VIEWS: 7 PAGES: 29 POSTED ON: 11/11/2011
2002 Manufacturing and Service Operations Management Conference Samuel Curtis Johnson Graduate School of Management, Cornell University, Ithaca, New York Adelman, Dan Hall, Joseph M. Savaskan, Canan Afeche, Philipp Hann, Il-Horn Savin, Sergei Aksin, Zeynep Ho, Teck H. (2) Scheller-Wolf, Alan Anderson, Eddie Holler, Samuel Schmidt, Charles P. Anderson, Jr., Edward G. (2) Hsu, Vernon Ning Schmidt, Glen M. Anily, Shoshana Hu, Xinxin Schwarz, Leroy B. (2) Armony, Mor Iyer, Ananth Schwarz, Maike Atkins, Derek Jackson, Peter L. Scudder, Gary D. Babich, Volodymyr Jain, Apurva Seshadri, Sridhar Balakrishnan, Anant (2) Janakiraman, Ganesh Shen, Yuelin Bassamboo, Achal Johnson, M. Eric (2) Shi, Dailun Benjaafar, Saif Kapuscinski, Roman Shumsky, Robert A. Bish, Ebru K. Karaesmen, Fikri (2) Simmons, Donald E. (2) Blackburn, Joseph D. Karaesmen, Itir Sloan, Thomas W. Bradley, James Karalli, Serge M. Smith, Craig Burnetas, Apostolos Kavadias, Stylianos (Stelios) Sobel, Matthew J. (2) Buzacott, John A. Kazaz, Burak Song, Jing-Sheng Cachon, Gérard P. (2) Kelly, Frank Sox, Charles R. Caggiano, Kathryn E. (2) Keskinocak, P. (2) Stavrulaki, Euthemia (2) Carlson, Carol W. Khmelnitsky, Eugene Steinberg, Richard Cattani, Kyle Kim, Joon-Seok Stuart, Jr., Harborne W. Chand, Suresh Klastorin, Ted Su, Xuanming Chen, Hong Koenigsberg, Oded Suri, Rajan Chen, Rachel R. Kouvelis, Panos Szmerekovsky, Joseph G. Cheng, Feng Kwasnica, Anthony Taylor, Terry A. (2) Cheng, Joseph Lapré, Michael A. Tayur, Sridhar (2), (3) Chod, Jiri Lee, Chung-Yee Terwiesch, Christian (2), (3), (4) Cohen, Morris A. (2) Lederer, Phillip J. Teyarachakul, Sunantha Corbett, Charles J. Leung, Joseph Y-T. Thomas, Doug Dahan, Ely Li, Jingshan Tsai, Weiyu Dallery, Yves Li, Qing Tzur, Michal Daniels, Richard L. Lin, Grace (2) Vairaktarakis, George L. de Kok, Ton G. Loch, Christoph H. van Houtum, Geert-Jan de Matta, Renato Maglaras, Constantinos van Ryzin, Garrett de Véricourt, Francis (2) Majumder, Pranab (2) Vulcano, Gustavo DeCroix, Gregory A. (2) Mehta, Tushar Wang, Qiong Deng, Shiming Mendelson, Haim Ward, Amy Deshpande, Vinayak (2) Moinzadeh, Kamran Ward, Jame Ding, Qing Morrice, Douglas J. Warsing, Don Dong, Lingxiu Muckstadt, John A. (2) Webster, Scott Duenyas, Izak Muriel, Ana Weng, Z. Kevin Durbin, Erik Napoleon, Karen Willems, Sean P. Elmaghraby, Wedad (2) Natarajan, Hari Wu, Owen Erhun, Feryal Netessine, Serguei (2) Wu, S. David Erkip, Nesim Newman, Alexandra Xu, Yi Erkoc, Murat Oh, Se-Kyoung Yadav, Prashant Ettl, Markus Olsen, Tava Yan, Houmin Ferguson, Mark Ozen, Ulas Yang, Wei (2) Flowers, A. Dale Pangburn, Michael Yano, Candace A. (2) Frank, Murray Parker, Geoffrey Yao, David D. (2) Gaur, Vishal Pinedo, Michael (2) Zeevi, Assaf Gerchak, Yigal Pinker, Edieal J. Zenios, Stefanos (2), (3) Gilbert, Stephen M. Plambeck, Erica L. (2), (3), (4) Zhang, Hanqin Giloni, Avi Ramdas, Kamalini Zhang, Hongtao Grey, William Randall, Taylor Zhang, Jiang Groenevelt, Harry (2) Rappold, James A. (2) Zhang, Rachel Gulcu, A. Ren, Justin Z. (2) Zhao, Hui Gullu, Refik Robinson, Lawrence W. Zhao, Xuan Gürkan, Korhan Roundy, Robin Zhou, Yong-Pin Gurnani, Haresh Rudi, Nils (2), Zipkin, Paul (2) Haksoz, Cagri Ryan, Jennifer K. 1 2002 Manufacturing and Service Operations Management Conference Samuel Curtis Johnson Graduate School of Management, Cornell University, Ithaca, New York Price-Directed Control of a Deterministic Inventory Routing System Dan Adelman—University of Chicago The idea of price-directed control is to use an operating policy that exploits optimal dual prices from a mathematical programming relaxation of the underlying control problem. We apply it to the problem of replenishing inventory to subsets of products/locations, such as in the distribution of industrial gases, so as to minimize long-run time average replenishment costs. Given a marginal value for each product/location, whenever there is a stockout the dispatcher compares the total value of each feasible replenishment with its cost, and chooses one that maximizes the surplus. We derive this operating policy using a linear functional approximation to the optimal value function of a semi-Markov decision process on continuous spaces. This approximation also leads to a math program whose optimal dual prices yield values and whose optimal objective function gives a lower bound on system performance. We use duality theory to show that optimal prices satisfy several structural properties. We obtain near optimal performance on many problem instances. As the number of items grows, the overall approach scales easily enough to demonstrate guaranteed performance of less than 3\% from optimality on a real-world instance with 62 products/locations and traveling salesman costs. Mechanism Competition in Delay-Sensitive Markets Philipp Afeche—Kellogg School of Management Delay is an increasingly important dimension of service quality in manufacturing and service operations. Examples include data transmissions over the Internet and outsourced IT-services, where delay or "Quality of Service" is determined by the system's congestion. We consider the case of two capacity-constrained service providers who compete for price- and delay- sensitive customers. How should these providers invest in, allocate, and price resources in such settings? The existing congestion pricing literature focuses on the advantage that a firm derives from having a higher capacity (or a lower capacity cost) than its competitor. However, it assumes that the providers use the same price-service mechanism, i.e., rules that specify the resource allocation as a function of prices, to charge and serve their customers. We focus on another important, and increasingly relevant, dimension of provider differentiation: the choice of the price- service mechanism. Specifically, we let each provider choose between a standard posted- price FIFO service mechanism and an auction mechanism, whereby customers may submit competing bids to increase their resource allocation and hence reduce their service time. We study a number of questions that arise in this context: First, what are the equilibria in customer and provider decisions for a given pair of price-service mechanisms? Second, how do providers make their mechanism design and capacity investment decisions in equilibrium? How do these equilibrium decisions depend on the market and provider characteristics? Third, what is the relative value of service differentiation vs. capacity investment? To what extent can the use of a clever price- service mechanism make up for a cost disadvantage? We study these questions by developing a model that embeds a queuing system (to describe the relationship between demand, capacity investment, service-mechanism and service quality) within a microeconomic framework (to capture the customer and provider payoffs and their strategic interactions). The work completed to date studies the value of a priority auction vs. uniform pricing for a monopoly provider. The work in progress considers the strategic interactions that arise in a duopoly market. Outsourcing Contracts with Service Level Agreements Zeynep Aksin—INSEAD Francis de Véricourt—Duke University Fikri Karaesmen—Ecole Centrale Paris Call center outsourcing is a growing market, where companies seeking low cost call center services or companies trying to absorb demand fluctuations in their call centers, resort to outsourcing some or all of their incoming calls to external service providers. Such outsourcing requires a detailed specification of how calls are shared between the contractor and the contract service provider. We investigate some of the outsourcing contracts in order to gain insights into contract design and system performance issues. The existing contracts in the industry specify the maximum number of calls that the contract service provider accepts and the desired service level. In order to explore such contracts, we introduce a discrete-time model with random customer demand. The two parties first agree on a contract, where under different contracting schemes they are setting either price or service level related parameters. Given contract parameters, the contractor and contract service provider both determine their optimal capacity. Random demand is realized after the capacity decision has been taken, and demand is shared according to contract specifications. In any service outsourcing setting, the contractor faces a trade-off between cost and quality of call content. While large contract service providers can use their scale advantage to offer low cost services, if capacity is not dedicated, often the expertise of the agents answering calls may not be sufficient, resulting in customers who are not satisfied and call again. We analyze the sensitivity of contract design and system performance with respect to these scale and quality effects. 2 Sharing Congested Networks and the Pricing of Bandwidth Eddie Anderson—Australian Graduate School of Management Frank Kelly, Richard Steinberg —University of Cambridge We propose a method for the sharing of bandwidth among users of a communication network. Bandwidth refers to the amount of data that can be transmitted per unit time—usually via fiber optic cables—and is expressed in bits per second. In a packet-switched network, e.g., the Internet, the message to be transmitted is first broken down into small units, called packets, each consisting of a header and a data portion, which then travel along various routes through the network to the destination, where the data is re-assembled into the original message. Congestion in packet networks has traditionally been handled in a simple way: When a resource within the network becomes overloaded, one or more packets are lost; loss of a packet is taken as an indication of congestion, and the source sending the packets slows down. This system of congestion control has its obvious disadvantages, and has led to the introduction of congestion marking, whereby packets that encounter long queues have certain bits in their headers set to indicate congestion. Our proposal for sharing bandwidth is based on use of the congestion marks. In our proposal, the owner of the network sells contracts for usage over a period, written in terms of the congestion marks to be generated by each user. At the end of the period, users participate in a balancing process in which each makes or receives payments according to whether he generated more or fewer congestion marks then his contracted amount. Optimal Control of a Two-Stage Service Supply Chain Edward G. Anderson Jr., Douglas J. Morrice—McCombs School of Business, The University of Texas at Austin We investigate the dynamic behavior of service-oriented supply chains by developing a two-stage serial capacity management model. Reflecting the reality of many service (and custom manufacturing) supply chains, each stage holds no finished goods inventory, but rather only backlogs that can be managed solely by adjusting capacity. Using control theory, we