) introducing lower bound constraints on the attribute levels in the final round. Secondly, the bidding behavior of suppliers may not follow the undistortedness assumption either because of strategic intentions of the supplier or the lack of sophistication on their part. This would result in an inconsistent set of simultaneous linear equations. The authors suggest using a weighted average least squares procedure to reduce the effects of bid distortion. We however would like to emphasize that this approach does not absolve buyer of need to compute true utility. D. Summary and Current Art In this section we reviewed mechanisms designed for the procurement of items with multiple attributes. We briefly reviewed the approaches to multi-criteria decision making and identified two broad categories of techniques - elimination methods and weighted average methods. The crucial issue in multi criteria procurement is the assignment of weights to each of the attributes to facilitate the development of a scoring function which captures the buyers’ utility. Two intelligent approaches - the first relying on a central agency to indicate a pairwise preference among a sample of received bids and the second based upon estimating the suppliers cost functions, were detailed. Currently, developing efficient approaches for multiattribute procurement is an active research area. The reader is referred to [105], [104], [102], [72], [58] for some recent work. The papers [105], [104], [102], [101] build upon the approaches described above. Kameshwaran and Narahari [72] have proposed an approach based on goal programming. Goal programming (GP) is one of the tools of choice for multi criteria decision analysis [106]. In [72], the authors show that GP can be used to model procurement scenarios where suppliers provide bids with configurable offers. Here the bids are assumed to be piece-wise linear and the buyer has a hierarchy of goals or aspiration levels which are to be satisfied. The authors propose the use of Weighted GP, Lexicographic GP, and Interactive Sequential GP techniques to solve the multi-attribute procurement problem. There are many issues that remain unresolved in multiattribute procurement. No single approach seems to work uniformly well and it is intrinsically a challenging problem. Much work remains to be done in all the areas: winner determination algorithms, payment rules, and achieving truth revelation.

form a set of simultaneous linear equations. By solving this set of equations for each supplier the true cost parameters for each attribute can be obtained. Computing the optimal scoring rule: After learning the suppliers’ cost functions, the optimal scoring rule for round is computed by solving the following optimization problem:

(35)

subject to:

(36) (37) (38)

The objective function maximizes the buyer’s utility with the last three terms of (24) making up the winning suppliers price; (36) is supplier ’s maximum drop-out score; (37) and (38) ensure that the top two suppliers participate in round .

q

  ¥  ãn¤n "(&'&'(" C n y "  n r &  o n d'('(" y C j"  §  " & & &¥ o   ¥ ¥ ¥ $¥ y ¥ E` ¥  ¥ y 0"&&&"  8V'('!™3£

ãn~n "(&'&(&'" C n " X¥  n r ¥ E`  ¥

} ' n F ¥ ` "'&(&'(Y)F ¥ ` " F X¥ &"C C  ¥ ¦ãn …  ¥ HŽ

}  ' C X¥  1¥ W

n ¥`"&&&" ¥` ¥  ¥  ¤n ''('UC n "  n r
¥ E` ¥ E` q q j j

W  êhw3 o n ¥ jV'('" fC ¥ d" ¥  8¥ § ¥ l  v … "&&& o   HŽ W  l †vx…” ¥  $¥ W  W  F @ Ö C @ W Ö F @ á C   F r ¥ 3 …  ¥ HŽ "'&'&('" C " C X¥ & ¥ á C  C ¥ 3 X  W … ¥  8¥ HŽ A C  ¥ HŽ C  ¥ F ¥X   ¥ W Ž  H C  |  § ¥ 'fc{ z Š q HŽ | £  8¥ £ C   ¥ oœ@  n HŽ   q q

C  ¥ cþ s s xwxwxw$¤s s r … u 'c s t Š q HŽ

22

Scenarios in Multi-attribute Procurement Additive Value Models Inverse Optimization Models

Math.Programming Models [101], [72], [102]

Game Theoretic Analysis [1], [103], [97], [104], [105], [98]

Case Studies, Surveys, References [55], [95], [2] [99]

TABLE X C LASSIFICATION OF M ULTI - ATTRIBUTE P ROCUREMENT AUCTIONS

We summarize the discussion on multi-attribute procurement auctions by presenting Table X which gives snapshot of the possible scenarios and a listing of relevant references. VII. A C ASE S TUDY
FROM