develop centralized and decentralized optimal policies that trade off backlog costs against capacity adjustment costs when information is shared. The resulting control policies are shown to be intuitive and reasonable. Under the decentralized policy we establish the potential for an increase in demand variability along an optimally managed supply chain. Contrary to conventional wisdom, we also show that, while lead-time reduction does generally reduce backlog variance, it also increases capacity variance, resulting in a trade-off between service quality and personnel costs at each stage. Furthermore, such lead-time reductions increase backlog variances at subsequent stages resulting in a service quality trade-off between stages. Finally, we show that sharing backlog information will not materially improve overall supply chain performance if the target lead- and capacity adjustment times of the stage closest to end-customer demand are much smaller than subsequent stages’. In the second part of this presentation, we develop similar results for the centralized control policy. Then we compare and contrast the centralized and decentralized policies in terms of the bullwhip effect, backlog and capacity variance trade-offs at each stage of the supply chain, and variance and cost performance trade-offs between stages of the supply chain. Although the centralized control rule has a more complicated form and may be more difficult to operationalize in certain supply chains, it merits special consideration due to the benefits resulting from managing the supply chain globally rather than myopically. Project Sourcing Decisions in the Presence of Learning and Integration Cost Edward G. Anderson, Jr.—University of Texas Geoffrey Parker—Tulane University We present a general model of project sourcing that includes the effects of integration costs and learning over time on both the execution of projects and the integration of project output into complete products and services. We demonstrate that myopic decisions can create path-dependent traps where firms that outsource to gain short-term advantage can experience higher long-run costs. We describe conditions under which insourcing a small fraction of project activity may dominate either complete insourcing or complete outsourcing. Further, we show that very rapid rates of technological change are likely to be associated with outsourcing only if underlying supplier factor inputs, such as labor costs, are lower. Finally, we show that firms can rationally cycle back and forth between subcontracting project activities and performing them in-house. Multi-Item Lot-Sizing with Multiple Capacitated Batches Shoshana Anily, Michal Tzur—Tel Aviv University We consider a two-echelon system in which items are transferred from the upper echelon (for example, a warehouse) to the lower echelon (for example, a retailer) through a fleet of trucks, each with a finite capacity. The lower echelon is facing dynamic deterministic demand for several items, over a finite planning horizon. A truck incurs a fixed cost for each trip made from the upper to the lower echelon. In addition, there exist item dependent variable procurement costs and inventory holding costs at the retailer, which are both constant over time. The objective is to find a delivery schedule that minimizes the total cost, while satisfying demand on time. We address and partially resolve the question of the problem’s complexity by introducing a dynamic programming algorithm whose complexity is polynomial for a fixed number of items, but exponential otherwise. In addition, we present an alternative algorithm, which is an enumeration procedure, based on two simple extreme solutions; this algorithm, as shown by our numerical study, is capable of finding the optimal solution quite efficiently for a fairly large number of items. We also 3 suggest that the best of the two extreme solutions may be used as a heuristic solution; the efficiency of this heuristic is demonstrated in our numerical study as well. The Impact of Duplicate Orders on Demand Estimation and Capacity Investment Mor Armony—Stern School of Business, New York University Erica L. Plambeck—Stanford Business School, Stanford University Motivated by a $2.5 Billion inventory write-off by Cisco Systems in the Spring of 2001, we investigate how duplicate orders can lead to an overestimate of the demand rate and customers’ sensitivity to delay, and hence to excess investment in capacity. We consider a system with one manufacturer and two distributors. The manufacturer produces for each of the distributors according to a Poisson process with rate , whenever that distributor’s inventory falls below the base stock level. Demand occurs according to a Poisson process with rate at each of the distributors. If a customer finds that the distributor is out-of-stock, then with probability he will seek to make a purchase from the other distributor; if the other distributor is also out-of-stock, the customer will order from both distributors. When his order is filled by one of the distributors, a customer cancels any duplicate orders. Furthermore, the customer cancels all of his outstanding orders after a period of time that is exponentially distributed with rate . We show that if the manufacturer is unaware of the double orders, then she will overestimate both the arrival rate and the reneging rate , and hence will overinvest in capacity . This estimation error increases with the double order probability . In addition, we spell out the maximum likelihood estimators for and , assuming that the manufacturer can only observe the distributors’ inventory levels periodically. Due to bias in the estimators for and , the manufacturer may overinvest in capacity even if he anticipates the potential for double orders. Our results support assessments made in the popular press that Cisco’s over-investment in capacity cannot be blamed only on the external economic turndown, but was also caused by forecast error in the presence of double orders. Market Segmentation, Channel Differentiation and Channel Structure under Price and Service Competition Derek Atkins, Xuan Zhao—Faculty of Commerce and Business Administration, University of British Columbia This paper explores market segmentation, channel differentiation and channel structure under price and service competition in a heterogeneous market. A market in which consumers differentially value the level of service (or quality) of a product is likely to be served by products with different (price, service) combinations. We wish to study how this heterogeneous market affects the structure of the supply chains delivering these different products. Should we expect such channels to remain integrated, or should the channel providing the higher/lower service or both prefer to decentralize? How does competition between the channels affect the choice of channel structure? What channel structures provide a stable equilibrium (Nash)? How are these results different between a "mature market" (represented by a fixed choice of service level) and an emerging market (where the service level is a decision variable)? What is the impact of channel power or leadership on these results? These and other results are presented in this paper. By studying a two channel multi-stage game with complete information, we analyze the Nash equilibrium supply chain structure under only price competition and the subgame perfect Nash equilibrium supply chain structure under both price and service competition. We provide an analytical framework and some managerial suggestions for companies both in a mature market and those planning to enter a new market. We find that in the pricing game, having channels vertically integrated (II) is always a Nash equilibrium. Having both channels decentralization (DD) can also be a Nash equilibrium if horizontal competition is strong, in agreement with the literature. But II may not be the equilibrium market structure in the whole multi-stage game (with price and service competition), rather DD can be the unique equilibrium. Furthermore, II and DD can both be equilibria at the same time. This is the key difference from the price only scenario studied in the literature. Our model does not use the market substitutability assumption in most related literature, but uses instead a consumer preference model for heterogeneous service levels to derive the demand functions experienced by competitive supply chains. An additional finding is that DD can make the industry better off even if it is not an equilibrium. So tacit collusion can sustain DD in the repeated game. Although unilateral decentralization by one channel may damage its own profit, it may increase the total industry profit. A cooperative contract between the two channels that allows for the strategic decentralization of one channel can be beneficial to both parties. Along the way of explaining our main results, many other related problems are also addressed. Finally we provide managerial implications and put forward hypothesis for further empirical studies. Operational Decisions of the Pre-IPO Firm Volodymyr Babich, Matthew J. Sobel—Weatherhead School of Management, Case Western Reserve University We consider operational and financial decisions that the owners (entrepreneurs) of a growing, privately owned firm are facing and the effect of these decisions on the timing of the Initial Public Offering (IPO). We assume that the owners (1) maximize the expected present value of the proceeds from the firm’s IPO, which they consider as a cash-out opportunity; and (2) they are willing to forgo immediate financial rewards and channel profits into further growth. This work continues a stream of research that links operational and financial decisions of the firm. The operational decisions in this model are productive capacity and production level. The financial decisions are the timing of the IPO, sources of external capital (bank loans or venture capital), and the amount borrowed by the firm. Financial and operational decisions cannot be decoupled because adequate capital is crucial for operational decisions to be feasible. Similarly, operational decisions affect the amount of the internally generated working capital and thereby influence the availability of external financing. We model the IPO event as a stopping time problem for an infinite horizon, discounted Markov Decision Process. Unlike traditional stopping time models, in our analysis we consider other decisions, such as, optimal production, sales and loan 4 size, at every stage of the model. The results include characterizations of an optimal capacity expansion policy and sufficient conditions for a monotone threshold rule to yield an optimal IPO decision. Delivery Date Commitment with Adaptive Scheduling Anant Balakrishnan, Hari Natarajan—Smeal College of Business, Penn State University As manufacturers strive to become more customer-oriented, they are reexamining their order negotiation systems and processes to ensure responsive but realistic delivery date commitments to customers. Commercial Available-To-Promise (ATP) systems focus on data integration across the sales and production planning functions of a firm to provide real-time information on product availability. However, successful supply chain performance requires not only data integration, but also proactive coordination of the activities across these functional areas. Motivated by a problem facing a commercial-pipe manufacturer, we discuss a model-based approach to integrate and systematically evaluate tradeoffs between order commitment and production scheduling decisions. The manufacturer produces and sells a wide variety of products to customers with different priority levels. Due to significant product changeover times, planners decide on production campaigns for different product families several weeks in advance using planned inventory targets, and progressively assign customer orders to planned production as orders arrive. Residual production is inventoried to meet demand until the next campaign for that product. Currently, the sales department uses static lead times, independent of facility loading and order size, to promise delivery dates; consequently, customers are often dissatisfied due to poor delivery performance. To support more principled delivery date negotiations with customers, we develop an optimization model that concurrently considers order commitments and production plans. The model, tailored for hybrid production environments in which planned production is partially committed to firm orders, permits modifying the existing schedule if needed, and incorporates options for reallocating planned production time across product families, and reducing planned inventory accumulation in order to accommodate quick delivery needs for important customers. We formulate the delivery date commitment problem as a mixed integer program that minimizes the cost of order tardiness, considering different customer priorities, and the exposure cost of reducing planned inventory. Since, this problem is NP-hard, the goal of this research is to develop and test effective solution methods. We develop optimization-based solution methods, and evaluate their computational performance. Stochastic Inventory Models with Demand Stimulation Anant Balakrishnan, Michael Pangburn, Euthemia Stavrulaki—Smeal College of Business, Penn State University For certain product categories, such as moderately priced novelty or impulse purchase items (e.g., toys, tools, consumer electronics products), there is evidence that higher stocking levels at retail stores can induce a greater likelihood of purchase by consumers. The operations management literature has analyzed optimal lot-sizing policies when inventory stimulates demand, but primarily in deterministic settings. In this paper, we analyze periodic review inventory management policies for products when demand increases with inventory level. Gerchak and Wang [1994] considered stochastic inventory models in the presence of demand stimulation, assuming a specific multiplicative inventory influence function (that describes how demand varies with inventories), and focused on characterizing conditions under which a myopic newsvendor policy is optimal for the multi-period model. We consider a general inventory influence function, and characterize the profit- maximizing order quantity for a single period stochastic inventory model. We demonstrate that when higher inventories can significantly increase demand, stocking more than the maximum possible demand can be optimal. We also compare the optimal single-period ordering policy when inventories stimulate demand with an adaptive policy that uses the standard newsvendor formula to choose order quantities but periodically updates the demand distribution based on observed demands. We show that the latter policy converges to a sub- optimal equilibrium quantity that is lower than the optimal order quantity. We employ the uniform distribution to provide additional insights on the structure of the optimal solution and to quantify the differences in order quantities and profits for the optimal and adaptive policies. We demonstrate that using a uniform distribution implies an optimality condition that resembles the critical fractile condition of the standard newsvendor model. Finally, as in Gerchak and Wang, we show how the single- period results extend to multiperiod settings with demand stimulation. Constructing an Optimal Portfolio of Capacity Options Achal Bassamboo, Stefanos Zenios—Stanford University Capacity options are used in the supply chain to leverage uncertainty in demand and to limit downside risk. Consider a firm that is launching a new product in the market. Capacity options with different maturities are available in the market. We use dynamic programming to determine the optimal dynamic portfolio of these capacity options. The new product’s life cycle is modeled by a Markov chain with discrete states. The product in each period can be in one of several stages. In each period, a random demand for the product is observed whose distribution depend on the stage. At the end of each period the firm purchases a portfolio of capacity options by observing the current state and the current portfolio. In the next period it observes the new state and demand. It exercises the options appropriately. The objective of the firm is to minimize the long run discounted procurement cost. Assuming that the Markov chain has an acyclic topology, we show that the optimal dynamic portfolio can be constructed iteratively. Each iteration involves solving a newsvendor problem, and the optimal portfolio is obtained in terms of the critical fractile solution to this problem. While one would expect that the optimal dynamic portfolio would involve purchasing options of different maturities, conditions are derived such that it is always optimal to purchase options with a single maturity because the long time maturity options tend to be costlier in discounted sense than short time maturity options. So it is always better for the firm to wait till they 5 have more information regarding the random demand it is going to observe in the following period and then decide how much to buy. When Does Higher Demand Variability Lead to Lower Safety Stocks? Saif Benjaafar, Joon-Seok Kim—University of Minnesota We study the effect of demand variability in multi-item inventory systems where supply lead-times are endogenously generated by a production system with finite capacity. Counter to intuition, we show that higher demand variability leads to a lower coefficient of variation in lead-time demand. We also show that higher demand variability leads to a smaller fraction of total stock being devoted to safety stock. More significantly, we show that a sufficiently large increase in demand variability can lead to the elimination of safety stocks altogether. These results suggest that strategies used to cope with lead-time demand variability may be less valuable when demand variability is high. We verify this intuition by examining the benefits of two such strategies, demand pooling and advanced order information. Under plausible conditions, we show that the value of pooling and advanced order information indeed decreases with demand variability. In particular, we find that the relative advantage of both strategies is insignificant when demand variability is high. Finally, we show that the effect of higher demand variability parallels the effect of higher capacity utilization. Optimal Investment Strategies for Flexible Resources Under Price Elasticities and Demand Correlation Ebru K. Bish, Qiong Wang, Ana Muriel We study the capacity investment decision faced by a two-product firm under price elasticities and demand correlation. The firm can invest in any amount of dedicated capacity for each specific product, and/or in flexible capacity, which can produce either of the products. Our objective is to characterize the properties of the optimal capacity investment strategy and to understand the impact of demand correlation, price elasticities, and investment costs on the optimal decision. We consider a two- stage model: In the first stage, the firm decides how much dedicated and flexible capacity to install under high demand uncertainty. In the second stage, demands are realized and the firm optimizes its profit through pricing and production decisions, constrained by the capacity investment vector determined in the first stage. Our analysis provides the structure of the optimal investment strategy as a function of price elasticities and investment costs. We also show that (1) it can be optimal for the firm to invest in the flexible resource even when product demands are perfectly positively correlated, and (2) it can be optimal not to invest in flexible capacity even when product demands are perfectly negatively correlated. Based on our analysis, we provide principles on the optimal capacity investment strategy and offer managerial insights. This study extends the works by Van Mieghem (1998) and Fine and Freund (1990) to include price elasticities. A Methodology for Valuing Time in Supply Chains and Make-to-Stock Manufacturing Joseph D. Blackburn—Vanderbilt University Although response time is a critical dimension of competition in manufacturing supply chains, the strategies applied in practice vary widely with respect to speed of response. Some organizations are actively compressing time in the supply chain while others within the same industry are making sourcing choices that embody slower, not faster, response time. This study outlines a methodology for valuing time in make-to-stock manufacturing to help clarify the costs and benefits of changing the speed of response. We develop simple bounds for the inventory-related effects on product cost of a one-week change in leadtime that apply under a wide range of conditions, including cases in which inventory is not managed optimally. These results are applied to the case of a footwear manufacturer considering sourcing decisions that would increase the leadtime but reduce product cost. The case demonstrates why quick response strategies have failed to make U.S. domestic manufacturers competitive in footwear and apparel. Improved Base-Stock Policies Under Order Crossover James R. Bradley, Lawrence W. Robinson—S.C. Johnson Graduate School of Management, Cornell University Robinson et al. [2001] have shown that using the common approach of setting base stock levels based on the variance of lead-time demand can lead to excess inventory and cost when order crossover is possible. ("Order crossover" occurs whenever replenishment orders do not arrive in the sequence in which they were placed.) They show how the effect of order crossover can be correctly taken into account by setting inventory levels using inventory shortfall rather than the lead-time demand. We suggest that the optimal base stock with respect to shortfall can be approximated by assuming that the shortfall is Normally