G ENERAL M OTORS

Procurement business professionals need to collect large numbers of bids from prospective suppliers, and, then, determine an allocation of awards to bids that minimizes procurement costs subject to various constraints on supplier delivery risk, vendor award count on commodities or subsets of commodities, and logistics costs. This need is present during the initial stage of awarding business to suppliers on new products, and is also present when primary suppliers are unable to deliver supplies (e.g., in the case of a strike, natural disaster, financial default, or other event that causes a work stoppage) to existing products. A team of researchers from General Motors Research (which included the last four authors of this paper) recently used an auction-optimization approach similar to those described earlier in this paper to solve this problem. This approach, soon to be deployed as a web application within General Motors Corporation (GM Corp), allows business users to determine an optimal allocation of awards to bids using the application over the company’s intranet. Because the optimization approach is math driven, automatic, and can be rolled out over the businesses intranet, business users can now focus on the business constraints that drive the award process. Business users can dynamically adjust the business constraints to understand the impact of the stated constraints on the award allocation. That is, they can understand the sensitivity of the awards to the constraint settings. This process can happen in real time, thus making the procurement business professionals’ much easier, and allowing them to focus on developing the relationship with suppliers rather than spending their time crunching numbers in spreadsheets in an effort to discover the optimal award allocation using more primitive manual type approaches. The sourcing corresponds to that of an important raw material for automotive manufacturing at GM. The overall commodity sourcing process is shown in Figure 2. Within GM, a huge amount of this commodity is sourced every year. To gain maximum cost savings (at a sufficiently high level of desired quality), GM uses a centralized demand aggregation and reselling application for the whole supply chain. This application attempts to combine the individual commodity requirements of its processors with GM’s direct commodity requirements to create large orders. These larger orders often qualify for significant volume discounts with the commodity

Fig. 2.

Overall Business Process

suppliers. GM then resells a portion of the purchased commodity to its processors to cover their material needs. Some of the processing is performed by external processors, and the rest is done by GM’s internal processing group. So, the overall process is very complex and manual approaches for determining an allocation of awards to the suppliers require enormous effort. Moreover, the solution may not be optimal. By using an optimization model similar to those described earlier in this paper, superior solutions can be obtained with less effort. A. System Overview A general commodity sourcing application in the automotive industry will look like the one shown in Figure 3. Here, the overall process is the same as shown in Figure 2. The requirements for the commodities are aggregated within a centralized system (sourcing tool) and an effective bundling of the commodity is found. The tool looks at the catalogs of the approved suppliers and sends RFQs to each of the suppliers for those bundles that they are capable of supplying. Each of these suppliers submits a list of configurable bids to the tool in response to the RFQ. The tool evaluates the bids and feeds them into the optimization model. The optimization model is also provided with plant specific constraints and other business constraints. The tool outputs a cost effective allocation to the suppliers.

23

Fig. 3.

Commodity Sourcing Application

to one or more particular plants. Each bid submitted has variable and fixed cost components. They are submitted for multiple units of multiple heterogeneous commodities destined for multiple plants. There are some bounds on the number of units of a particular commodity that a supplier can supply to a plant. Also, there are some restrictions on the number of suppliers that can be selected for a plant and also on the number of suppliers that can be awarded business for a commodity. All of these constraints are determined by taking into consideration the suppliers’ capacities, their disruption risk, their overhead costs, etc. The model that the team created to solve the problem determines an efficient allocation of awards to bids so that overall procurement cost is minimized, and the various constraints on the number of suppliers for a plant/commodity, bounds on the award to each supplier, plant preferred suppliers, and other constraints are satisfied. The mathematical programming model used combines several features of the formulations discussed in IV.B and V.B and extends those formulations with business constraints unique to GM. C. Results and Conclusions Using the above approach, the team was able to determine a cost minimizing potential re-allocation of awards that could be used for negotiations with the current suppliers (A and B) regarding the status of expiring contracts. The set of specialty commodities that the team considered were C1, C2, C3, and C4. The solution obtained by the conventional method (using current suppliers A and B) gave an Annual Purchase Value (APV) of $ 31.11 M while that obtained using our optimization approach gave an APV of $ 30.4M. The results are shown in Table XI and Table XII respectively. These preliminary results gave significant savings of about $ 0.71M (2.3% of APV) in a solution time in the single digit seconds range. Also, the team considered how the optimization model might be used to help determine an optimal supply reallocation strategy in the event of a supplier disruption event. For this, the team considered the reallocation of a portion of the commodity from one of the suppliers to two other suppliers. Three scenarios were investigated: 1) Reallocating only the highest volume parts 2) Reallocating so as to minimize the total cost impact 3) Same as Scenario 1 above, but excluding certain parts for reallocation The results are shown in Table XII. So, if the number of reallocations (Scenario 1) is sought to be minimized, the cost impact is high ($ 3.2M). On the other hand, trying to minimize the total cost impact increases the number of reallocations to 280. Assuming a switching cost of approximately $10,000 for reallocating a part to a new supplier, the cost minimizing reallocation has an impact of $ 2.7M with 146 reallocations. In conclusion, the team demonstrated that significant cost savings are possible when applying auctions and optimization to a real-world procurement problem. The savings could be as high as 3%. Of course, given that the analysis was limited to a particular commodities group, there is reason to believe that this estimate is only a lower bound on what is possible using

Fig. 4.

Main Elements of a Configurable Bid

The main elements of a configurable bid submitted by a supplier are shown in the tree-like structure shown in Figure 4. A configurable bid gives either a base price for a bundle and quantity, or a volume discount price, which is a function of quantity. The bid consists of various attributes, which can be fixed attributes, range attributes, or optional attributes. Fixed attributes are simple attributes with one corresponding value. Range attributes allow suppliers to specify an attribute as a step function, where each step has a fixed impact on price. Optional attributes can either be included or not. A supplier can also specify some logical rules for assigning some discounts to specific combination of selected attributes. B. A Model for Minimizing Procurement Spend The commodity requirements of all the plants are aggregated and bids are invited from the suppliers. Each plant has a set of preferred suppliers for procuring the commodities. Also, suppliers are capable of supplying certain commodities

24

Mill/Commodity A B Grand Total

C1 $107,010 $303,039 $ 410, 049

C2 $ 2,863,928 $ 1, 739, 721 $ 4, 603, 649

C3 $ 7,069,270 $ 10, 153, 908 $ 17, 223, 178

C4 $1,502,302 $ 7, 367, 029 $ 8, 869, 331

Grand Total $ 11, 542, 510 $ 19, 563, 697 $ 31, 106, 207

TABLE XI C URRENT C OMMODITY A LLOCATION (APV: $ 31.22M)

Mill/Commodity A B C D E Grand Total Savings ($) Savings (%)

C1 $ $ $ $

C2 $ $ $ $

C3 $ $ $ $

C4 $5, 117, 764 $1, 276, 440 $ 2, 182, 372 $ 8, 576, 576 $ 292, 755 3.301%

Grand Total $ $ $ $ $ 5, 117, 764 10, 578, 348 8, 739, 365 4, 243, 101 1, 718, 001

71, 708 5, 981 277, 243 27, 945

1, 629, 282 1, 763, 698 1, 007, 565 176, 621

7,600, 918 6, 970, 686 775, 921 1, 513, 435

$ 382, 877 $ 27, 172 6.627 %

$ 4, 577, 166 $ 27, 483 0.597%

$ 16, 860, 960 $ 362, 218 2.103 %

$ 30, 396, 579 $ 709, 628 2.2813 %

TABLE XII O PTIMAL C OMMODITY A LLOCATION (APV: $ 30.40M)

Scenario 1 2 3

Cost Impact (Increase) $ 3.2M $ 2.5 M $ 3.8M

No. of Reallocations 77 280 81

TABLE XIII R EALLOCATION OF M ATERIAL

this approach. It was also shown that using this type of modeling approach allows business analysts to explore multiple different constraints on the award and reallocation process. This leads to a much better understanding of the sensitivity of the optimal sourcing decisions to business constraints than possible using a more traditional manual process. In the following section, we summarize the discussion in this paper by providing a few concluding remarks and pointers to potential research opportunities. VIII. C ONCLUSIONS AND F UTURE W ORK In this paper, we have surveyed the state-of-the-art in auction-based mechanisms for automating negotiations in electronic procurement. We have discussed these mechanisms under three categories: Procurement of single unit or multiple units of a single item based on a single attribute Procurement of single unit or multiple units of multiple items based on a single attribute Procurement of multiple units of a single item based on multiple attributes

A fourth category would be: procurement of multiple units of multiple items based on multiple attributes. Given that the first three categories are areas of active research with unresolved issues, this fourth category would be even more challenging. In each of the three categories of procurement, we discussed several procurement scenarios, provided problem formulations, and discussed issues of computational complexity related to the mechanism and the bidding process. We have also provided a description of several case studies, including a detailed one from General Motors. Currently, procurement auctions is an active area of research with many interesting problems which merit further study. Here, we provide pointers to a few selected categories: Winner Determination Problem in Procurement Auctions Though extensive work has been done on solving the winner determination problem in different types of procurement auctions, there is still wide scope for future work here. One of the promising directions is to come up with efficient approximation algorithms with provable bounds for solving the winner