distributed, which requires that the variance of the number of orders outstanding be computed. Computing this variance precisely, however, requires more information than may be available in practice, and also may be cumbersome. This paper presents easily computed upper bounds on the variance of the number of orders outstanding, which provide a means to set inventory policies that take order crossover into account, and improve significantly upon policies set according to lead-time demand. We show that this approximation is close to optimal for a family of negative binomial-binomial lead-time distributions. Quantity Discounts in Single Period Supply Contracts with Asymmetric Demand Information Apostolos Burnetas—Case Western Reserve University Stephen M. Gilbert—University of Texas 6 Craig Smith—McKinsey & Co., Inc. We investigate how a quantity discount schedule can be used to influence stocking decisions and supply chain performance in single period interactions between a supplier and buyer(s). In contrast to much of the work that has been done on single period supply contracts, we assume that there is no opportunity for ongoing interactions between the supplier and the buyer(s) after demand information is revealed. Furthermore, we assume that there are either heterogeneous buyers that face different distributions of demand or that there is a single buyer that has better information about the distribution of demand than does the supplier. We discuss the relationship between the design of a quantity discount schedule and the product-line design problem. Although these two problems have similarities, the introduction of quantity as the dimension of differentiation among market segments alters the structure of the problem. Specifically, in the design of a quantity discount schedule, the buyer can select along a continuum of quantities, whereas in the traditional product-line design problem, the buyer is restricted to a discrete set of products. We drive structural properties of this unique problem that facilitate managerial insights and solution procedures. Risk Analysis of Take-or-Pay Contracts with Forecast Updates John A. Buzacott—Schulich School of Business, York University Houmin Yan—Dept of Systems Engineering and Engineering Management, Chinese University of Hong Kong Hanqin Zhang—Institute of Applied Mathematics, Academy of Mathematics and Systems Science A take-or-pay contract is an agreement between a buyer and a supplier. It specifies a minimum volume the buyer must purchase (take), and the maximum volume the buyer can obtain (pay). Take-or-pay contracts belong to the class of volume flexible contracts which have attracted much attention recently. However, the supply contract literature mainly deals with models for the optimization of the expected value of a given cost or profit function. Although risk analysis is a standard component of finance related studies, very little has been done in supply contract models. With highly uncertain supply and demand conditions, it is absolutely necessary to consider risk issues. Furthermore, it is essential to understand how the quality of demand forecast updates influence decision making. In this paper, we study take-or-pay supply contracts using a mean-variance formulation. We show that a mean-variance trade-off analysis can be carried out efficiently. Our major result are: (1) the mean-variance objective is concave with respect to both the "take" and "pay" portions of the contract; (2) for the case of worthless and perfect demand forecast revisions, we derive an explicit optimal solution. With the above results, we further develop insights into risk and contract management. We show that the behavior of mean-variance objective targets at precision. Moreover, we also develop a monotonicity result with respect to the quality of forecast revision. For general forecast revisions, we implement algorithms for finding the optimal decision numerically. Our numerical experiments reveal that the mean-variance type of risk analysis is practical. As expected, the strategies developed with risk measures out-perform the strategies based on expected value of objective measures. In particular, risk analysis is efficient and effective when the quality of forecast revision is low to medium. Furthermore, our study indicates that it is possible to reduce the risk (measured by its variance) by six to eight fold, while the loss in the expected profit is almost invisible. At the same time, our study indicates that the strategies developed for the expected value of objective measures can only be considered when the quality of forecast revision is high. Advanced Purchase Discounts and Supply Chain Inventory Management Gérard P. Cachon—The Wharton School of Business, University of Pennsylvania O’Neill Inc is a large supplier of clothing for water sports: surfing, diving, water ski, wake boarding, etc. Like other producers of fashion apparel, they face uncertain demand and short selling seasons. Hence, matching supply to demand is one of their key challenges. O’Neill’s retailers also face the same problem. Ideally, retailers would like to order and receive delivery of at least some inventory during the selling season. They could then use informative early season sales to direct their replenishment towards the faster selling products. On the other hand, all else being equal, O’Neill prefers to receive the retailers’ orders well in advance of the selling season so that cheaper production can be implemented and capacity constraints are not binding. In fact, O’Neill has some control over when the retailers order. At one extreme, O’Neill could insist that retailers pre-book their entire season commitment well before the season begins. If O’Neill then produces no more than the retailers order, O’Neill would be unable to fill any replenishment requests during the season. At the other extreme, O’Neill could give no incentive to pre-book, in which case the retailers would then delay their orders as much as possible. The intermediate solution is to offer retailers an advanced purchase discount on the units they pre-book before the season to encourage them to at least pre-book some inventory. This paper studies a model that captures these issues. There are two periods and in the initial version of the model there are two firms, a supplier and a retailer. (The extension to multiple retailers is also discussed.) Before the game begins the firms have a common forecast for demand. The following sequence of events occurs in period one: the supplier announces the wholesale prices for each period; the retailer submits a pre-book order; the supplier chooses a production quantity; and at the end of the period the supplier delivers the retailer’s pre-book order. At the start of period two the retailer receives some information regarding demand and then submits a second order. The supplier may be able to do a second production in period two, but that production is more expensive than period one production. Furthermore, the inventory from that production can be delivered to the retailer before demand occurs in the period. In this setting the supply chain must balance the cost advantage of early production with the information advantage of later production. The equilibrium pricing, production and ordering strategies are derived for both rms. The settings in which the supplier offers an advanced purchase discount are characterized. Retailing Assortment Decisions under Consumer Search Gérard Cachon, Christian Terwiesch, Yi Xu—The Wharton School of Business, University of Pennsylvania 7 Consumers often know what kind of product they wish to purchase, but do not know which particular model best fits their needs. For example, a consumer shopping for a new bicycle might be able to assess the utility associated with a bicycle in the store she visits, but typically faces uncertainty about the utilities associated with bicycles outside of the retailer’s assortment. Hence, even if she finds an acceptable bicycle at the retailer, she may nevertheless continue her search at other retailers in the hope of finding an even better bicycle. As a result, since search impacts the likelihood a consumer purchases in any given store, search should be considered in the retailer’s assortment decision. We develop a general decision framework for the retailer’s assortment problem that incorporates consumer search, thereby extending current approaches to assortment planning which do not explicitly account for consumer search. With our model we obtain the following three results. First, a retailer who decides whether or not to add a product to the assortment should not only look at the direct costs and revenues of the product, but also anticipate the indirect benefit that an extended assortment has in preventing consumer search; it may be optimal to add a product to the assortment even if the additional assortment costs exceed the revenue gains associated with that product. Second, we consider a retailer who chooses his product assortment by applying the Multinomial Logit (MNL) model, which is a special case of our general consumer choice model. Following the MNL, the retailer estimates the fraction of consumers who abstain from purchase. However, this estimation does not distinguish between consumers who abstain from purchase entirely and those who don’t purchase at the retailer because of search. As a result, we demonstrate that this traditional approach may result in an assortment with fewer items than optimal. Third, we consider the cyclical application of the traditional approach: the retailer estimates the parameters of the MNL, chooses an assortment, and then continues to repeat those two steps. We show that this iteration may lead to a non-optimal assortment even though the parameter estimates are consistent with the chosen assortment. Real-time Capacity and Inventory Allocation for Reparables in a Two-Echelon System with Emergency Shipments Kathryn E. Caggiano, John A. Muckstadt, James A. Rappold—University of Wisconsin-Madison In this paper, we examine a multi-echelon system for reparable service parts with finite repair capacity under periodic- review. We develop a finite-horizon decision model for determining the optimal allocation of repair capacity as well as the optimal allocation of service parts across multiple field stocking locations. The decision model considers demand uncertainty explicitly and is designed to maximize the likelihood of having sufficient service parts for allocation to stocking locations. Specifically, we employ a novel capacity allocation function that gives higher repair priority to those parts that are least likely to remain unused in inventory for long periods of time. We compare the performance of simple allocation rules, such as first-come first-served, with more sophisticated rules that consider various amounts of system information. Of particular concern is the computational efficiency of the approach and its applicability to large-scale service parts repair and distribution