¢ ¢ ¢

25

determination problem. There are many efforts in this direction already, see for example, [58], [68]. The rich body of literature available in combinatorial optimization and approximation and randomized algorithms will be extremely useful here. Another key opportunity here is in the area of on-line algorithms. In many situations, it becomes imperative to quickly evaluate the bids received and allocate quantities/orders to suppliers. In such situations, on-line algorithms would be useful. As of the present art, there is not much work in this direction and this would be a valuable topic for further investigation. Iterative Procurement Auctions As already stated in the paper, iterative mechanisms have several advantages over one shot mechanisms. Also, there have been several iterative mechanisms developed for procurement [42], [77], [45], [93], [107]. Iterative mechanisms are appealing to the buyer and suppliers because they enable the bidders to apply corrections to their bids in a continuous way. More work is required in ensuring that iterative mechanisms satisfy desirable economic properties also. The use of on-line algorithms for rapid winner determination in successive rounds of bidding is an important direction for future work. Design of Incentive Compatible Mechanisms Inducing truth revelation will continue to be a major issue in the design of procurement mechanisms. Designing such mechanisms has intrinsic difficulties such as very high computational complexity and loss of efficiency. It would be interesting to study how the use of approximate algorithms for winner determination and payment computation would affect incentive compatibility. There is already a fair amount of work in this direction [68], [86], [60]. Multi-Attribute Procurement Multi-attribute procurement is an intrinsically difficult problem, but at the same time an important problem that needs immediate attention. There are a few results available [95], [104], [107] and there are a few promising approaches such as goal programming [72], but much more needs to be researched in this area. Use of Learning in Procurement Auctions Procurement auctions provide an ideal platform for use of machine learning techniques in improving the efficiency of the process. The history of bidding by a supplier is an important parameter for winner determination. To incorporate history into procurement decision making will call for use of appropriate machine learning techniques such as reinforcement learning. Learning based models would be useful in iterative procurement auctions to help the buyer estimate the cost functions of the suppliers and in optimally incrementing the procurement budgets. One such application is discussed in [82]. Another interesting application is discussed in [108]. We believe machine learning techniques have powerful applications in procurement auctions.

Procurement Auctions from a Total Supply Chain Perspective It is important to design procurement mechanisms based on a total cost approach where the total cost captures all aspects of the entire supply chain. This has been explored in a few papers already [59] but a deeper understanding and a more systematic approach is required here. Deployment Issues Practical deployment of procurement auctions will throw up numerous challenges. Several authors have addressed these issues: security of transactions [109]; collusion of suppliers [61]; user interface issues [105]; fairness issues [32]; failure freeness and robustness against failures [32], [109], [41]. These issues need immediate attention for successful adoption of auctions by purchasing departments. Designing software implementation frameworks so as to allow sensitivity analysis of procurement decisions in complex supply chain environments is also an important issue. Use of multi-agent agent technology in automating standard electronic procurement problems is one more issue. Using emerging Internet technology standards such as ebXML in implementing eprocurement solutions is an immediate practical issue. Procurement Exchanges Procurement exchanges are those where there are multiple buyers and multiple suppliers and the exchange facilitates matching of buyers with suppliers. All the issues become more complex with exchanges because of the presence of multiple buyers. There is a large body of literature on exchanges; for example, see [110], [23], [111]. ACKNOWLEDGMENT The first two authors would like acknowledge the support of General Motors Research and Development Center, Warren, Michigan, USA for carrying out this research under the IMPROVE research grant. R EFERENCES
[1] M. Bichler, M. Kaukal, and A. Segev, “Multi-attribute auctions for electronic procurement,” in Proceedings of the First IBM IAC Workshop on Internet Based Negotiation Technologies, Yorktown Heights, NY, USA, 1999. [2] M. Bichler, G. Kersten, and S. Strecker, “Towards a structured design of electronic negotiations,” Group Decisions and Negotiations, vol. 12, no. 4, pp. 311–335, 2003. [3] G. E. Kersten, S. J. Noronha, and T. Jeffrey, “Are all e-commerce negotiations auctions?” in Fourth International Conference on the Design of Cooperative Systems, Sophia-Antipolis, France, May 2000. [4] C. Beam and A. Segev, “Automated negotiations: A survey of the state of the art,” Wirtschaftsinformatik, vol. 39, no. 3, pp. 263–268, 1997. [5] A. Chavez and P. Maes, “Kasbah: An agent marketplace for buying and selling goods,” in First International Conference on the Practical Application of Intelligent Agents and Multi-Agent Technology (PAAM’96), 1996, pp. 75–90. [6] D. Zeng and K. Sycara, “Bayesian learning in negotiation,” in Working Notes for the AAAI Symposium on Adaptation, Co-evolution and Learning in Multiagent Systems, S. Sen, Ed., Stanford University, CA, USA, 1996, pp. 99–104. [7] A. Mas-Collel, M. D. Whinston, and J. R. Green, Microeconomic Theory. Oxford University Press, 1995. [8] R. B. Myerson, Game Theory: Analysis of Conflict. Cambridge, Massachusetts: Harvard University Press, 1997.

26

[9] G. E. Corporation, “Letter to share owners,” GE Annual Report, 2000. [10] S. Beall, C. Carter, and et. al., “The role of reverse auctions in strategic sourcing: Case Study Glaxo Smith Kline,” Center for Advanced Purchasing Studies, Temple, AZ, USA, Tech. Rep., 2003. [11] ——, “The role of reverse auctions in strategic sourcing: Case Study METRO,” Center for Advanced Purchasing Studies, Temple, AZ, USA, Tech. Rep., 2003. [12] ——, “The role of reverse auctions in strategic sourcing: Case Study Volkswagen,” Center for Advanced Purchasing Studies, Temple, AZ, USA, Tech. Rep., 2003. [13] ——, “The role of reverse auctions in strategic sourcing: Case Study Bechtel,” Center for Advanced Purchasing Studies, Temple, AZ, USA, Tech. Rep., 2003. [14] ——, “The role of reverse auctions in strategic sourcing,” Center for Advanced Purchasing Studies, Temple, AZ, USA, Tech. Rep., 2003. [15] W. Elmaghraby and P. Keskinocak, “Technology for transportation bidding at the home depot,” in Practice of Supply Chain Management: Where Theory and Practice Converge. Kluwer Academic Publishers, 2003. [16] ——, “Combinatorial auctions in procurement,” in Practice of Supply Chain Management: Where Theory and Practice Converge. Kluwer Academic Publishers, 2003. [17] G. Hohner, J. Rich, E. Ng, G. Reid, A. J. Davenport, J. R. Kalagnanam, S. H. Lee, and C. An, “Combinatorial and quantity discount procurement auctions provide benefits to Mars, Incorporated and to its suppliers,” INTERFACES, vol. 33, no. 1, pp. 23–35, 2003. [18] J. Ledyard, M. Olson, D. Porter, J. Swanson, and D. Torma, “The first use of a combined value auction for transportation services,” Interfaces, vol. 32, no. 5, pp. 4–12, 2002. [19] C. Caplice and Y. Sheffi, “Combinatorial auctions for truckload transportation,” in Combinatorial Auctions, P. Cramton, Y. Shoham, and R. Steinberg, Eds. The MIT Press, Cambridge, Massachusetts, USA, 2005. [20] Y. Sheffi, “Combinatorial auctions in the procurement of transportation services,” Interfaces, vol. 34, 2004. [21] R. Epstein, L. Henriquez, J. Catalan, G. Weintraub, and C. Martinez, “A combinatorial auction improves school meals in Chile,” Interfaces, vol. 32, no. 6, pp. 1–14, 2002. [22] M. Bichler, A. Davenport, G. Hohner, and J. Kalagnanam, “Industrial procurement auctions,” in Combinatorial Auctions, P. Cramton, Y. Shoham, and R. Steinberg, Eds. The MIT Press, Cambridge, Massachusetts, USA, 2005. [23] P. Milgrom, Putting Auction Theory to Work. Cambridge University Press, 2004. [24] V. Krishna, Auction Theory. Academic Press, 2002. [25] R. McAfee and J. McMillan, “Auctions and bidding,” Journal of Economic Literature, vol. 25, pp. 699–738, 1987. [26] P. Milgrom, “Auctions and bidding: A primer,” Journal of Economic Perspectives, vol. 3, no. 3, pp. 3–22, 1989. [27] P. Klemperer, Auctions: Theory and Practice. The Toulouse Lectures in Economics. Princeton University Press, 2004. [28] E. Wolfstetter, “Auctions: An introduction,” Economic Surveys, vol. 10, pp. 367–421, 1996. [29] J. Kalagnanam and D. C. Parkes, “Auctions, bidding, and exchange design,” in Handbook of Quantitative Supply Chain Analysis: Modeling in the E-Business Era, ser. Int. Series in Operations Research and Management Science, D. Simchi-Levi, D. S. Wu, and M. Shen, Eds. Kluwer Academic Publishers, 2005, ch. 10. [30] S. de Vries and R. V. Vohra, “Combinatorial auctions: A survey,” INFORMS Journal of Computing, vol. 15, no. 1, pp. 284–309, 2003. [31] S. de Vries and R. H. Vohra, “Design of combinatorial auctions,” in Handbook of Quantitative Supply Chain Analysis: Modeling in the E-Business Era. International Series in Operations Research and Management Science, Kluwer Academic Publishers, Norwell, MA, USA, 2005, pp. 247–292. [32] A. Pekec and M. Rothkopf, “Combinatorial auction design,” Management Science, vol. 49, pp. 1485–1503, 2003. [33] Y. Narahari and P. Dayama, “Combinatorial auctions for electronic business,” Sadhana (Special Issue on Electronic Commerce and Electronic Business), vol. 30, pp. 179–211, April 2005. [34] P. Cramton, Y. Shoham, and R. Steinberg, “Introduction to combinatorial auctions,” in Combinatorial Auctions, P. Cramton, Y. Shoham, and R. Steinberg, Eds. The MIT Press, Cambridge, Massachusetts, USA, 2005. [35] W. J. Elmaghraby and P. Keskinocak, “Dynamic pricing: Research overview, current practices and future directions,” Management Science, vol. 49, no. 10, pp. 1287– 1309, October 2003.