systems. Textured and Spackled: Dual Strategies of Make-to-Stock and Make-to-Order That Integrate Marketing and Operations Perspectives Kyle Cattani—The Kenan-Flagler Business School, UNC Chapel Hill Ely Dahan—MIT Sloan School of Management Glen M. Schmidt—The McDonough School of Business, Georgetown University Using a manufacturer of messenger bags as motivation, we model the choice of make-to-order (MTO) versus make-to- stock (MTS) production from a joint marketing/operations perspective. From a marketing view, MTO can help satisfy diverse customer preferences through customized production, but some customers may prefer standard off-the-shelf products made via MTS. From an operations perspective, MTS may be the lower-cost method, because more efficient production is possible given smoother schedules and less need for flexibility, but obsolescence and stockout costs may favor MTO. Our model suggests that even when most customers prefer standard product(s), the firm may want to adopt a dual approach utilizing both MTO and MTS production. In this case, variable MTO production volume can be layered onto constant MTS output to meet variable demand. This uneven, or textured production schedule balances the higher production cost of MTO against the ability to more closely match demand volume. When some customers prefer a standard product but many others opt for a customized product, we again find situations where the firm may want to use a dual approach even within the same production facility - MTS production yields standard products that are used to restock inventory. This filled in, or spackled production mitigates the daily swings in MTO production, thus reducing overall production cost through improved capacity utilization. In this paper we identify the conditions under which a dual MTS/MTO production setting may be preferred over strict MTS or strict MTO. The JIT Inventory Revolution. What Actually Happened to the Inventories of American Companies Between 1981-2000? Hong Chen, Murray Frank, Owen Wu—Faculty of Commerce and Business Administration, University of British Columbia In the 1980s Japanese manufacturing companies made substantial market share gains in the U.S. markets in a range of industries, including most notably the car industry. This stimulated a significant search for the key to their success. There were many calls for a revolution in inventory policies of American firms. The ―Just-In-Time‖ inventory system was often identified as a key element. It was said that American firms needed to reduce their inventory holdings. Two decades later such calls are much less frequent. Did the revolution actually take place? In this paper we study the inventory holdings of publicly traded companies in the manufacturing, wholesale and retail sectors of the American economy over the period 1981-2000. Our study shows that on average American firms did reduce their inventory holdings, but the observed reduction was mainly restricted to the work-in-process inventory rather than raw materials or finished goods inventories, and to manufacturing rather than wholesaling or retailing. That is, the JIT revolution seems to have achieved some success in reducing the work-in-progress inventory of manufacturing firms. However, there is no evidence 8 that JIT has made any impact on the reduction of the inventory used for smoothing the material flows across firms. This supports the current emphasis on improving the whole supply chain. Supporters of JIT often suggested that inventory reduction would provide financial benefits. Therefore we study the financial market evaluation of the inventory holdings. First we find that there is little evidence that inventory holdings had any impact on the financial market valuation of firms; for good or for ill. Second, there is no evidence that excess inventory results in firm failure. Typically firms have unusually low inventory during the two years prior to bankruptcy, liquidation, or acquisition. Thus excessive inventory does not appear to commonly contribute to corporate failure. Efficient Auction Mechanism for Supply Chain Procurement Rachel R. Chen, Ganesh Janakiraman, Robin Roundy, Rachel Q. Zhang—Cornell University In this paper, we introduce multi-unit Vickrey auctions that incorporate both production costs and transportation costs in supply chains involving multiple supplier locations and demand centers. To the best of our knowledge, this is the first attempt to analyze auctions in complex supply chains. We propose two auction mechanisms with VCG payment structure, Auction T and Auction R. Both auctions are incentive compatible for the suppliers and efficient for the quantities awarded. While Auction T is simpler and easier to implement, the buyer's payments may be unreasonably high at times, or hard to determine with asymmetric suppliers. So we introduce Auction R under which the buyer reveals a consumption utility function that the auctioneer will take into consideration while making production and transportation decisions. We show that under Auction R, the buyer is guaranteed to pay less. However, the buyer does not in general have the incentive to report her true consumption utility. We then derive the buyer's bidding function that maximizes her net utility under full information. It turns out that, in Auction R, when the buyer follows the optimal strategy, she pays a uniform price for all the units awarded to a demand center. This property makes Auction R potentially attractive for implementation, as in most uniform price auctions, incentive compatibility usually does not hold. Note that, under Auction R, the quantity awarded to the buyer is one of the outputs of the auction. We also consider buyer's bidding strategy in Auction R when she is uncertain about the suppliers' true production costs. When there is uncertainty in the system, both the awarded quantities and the buyer's payments are no longer predetermined. One needs to consider the tradeoffs among the expectation and variability of the awarded quantities, and the expectation and variability of the payments. Since optimization over a set of functions is in general very challenging, we focus on a special set of bidding functions for the buyer. Numerical examples show that, choosing the right parameters in the functions, one can predict the awarded quantities and payments with little or zero uncertainty. To illustrate the importance of incorporating transportation costs into auctions, we consider auctions, called S, under which the auctioneer decides the production quantities at all supplier locations solely based on the suppliers' bids and the demand, and the transportation decisions are made subsequently. Although Auction S is incentive compatible, numerical examples show that considerable supply chain cost savings can be achieved if one runs Auction T rather than Auction S. However, the buyer may favor Auction S due to lower total payments under certain circumstances. This work can be easily extended to the case with one supplier and multiple buyers and to include some of the other costs associated with integrating a supply chain. Future work needs to be done to extend this to supply chains with multiple suppliers and buyers. Risk and Flexibility in Managing Supply Contracts Feng Cheng, Markus Ettl, Grace Lin—IBM TJ Watson Research Center Maike Schwarz—Dept. of Statistics, University of Hamburg David D. Yao—IEOR Dept, Columbia University We develop a framework that describes the effect of contract terms and conditions on managing supply chain risk. The model considers the efficiency of the whole supply chain, and the distribution of gains between buyers and suppliers. We discuss how each party should behave in contract negotiations by numerically investigating the impact of contract parameters such as base price, risk premium, and quantity flexibility. We consider a simple two-party supply chain that consists of a supplier and a manufacturer. The manufacturer orders products from the supplier, and then sells them to end customers. End customer demand is uncertain. At the beginning of a period, the manufacturer places an order of quantity Q, paying a base price v for each unit. In addition, the manufacturer can purchase q option contracts from the supplier, at a cost of c per contract. Each option contract gives the manufacturer the right, but not the obligation, to receive an additional unit from the supplier at a premium price w at the end of the period, after demand is realized. We develop an optimization model to obtain optimal order and pricing policies. The manufacturer's problem is to decide the committed purchase quantity, and the number of option contracts that should be purchased. The supplier's problem is to decide the option price, and the price premium, based on the flexibility that the manufacturer requested. We analyze the model under three different scenarios: first, when the supplier has no information on the manufacturer's demand; second, when the manufacturer and supplier share demand information; and third, when the supplier and manufacturer share demand information, and use a profit-sharing agreement to maximize the total expected supply chain profit. We develop conditions for a win-win situation where both manufacturer and supplier are better off than in a traditional newsvendor setting. Deriving The Economic Run Quantity in a Perfectly Competitive Market Joseph Cheng, Donald Simmons—Ithaca College 9 The current approach to economic run quantity or optimal lot sizes is often based on the principle of cost minimization for a predetermined quantity of output, which is independent of the economic run quantity. This paper focuses on profit maximization, which is more general than cost minimization since it allows for variation in product quantity and product mix. In economic theory, the most commonly used model is the perfectly competitive market where firms face a perfectly elastic demand. Under such an environment, firms may sell any quantity at the current market price and, therefore, seek to produce a level of output that would maximize profit at the current market price. This means that the total quantity to be produced by a firm is not predetermined but rather is a critical decision variable. In addition to the drawback of assuming fixed output, cost minimization may ignore an opportunity cost in terms of profit foregone. For example, in a real-world manufacturing setting, it is uneconomical to allow the facility to become idle after each production run. In most cases, the facility would be re-arranged to manufacture another product. During this setup, there would be a loss of potential profit, which could otherwise be earned if these man-hours were used for producing the first product. No explicit consideration for profit foregone has been made in the traditional formula. Would simply adding such foregone profit to the setup cost in the traditional formula solve the problem? The answer is "not quite". The reason for this is that there are setup and carrying costs for a second product, which should be minimized as well. In a multi-product firm, one cannot reach an optimal solution for any single product in isolation from other products, which is the fundamental flaw of the traditional formula. In order to assure that the decision is optimal for the firm as a whole, the profit function to be maximized must take into account all products and all costs, including opportunity costs. The traditional economic run formula will only attain a sub-optimal solution. For a truly optimal solution, the product volume and product mix decisions must be coordinated with the economic run decisions to ensure that profit is maximized overall. This paper offers a superior solution to the problem of product lot sizing by focusing on the bottom line (profit) rather than on the middle line (cost). Resource Flexibility with Responsive Pricing Jiri Chod, Nils Rudi—University of Rochester Flexibility enables firms to better respond to volatile market conditions. This article studies two types of flexibility, namely resource flexibility and responsive pricing. We consider a situation where a single resource can be used to satisfy demands for two distinct products. While the resource's capacity must be decided based on uncertain demand curves, it's allocation to the two products is made after accurate demand information is available. By responsive pricing, the firm gains additional flexibility since resources are not only allocated based on demands but rather the demands themselves can be influenced. Our model reflect the interdependency between these two types of flexibility and highlights the effect of key drivers such as demand elasticity, demand variability and demand correlation. We further illustrate the value of flexibility by contrasting our results to the situation where the capacities for two specific resources are locked in before accurate demand information is available. Finally, we consider supply chain issues by incorporating the supplier of the firm's resource into the model. Measuring Imputed Costs in the Semiconductor Equipment Supply Chain Morris A. Cohen, Teck H. Ho, Justin Z. Ren, Christian Terwiesch—The Wharton School, University of Pennsylvania We consider the order fulfillment process of a customized capital equipment supplier. Prior to receiving a firm purchase order from the customer, the supplier receives a series of shared forecasts, which are called ―soft orders‖. Facing a stochastic internal manufacturing lead-time, the supplier must decide at what time to begin the fulfillment of the order. This decision requires a trade-off between starting too early, leading to potential holding or cancellation cost, and the cost of starting too late, leading to potential loss of goodwill. We collect detailed data of shared forecasts, actual purchase orders, production lead-times, and delivery dates for a supplier-buyer dyad in the semiconductor equipment supply chain. Under the assumption that the supplier acts rationally, optimally balancing the cancellation, holding, and delay costs, we are able to estimate the corresponding cost parameters based on the observed data. Our estimation results suggest that the cost of cancellation is four times higher and the holding cost is two times higher than the delay cost. In other words, the supplier is very conservative when commencing the order fulfillment, which negates the effectiveness of the overall forecast sharing mechanism. Drivers of Diffusion of ISO 9000 and ISO 14000: Findings from a Global Survey Charles J. Corbett—The Anderson School at UCLA ―Greening the supply chain‖ is often mentioned as a way of promoting environmentally responsible behavior in emerging economies without resorting to multilateral governmental agreements. Firms in leading countries would set the standard and then require their suppliers abroad to follow these standards, for instance by requiring ISO 14001 certification. This paper examines whether such management standards do indeed diffuse through global supply chains, by focusing on the first such standard, ISO 9000. We modify the classical Bass diffusion model to a global context, and use it as a theoretical basis from which we derive a number of hypotheses about the relations between relative penetration levels and the role of exports in certification behavior. We use macro-level certification data to compare penetration levels across different countries and region. We then use firm-level data from a global survey of over 5000 firms in 9 countries to test the hypotheses. We find strong support for the commonly held assumption that export pressure was indeed a key driver of global diffusion of ISO 9000, which in turn suggests that it is possible to promote management practices through supply chains. Analysis of General Assembly Systems: Approximations versus Assumptions 10 Dr. Ton G. de Kok—Director, European Supply Chain Forum, Department of Technology Management The analysis of general assembly systems has received considerable attention during the last two decades. Due to its structural complexity one typically finds exact analysis only for simple (e.g. two-stage) systems under simplifying (e.g. zero lead time) assumptions and suboptimal (e.g. base stock) policies. Alternatively, one analyses general structures heuristically based on simplifying assumptions on the interaction between subsequent assembly stages and on the interaction between items that are assembled into a (sub)assembly at an assembly stage. In this presentation we discuss the pitfalls that one may encounter when making particular assumptions on the interaction between items in a general assembly systems. In particular we discuss the so-called decomposition assumption and assembly assumption. These two assumptions play a major role in the heuristic analysis of general structures in several recent papers. Throughout the presentation we restrict to periodic (echelon) order-up-to-policies, i.e. base stock policies. The decomposition assumption states that the probability that a child (i.e. upstream) item of an item is not available is negligible. This assumption is complied with by assuming a non-stockout probability for each item of at least 90%, say. The assembly assumption states that if upon release of an order for an item this not all child items are available, then with probability 1 only a single child item is not available. Our assessment of the consequences of these assumptions is based on cost comparisons between resulting optimal solutions for the base stock levels. As a benchmark for comparison we use the recently postulated concept of Modified Base Stock policies, that enable an exact analysis of general assembly systems and which is (close-to) optimal for serial and divergent supply chain structures. The main insight obtained is that both the decomposition and assembly assumption restricts the set of feasible solutions for the base stock levels to suboptimal solutions. Scheduling Patient Operations To Minimize the Number of Post Anesthesia Care Unit Nurses Renato de Matta—University of Iowa Vernon Ning Hsu—Hongkong University of Science and George Mason University Chung-Yee Lee—Hongkong University of Science and Texas A&M University Health care managers pay close attention to contain and reduce the post anesthesia care unit (PACU) cost of an ambulatory surgical center (ASC) that specializes in providing elective surgeries to patients. Studies suggest that scheduling patients in the ASC's operating rooms to regulate patient flow to the PACU might potentially reduce the peak number of patients in the PACU and therefore reduce the PACU capacity needs, i.e, number of nurses, and the PACU cost thereof. But to date, no in depth study of this important problem is available in the scheduling literature. Motivated by applications in process industries, we formulate the patient scheduling problem as a no-wait, two-stage process shop problem with two objectives: (a) minimize the peak number of patients subject to constraints that all patient surgeries scheduled daily are completed, and that the last patient leaves the PACU within the ASC designated daily closing time; and (b) minimize makespan. Objective (a) ensures the smallest number of nurses, while (b) maximizes PACU nurses’ utilization. We present computational complexity results and define the boundary between easy and hard cases. We also develop an exact polynomial algorithm to solve the easy case and a tabu search- based heuristic algorithm to solve the hard case. Our algorithms are shown to be very effective in finding optimal and near optimal schedules, respectively, on a set of hospital data. Managing Response Time and Service Quality in a Call Allocation Problem Francis de Véricourt—Fuqua School of Business Yong-Pin Zhou—University of Washington-Business School As call centers strive to lower customer-waiting time, increase capacity utilization, and lower cost (or, increase revenue), a common practice is to encourage customer service representatives to treat calls as fast as possible. This is usually achieved by compensating representatives on the number of served calls over a period of time. Even though call handle time will be reduced, managers are also aware that the provided service quality may deteriorate. The link between service quality and revenue is often viewed only on a strategic level (through customer loyalty for instance). On an operating level, the main approach to ensure service quality remains call monitoring and recording. The performance measures of quality are rarely taken into account when calls are allocated to available agents. In this paper, we analyze a call allocation problem in which customer service representatives show different service speeds and quality levels. We model service quality as the probability that a customer is not satisfied and reenters the system after leaving. There may be certain representatives who handle calls very quickly but the customers they server are more likely to reenter the waiting line. Other representatives may provide longer but better answers. The manager has then to solve this tradeoff when deciding to which agent an arriving call should be sent in order to minimize the average waiting time. We formulate the problem as a Markov Decision Process, and we partially characterize the optimal allocation policy in this framework. This result allows us to derive insights on how to take both response time and quality of the answer into account when operating the system. Policy Evaluation and Optimization for a Series System with Returns Gregory A. DeCroix—Fuqua School of Business, Duke University Jing-Sheng Song—Graduate School of Management, University of California Paul Zipkin—Fuqua School of Business, Duke University Consider a series system with stochastic demand, constant leadtimes and no scale economies. The standard model has nonnegative demand, but here we relax this assumption. The aim is to model systems with returns. 