[36] M. Bichler, R. Lawrence, J. Kalagnanam, H. Lee, K. Katircioglu, G. Lin, A. King, and Y. Lu., “Applications of flexible pricing in business-to-business electronic commerce,” IBM Systems Journal, vol. 41, no. 2, pp. 287–302, 2002. [37] W. Elmaghraby, “Pricing and auctions in emarketplaces,” in Handbook of Quantitative Supply Chain Analysis: Modeling in the E-Business Era. International Series in Operations Research and Management Science, Kluwer Academic Publishers, Norwell, MA, USA, 2005, pp. 213–246. [38] T. A. Minahan, H. Francis, and V. Mark, “Making e-sourcing strategic: From tactical technology to core business strategy,” Aberdeen Group Inc, Boston, Massachusetts, Tech. Rep., 2002. [39] W. Vickrey, “Counter speculation, auctions, and competitive sealed tender,” Journal of Finance, vol. 16, pp. 8–37, 1961. [40] M. Herschlag and R. Zwick, “Internet auctions-a popular and professional literature review,” Electronic Commerce, vol. 1, no. 2, pp. 161– 186, 2000. [41] K. Manoj and S. I. Feldman, “Internet auctions,” in Proceedings of Third Usenix Workshop on Electronic Commerce, 1999. [42] D. Parkes, “Iterative combinatorial auctions: Achieving economic and computational efficiency,” Ph.D. dissertation, Department of Computer and Information Science, University of Pennsylvania, May 2001. [43] H. R. Varian, “Economic mechanism design for computerized agents,” in Proceedings of the USENIX Workshop on Electronic Commerce, 1995, minor update, 2000. [44] J. Laffont and J. Tirole, A Theory of Incentives in Procurement and Regulation. The MIT Press, Cambridge, MA, USA, 1993. [45] L. Ausubel and P. Milgrom, “Ascending auctions with package bidding,” Frontiers of Theoretical Economics, vol. 1, no. 1, 2002. [46] E. Clarke, “Multi-part pricing of public goods,” Public Choice, vol. 11, pp. 17–23, 1971. [47] T. Groves, “Incentives in teams,” Econometrica, vol. 41, pp. 617–631, 1973. [48] L. Hurwicz, “On informationally decentralized systems,” in Decision and Organization: A Volume in Honor of Jacob Marchak. NorthHolland, 1972. [49] K. Arrow, “The property rights doctrine and demand revelation under incomplete information revelation,” in Economics and Human Welfare. Academic Press, New York, 1979. [50] R. B. Myerson and M. A. Satterthwaite, “Efficient mechanisms for bilateral trading,” Journal of Economic Theory, vol. 28, pp. 265–283, 1983. [51] C. d’Aspremont and L. Gerard-Verat, “Incentives and incomplete information,” Journal of Public Economics, vol. 11, pp. 25–45, 1979. [52] R. B. Myerson, “Optimal auction design,” Mathematics of Operations Research, vol. 6, pp. 58–73, 1981. [53] J. Green and J. Laffont, Incentives in Public Decision Making. NorthHolland, Amsterdam, 1979. [54] A. Davenport and J. Kalagnanam, “Price negotiations for direct procurement,” IBM Research, Yorktown Heights, NJ, USA, Research Report RC 22078, 2001. [55] L. Benyoucef, H. Ding, and X. Xie, “Supplier selection problem: Selection criteria and methods,” INRIA, Lorraine, Tech. Rep., 2003. [56] M. Eso, S. Ghosh, J. Kalagnanam, and L. Ladanyi, “Bid evaluation in procurement auctions with piece-wise linear supply curves,” IBM Research, Yorktown Heights, NJ, USA, Research Report RC 22219, 2001. [57] D. Pisinger and P. Toth, “Knapsack problems,” in Handbook of Combinatorial Optimization, D. D.-Z and P. M. Pardolas, Eds. Kluwer Academic Publishers, 1998, pp. 299–428. [58] S. Kameshwaran, “Algorithms for piecewise linear knapsack problems with applications in electronic commerce,” Ph.D. dissertation, Department of Computer Science and Automation, Indian Institute of Science, Bangalore, India, August 2004. [59] R. R. Chen, G. Janakiraman, R. Robin, and R. Q. Zhang, “Efficient auctions for supply chain procurement,” Johnson Graduate School of Management, Cornell University, Ithaca, NY, Tech. Rep., 2002. [60] N. Nisan and A. Ronen, “Computationally feasible VCG mechanisms,” in Proceedings of the Second ACM Conference on Electronic Commerce, 2000, pp. 242–252. [61] P. Bajaria and G. Summers, “Detecting collusion in procurement auctions: A selective survey of recent research,” Stanford University, Department of Economics, Working Paper 01014, Tech. Rep., 2002. [62] R. Narasimhan, S. Talluri, and et.al., “Evaluating e-procurement solutions,” Center for Advanced Purchasing Studies, Temple, AZ, USA, Tech. Rep., 2003.