11 We show how to evaluate echelon and local base-stock policies (which are different in this context). The calculations required are considerably more intricate than in the nonnegative-demand case. We also assess some simpler approximations. Finally, we discuss methods for policy optimization. Inventory Management for an Assembly System with Product Returns Gregory A. DeCroix, Paul Zipkin—Duke University We study an assembly system that experiences product returns. A returned product may be entirely reusable or only partially reusable—i.e., it may only be possible to reuse some parts or subcomponents of the product. We explore the impact of such returns on the management of the system. In particular, we examine the impact of partial reusability on the optimal stocking decisions for reusable and non-reusable components. For the steady-state version of the problem we also explore the relationship between this system and a related series system. Combining Spot Purchases with Contracts in a Two-Echelon Supply Chain Shiming Deng, Candace Arai Yano—University of California, Berkeley We develop a model for a supply chain with a single supplier and single buyer that incorporates the possibility of a second purchase at a “spot” price, as a supplement to a fixed-price contract. The supply arrangement consists of two parts: (1) a contract purchase of the component at a price determined by the supplier and a quantity decided by the buyer in advance of observing demand, and (2) a potential spot procurement of additional units (to be determined by the buyer) at a negotiated “spot” price after demand is realized. Because the component has a long production lead time, the supplier can make only a single production run that must commence far before demand is observed by the buyer. We formulate the problem as a four-stage game and characterize the equilibrium wholesale price, supplier’s production quantity and buyer’s order quantities. We also discuss the impact of the “spot” purchase option on supply chain coordination. Optimal Capacity-Allocation Mechanisms in Supply Chains Under Asymmetric Information: An Auctions Approach V. Deshpande, Leroy B. Schwarz—Krannert School of Management, Purdue University We consider a supply chain in which a single supplier with fixed capacity sells to several independent retailers. The retailers have private information about their individual markets (e.g. mean market demand), which influences the size of their orders to the supplier. If the sum of all retailer orders exceeds the supplier’s capacity, the supplier uses a pre-declared rule, which maps retailer’s orders to allocations. A broad class of allocation mechanisms are prone to manipulation by retailers, as shown by Cachon and Lariviere (1999). In this research we consider the following question: What is the optimal capacity-allocation rule and pricing mechanism for the supplier? We use a mechanism-design approach to solve this problem. First derive the optimal allocation rule under general assumptions about the retailer profit functions and the nature of information asymmetry. Second, we derive an incentive- compatible pricing mechanism to implement this allocation rule. We also determine the optimal capacity for the supplier conditional on the optimal allocation rule. We show that an incentive-compatible implementation of the relaxed linear-allocation mechanism is optimal for the supplier when retailers face linear demand, and the intercept of the demand curve has a uniform distribution. We also provide conditions for deriving the optimal allocation rule when retailers face a newsvendor problem. Finally, in order to implement the optimal allocation rule, we design an auction mechanism wherein retailers bid for supplier capacity. On the Interaction of Production and Financial Hedging Decisions in Global Markets Qing Ding, Panos Kouvelis—Washington University We study the interaction of operational and financial hedging decisions of risk averse global firms in the presence of demand and exchange rate uncertainty. We consider a simple two stage stochastic programming model of a firm producing in its home country and selling to a foreign market. In the first stage a capacity/production plan and a financial hedging contract are decided in the presence of demand and exchange rate uncertainty. In the second stage, after the realization of demand and exchange rate, a production allocation decision (e.g., how many units to localize and distribute to the foreign market) is made in order to optimize profits. The second stage allocation decision (referred to as an allocation option) is the firm's real option serving as an operational hedge of the demand and exchange rate uncertainty. We want to understand the role of the allocation option in relation to financial hedging policies. We provide optimal hedging policies for the general stochastic demand problem, but stronger results are obtained for the "independent case" (i.e. demand and exchange rate are independent random variables). Our analysis, clearly establishes the value of the operational hedge (allocation option) for risk averse firms. Markets for Surplus Components with a Strategic Supplier: Implications for Channel Profits Lingxiu Dong, Erik Durbin We study markets for surplus components, which allow manufacturers with excess component inventory to sell to firms with a shortage. Recent developments in internet commerce have the potential to greatly increase the efficiency of such markets. We consider a general supply network where one monopolist component supplier sells to a large number of manufacturers who use the component to produce and sell final goods and ask how trade in surplus components affects the profits of firms in the 12 channel. Three scenarios are studied, namely no surplus market, a frictionless surplus market, and a surplus market with costly trading. The key findings are that: A surplus market will not necessarily improve efficiency of the supply chain. We characterize conditions under which a surplus market will improve supplier profits, manufacturer profits, and overall efficiency of the channel. Increased costs of transacting on the surplus market may benefit manufacturers who trade on the market, because of the impact of these costs on the component supplier's pricing power. Advanced Demand Information vs. Outsourcing Option in Production-Inventory Model with Uncertain Capacity Izak Duenyas, Xinxin Hu, Roman Kapuscinski—University of Michigan Business School To control a production-inventory system efficiently, the manager has to consider jointly the effects of stochastic demand and uncertainty in the production process. Both uncertainties hurt the system and by alleviating either of them, the reliability of the system can be improved. Types of investments that we see in practice include both investments in information system geared towards better demand forecasting, as well as various forms of contracts that assure additional capacity. This motivates us to jointly analyze advanced-demand information and outsourcing option in a finite-horizon dynamic-inventory model with backordering, in order to explore the characteristics of the resulting optimal policies and to evaluate their performance. In the paper we analyze different versions of the model, which correspond to the different features of advanced-demand information and outsourcing options, including extensions to Markov-Modulated demand and different structures of lead times. We prove that the optimal policy is a state-dependent double-threshold policy, with a lower outsourcing threshold and a higher production threshold. We explore the sensitivity of the optimal costs and of thresholds to unreliable capacity, and show that stochastically larger capacity decreases both the optimal cost and the value of thresholds, and that stochastically more variable capacity increases optimal cost, while it does not determinate of the relative ordering of thresholds. Through a numerical study, we evaluate the individual and joint benefits of advanced-demand information and outsourcing. We show that in some situation, the joint value of both options employed is significantly bigger than sum of individual benefits. We also find that the system parameters, demand, capacity variance, utilization, and service level, have exactly the opposite effects on the individual value of information and on the value of outsourcing. Among them, the effects of demand variance are most complicated. Optimal Markdown Mechanisms in the Presence of Rational Customers with Multi-unit Demands W. Elmaghraby A. Gulcu—Georgia Institute of Technology P. Keskinocak The Internet provides tremendous opportunities for implementing dynamic pricing mechanisms, since it is easier to collect information about markets and customers, and to change prices electronically rather than physically. In this paper, we analyze the optimal design of a markdown pricing mechanism, which is a form of dynamic pricing. Our focus is the structure of the optimal markdown mechanisms in the presence of rational or "strategic" buyers. While markdown pricing mechanisms have been in use long before the advent of the Internet, they have received only a brief glance from both the economic and operations management literature. In particular, many fundamental design and associated strategic behavior dimensions have not yet been addressed: At which price levels will a buyer bid (and how much will he bid) given a particular markdown strategy? Will the buyer submit all of his demand at a single price level or at multiple price levels? What is the optimal number of price levels? Do the optimal number and value of prices depend on the distribution of buyers' valuations? Under what conditions would the seller be better off implementing a single take-it-or-leave-it price vs. markdown pricing? With this paper we shed light to these fundamental questions. We first focus on a markdown mechanism under complete information in which only two customers participate. For this base case, we characterize the optimal behavior of the buyers and then we establish the optimal markdown mechanism for the seller. Later, we leave the complete information assumption aside and repeat the analysis with a setting where the seller does not know the demands of each customer type. For both settings, we compare the seller's revenues resulting from the optimal markdown prices and the optimal single price. Capturing Supplier Cost Reductions in Buyer-Supplier Relationships Wedad Elmaghraby, Se-Kyoung Oh—ISYE, Georgia Institute of Technology In a repeat-purchase environment, the monitoring of incumbent supplier(s)’ costs becomes critical. This is particularly true when supplier unit costs decline with accumulated output or supplier investments. When a supplier’s costs decrease over time, it may be advantageous for the buyer to keep an incumbent supplier, in the hopes of reducing her procurement costs. However, this will only occur if the supplier is forced to pass along a portion of his cost savings to the buyer. It is common practice in industry for buyers to set critical performance(cost) bar (CPB) for their suppliers. Under a CPB mechanism, the buyer offers subsequently lower prices to her incumbent suppliers. If the supplier(s) accepts the lower price, then he is retained; however, if he rejects the lower prices, the buyer will drop the supplier(s) and select a new one. In this paper, we investigate the performance of a CPB mechanism when compared to common alternative procurement mechanisms. In particular, we study the performance of a CPB mechanism when suppliers experience cost reductions as a result of (i) learning by doing (experience) and (ii) investment in cost-reducing activities. We compare the performance of a CPB mechanism against a standard sequential second price procurement auction. Our goal is to understand when a CPB outperforms a sequential second price auction, i.e., a CPB results in a lower total procurement cost for the buyer than a standard auction. 