27

[63] J. Kagel, “Auctions: A survey of experimental research,” in The Handbook of Experimental Economics. Princeton University Press, Princeton, 1995, pp. 501–587. [64] L. Ausubel and P. Cramton, “Vickrey auctions with reserve pricing,” Working Paper, University of Maryland, Tech. Rep., 1999. [65] M. Eso, S. Ghosh, J. R. Kalagnanam, and L. Ladanyi, “Bid evaluation in procurement auctions with piece wise linear supply curves,” IBM, Tech. Rep. RC22219 (W0110-087), 2001. [66] A. Davenport and J. Kalagnanam, “Price negotiations for procurement of direct inputs,” in Mathematics of the Internet: E-Auction and Markets, B. Dietrich and R. Vohra, Eds. Springer Verlag, 2001, IMA Volumes in Mathematics and its Applications. [67] V. Dang and N. Jennings, “Optimal clearing algorithms for multiunit single-item and multi-unit combinatorial auctions with demandsupply function bidding,” Dept of Electronics and Computer Science, University of Southampton, UK, Tech. Rep., 2003. [68] A. Kothari, D. Parkes, and S. Suri, “Approximately strategy proof and tractable multi-unit auctions,” in Proceedings of ACM Conference on Electronic Commerce (EC-03), San Diego CA, June 9-12, 2003. [Online]. Available: citeseer.ist.psu.edu/kothari03approximatelystrategyproof.html [69] J. Gallien and L. Wein, “A smart market for industrial procurement with capacity constraints,” Management Science, vol. 51, no. 1, pp. 76–91, January 2005. [70] N. Nisan, “Bidding and allocation in combinatorial auctions,” in Proceedings of the Second ACM Conference on Electronic Commerce, 2000. [71] C. Boutlier and H. Hoos, “Bidding languages for combinatorial auctions,” in Proceedings of the 17th International Joint Conference on Artificial Intelligence, 2001. [72] S. Kameshwaran and Y. Narahari, “E-procurement using goal programming,” in Proceedings of International Conference on Electronic Commerce and Web Technologies, DEXA 2003 Conferences, Linz, Austria, 2003. [73] S. Biswas and Y. Narahari, “Iterative reverse dutch auction for electronic procurement,” in Proceedings of the International Conference on Electronic Commerce Research, ICECR-5, Montreal, Canada, 2002. [74] D. Parkes, “iBundle: An efficient ascending price bundle auction,” in Proceedings of ACM Conference on Electronic Commerce (EC-99), 2000, pp. 148–157. [75] M. Wellman, W. Walsh, P. Wurman, and J. MacKieMason, “Auction protocols for decentralized scheduling,” University of Michigan, Tech. Rep., 1998. [Online]. Available: citeseer.nj.nec.com/article/wellman98auction.html [76] V. Chvatal, “A greedy heuristic for the set cover problem,” Mathematics of Operations Research, vol. 4, pp. 233–235, 1979. [77] S. Biswas and Y. Narahari, “Iterative Dutch combinatorial auctions,” Annals of Mathematics and Artificial Intelligence, 2005, to appear in the special issue on the Foundations of Electronic Commerce. [78] T. Sandholm, “An algorithm for optimal winner determination in combinatorial auctions,” Artificial Intelligence, vol. 135, no. 1, pp. 1– 54, 2002. [79] T. Sandholm and S. Suri, “Bob: Improved winner determination in combinatorial auctions and generalizations,” Artificial Intelligence, vol. 145, pp. 33–58, 2003. [80] R. Gonen and D. Lehmann, “Optimal solutions for multi-unit combinatorial auctions: Branch and bound heuristics,” in Proceedings of ACM Conference on Electronic Commerce (EC-00), 2000, pp. 13–20. [81] E. Zurel and N. Nisan, “An efficient approximate allocation algorithm for combinatorial auctions,” in Proceedings of ACM Conference on Electronic Commerce (EC-01), 2001, pp. 125–136. [82] C. V. L. Raju and Y. Narahari, “Use of reinforcement learning in iterative bundle auctions for procurement,” in Proceedings of the International Conference on Automation, Energy, and Information Technology, EAIT-2001, Indian Institute of Technology, Kharagpur, 2001. [83] C. DeMartini, A. Kwasnica, J. Ledyard, and D. Porter, “A new and improved design for multi-object iterative auctions,” Social Science Working Paper No. 1054, Pennsylvania State University, Tech. Rep., 1998. [84] D. Parkes, “Iterative combinatorial auctions,” in Combinatorial Auctions, P. Cramton, Y. Shoham, and R. Steinberg, Eds. The MIT Press, Cambridge, Massachusetts, USA, 2005. [85] S. Bikhchandani and J. Ostroy, “From the assignment model to combinatorial auctions,” in Combinatorial Auctions, P. Cramton, Y. Shoham, and R. Steinberg, Eds. The MIT Press, Cambridge, Massachusetts, USA, 2005.

[86] D. C. Parkes, “An iterative generalized vickrey auction: Strategyproofness without complete revelation,” in Proceedings of AAAI Spring Symposium on Game Theoretic and Decision Theoretic Agents, 2001. [87] Y. Bartal, R. Gonen, and N. Nisan, “Incentive compatible multiunit combinatorial auctions,” in Proceedings of Dagstuhl Seminar on Electronic Market Design, 2002. [88] M. Rothkopf, A. Pekec, and R. Harstad, “Computationally manageable combinatorial auctions,” Management Science, vol. 44, pp. 1131–1147, 1998. [89] S. de Vries and R. V. Vohra, “Combinatorial auctions: A survey,” INFORMS Journal on Computing, 2002. [90] K. Leyton-Brown, Y. Shoham, and M. Tennenboltz, “An algorithm for multi-unit combinatorial auctions,” in Proceedings of National Conference on Artificial Intelligence (AAAI-00), 2000. [91] D. Mishra and D. C. Parkes, “Ascending price Vickrey auctions using primal-dual algorithms,” Harvard Univeristy, Tech. Rep., 2004. [92] ——, “A Vickrey-Dutch clinching auction,” Harvard University, Tech. Rep., 2004. [93] P. R. Wurman and M. P. Wellman, “Akba: A progressive, anonymousprice combinatorial auction,” in Proceedings of ACM Conference on Electronic Commerce (EC-00), 2000, pp. 21–29. [94] M. Tennenholtz, “Some tractable combinatorial auctions,” in Proceedings of National Conference on Artificial Intelligence (AAAI-00), 2000. [95] Y. K. Che, “Design competition through multidimensional auctions,” RAND Journal of Economics, vol. 24, pp. 668–679, 1993. [96] R. Keeney and H. Raiffa, Decisions with multiple objectives: Preferences and value tradeoffs. Wiley, 1976. [97] M. Bichler, J. Lee, H. S. Lee, and J. Y. Chung, “Absolute: An intelligent decision making framework for e-sourcing,” in Proceedings of the Third Workshop on Electronic Commerce and Web Based Information Systems, San Jose, CA, 2001. [98] D. R. Beil and L. M. Wein, “An inverse optimization based auction mechanism to support a multi-attribute rfq process,” Management Science, vol. 49, no. 11, pp. 1529–1545, November 2003. [99] R. K. Ahuja and J. B. Orlin, “Inverse optimization,” Operations Research, vol. 49, no. 5, pp. 771–783, 2001. [100] D. C. Parkes and L. H. Ungar, “Iterative combinatorial auctions: Theory and practice,” in Proceedings of the 17th National Conference on Artificial Intelligence, 2000, pp. 74–81. [101] M. Bichler and J. Kalagnanam, “Configurable offers and winner determination in multi-attribute offers,” European Journal of Operations Research (to appear), 2005. [102] T. W. Sandholm and S. Suri, “Side constraints and non price attributes in markets,” in In IJCAI-2001 Workshop on Distributed Constraint Reasoning, Seattle, WA, 2001. [Online]. Available: citeseer.ist.psu.edu/sandholm01side.html [103] M. Bichler and H. Werthner, “Simulation of multidimensional procurement auctions,” in Proceedings of the European Simulation MultiConference, Gent, Belgium, 2000. [104] M. Bichler and J. Kalagnanam, “Configurable offers and winner determination in multi-attribute auctions,” European Journal of Operational Research, vol. 160, no. 2, 2005, iBM Research Report RC22478. [105] M. Bichler, J. Kalagnanam, and H. S. Lee, “Reco: Representation and evaluation of configurable offers,” IBM Research Report, RC 22288, Tech. Rep., 2002. [106] R. E. Steur, Multiple Criteria Optimization: Theory, Computation and Application. Wiley, 1986. [107] D. C. Parkes and J. Kalagnanam, “Iterative Multiattribute Vickrey Auctions,” Management Science, 2004, special Issue on Electronic Markets. [108] S. Lahaie and D. C. Parkes, “Applying learning algorithms to preference elicitation in the generalized vickrey auction,” Harvard University, Tech. Rep., 2004. [109] Y. Sakurai, M. Yokoo, and S. Matsubara, “A limitation of generalized Vickrey auction in electronic commerce: Robustness against false-name bids,” in Proceedings of National Conference on Artificial Intelligence (AAAI-99, vol. 16, 1999, pp. 86–92. [110] S. Biswas, “Iterative algorithms for combinatorial auctions and exchanges,” Ph.D. dissertation, Department of Computer Science and Automation, Indian Institute of Science, Bangalore, India, April 2004. [111] D. C. Parkes, R. Cavallo, N. Elprin, A. Juda, S. Lahaie, B. Lubin, L. Michael, J. Shneidman, and H. Sultan, “ICE: An iterative combinatorial exchange,” in Proc. 6th ACM Conf. on Electronic Commerce (EC’05), 2005.