13 Under the CPB mechanism, suppliers bid in the first period to be selected. In the second period, the buyer offers the incumbent suppliers a per-unit price of R to continue doing business with her. If a supplier accepts price R, then the buyer keeps him in the second period. But if there are any suppliers who do not accept this offer, the buyer conducts competitive bids among other potential suppliers for the unfilled orders. Enterprise-Wide Optimization of Total Landed Cost at a Grocery Retailer Feryal Erhun, Sridhar Tayur—Graduate School of Industrial Administration, Carnegie Mellon University We describe an internet-based enterprise-wide tool for tactical planning at a grocery retailer, which we call the Total Cost Solution Application. The tool enables a ―total landed cost‖ perspective by coordinating decisions across functions of the supply chain. It dynamically optimizes across logistics, purchasing, and warehouse management, while considering the necessary joint replenishment economies and accounting for a wide variety of complexities, such as discounts on total order quantity, intermittent demand, multiple warehouses, vendor deals and forward buys, and promotions. The application, we believe for the first time, combines appropriate operations research techniques to solve the problem; furthermore, it is created in the newly available information technology infrastructure. The entire application ―sits above‖ and coordinates existing execution tools such as Manugistics (for logistics), BICEPS (for purchasing), and Exceed and TRICEPS (for warehouse management) by setting key operating targets and providing operational guidance. The results were dramatic and quick. Within the first six weeks of implementation, we observed reduction in on-hand inventories, an increase in service levels, and substantial improvements in logistics decisions. The total landed cost is substantially lower – with an improvement of $3.1 million, or 0.24% of COGS and 3.5% of net profits – while providing superior fulfillment to the stores. Analysis of An Assembler-Distributor Network Under Revenue Sharing and Wholesale Price Contracts Nesim Erkip, Ulas Ozen, Refik Gullu—Middle East Technical University We consider a decentralized assembler-distribution network with several suppliers. There is a single product with stochastic demand. The suppliers are of two kinds. There is one type-1 supplier that operates with a wholesale contract and there are n different type-2 suppliers, each having a revenue-sharing contract with the assembler. Each supplier type represents a certain market structure with respect to the assembler. Type-1 supplier, on the other-hand, is practically a monopoly (either monopoly or one of the firms in an oligopolistic market), powerful enough to set the wholesale price. Type-2 suppliers are follower type companies, which are usually clustered around the parent company (assembler), having limited resources, and usually their main customer is the assembler. The distributor is considered to be available as an option to increase the demand of the assembled item. We first analyze a system that consist of an assembler, one Type-1 supplier and a number of Type-2 suppliers where the assembler is supplied with complementary components by both suppliers, performs assembly an kitting operations and finally sells the product to the market. The demand is random. Same information on demand distribution is available to all. According to revenue sharing contract the assembler sets the revenue shares of the suppliers and the suppliers determine how much to deliver where they are paid only for the complete products sold. And for the wholesale price contract, the Type-1 supplier sets the wholesale price of its product and the assembler decides on the order quantity. We find optimal solutions and analyze system behavior under different cost and demand settings where each actor tries to maximize its expected profit. Then, we compare the findings with the centralized system. We observed that the relatively powerful Type-1 supplier and the assembler dominate the system. A set of incentive mechanism is proposed to get a solution that is closer to a coordinated system. In another setting, we extend our previous system with a distributor. Here, the assembler faces with an opportunity to enter a new market with a distributor. We specify a revenue sharing contract between the distributor and assembler. Using previous results, we analyze this system under correlated and independent market demands to provide managerial insight. High-Tech Capacity Reservation with Deductible Fees Murat Erkoc, S. David Wu—Lehigh University We study capacity reservation contracts in a high-tech manufacturing environment. Motivated by our work at a major telecommunications component manufacturer in the U.S., we consider contracts that allow the manufacturer (the supplier) to share the risk of capacity expansion with her OEM customers (the buyer). This is important as the capacity cost is enormous in this industry, while the market demand highly volatile. Without a risk sharing mechanism, the supplier will adopt an overly conservative expansion policy, which leads to profit loss for the channel. We first analyze a one-supplier, one-buyer system with stochastic demand, and then generalize the results to the one-supplier, two-buyer case. The reservation contract works as follows: the supplier announces a per unit capacity reservation fee that can be partially deducted from the total payment when the capacity is later utilized. In turn, the buyer places reservation on a specified amount. Given the reservation, the supplier makes decisions to expand capacity. We first establish the benefit of the capacity reservation contract from the viewpoint of channel efficiency. We then identify the conditions under which the supplier would voluntarily comply with the contract. We propose two forms of channel coordination: First, a contract where a portion of the reservation fee is deductible. Second, a contract where the reservation fee is proportional to the capacity cost. We prove that the latter can be reduced to a profit sharing contract. We first generalize the analysis to the two-buyer case where the buyers are from the same market with fixed split. We show that the capacity reservation problem in this case can be reduced to the one-buyer case. We then consider the case where the buyers compete in the market and analyze the impact of competition on the reservation contracts. Inventory Decisions Considering Product Quality and Inter-Company Competition 14 Mark Ferguson—Georgia Institute of Technology Oded Koenigsberg—Duke University Firms often make production or purchasing decisions without knowledge of the exact market size of their product. When demand turns out to be lower than the chosen product quantity, the firm is left with unsold inventory that it may choose to carry over into the next period. Given a perishable product whose quality level decreases over time, how does a firm make pricing and inventory decisions for this unsold inventory. This paper addresses this issue and finds conditions under which the firm is better off scrapping either all or part of its leftover inventory in order to avoid competition with any new product obtained in the future periods. Deciphering Demand Signatures Vishal Gaur—Stern School of Business, NYU Avi Giloni—Yeshiva University, New York Sridhar Seshadri, Stern School of Business, NYU We consider a two-stage supply chain where multiple retailers order from a single supplier. The frequency of ordering varies across retailers due to differences in their costs. The supplier may only see the aggregation of the orders from the retailers and would like to forecast the future order stream using historical data in order to plan its inventory. We present a forecasting method based on the ordering behavior and frequency domain time series analysis. We use several datasets from the industry and generated through simulation to benchmark our method with the existing time-series forecasting methods. This problem is motivated by a candy manufacturer that supplies candy to a large video rental chain. Each video store orders candy separately, but the orders are aggregated at the corporate office of the company and sent to the candy manufacturer. The naïve method of forecasting in this scenario is to estimate the mean and standard deviation of the aggregate demand from the historical data assuming stationarity, and use it to plan inventory. Other methods for forecasting time-series data that may be used include ARIMA, seasonal ARIMA, exponential smoothing, etc. We show that these methods are inadequate for forecasting the aggregated order series because such data violate the assumption of stationarity. Our paper gives a forecasting technique that is based on the ordering behavior induced by inventory models. It may be applied in a multi-stage supply chain to forecast future orders and reduce the incidence of bullwhip effect. It also contributes to the theory of time-series analysis by analyzing non-stationary time-series data formed by the aggregation of multiple periodicities. Our analysis generalizes to the situation when each individual time-series has a different trend component. The technique could also be useful when demand is intermittent. A Consignment System where Supplier Cannot Verify Retailer's Sales Reports Yigal Gerchak, Eugene Khmelnitsky—Tel-Aviv University Newspapers are often sold through stores via consignment arrangements, which involve vendor (publisher) managed inventory and revenue sharing. Since retailers are not required to actually return unsold copies, it is said that some of them occasionally under-report sales. That hurts the publisher on the short run and also interfere with his rational stocking decisions as these are based, to some extent, on previous sales reports. We construct a discounted dynamic framework for the retailer's optimal reporting as a function of the publisher's delivery-response function to these reports. It turns out the optimal report does not depend on actual sales. We then show that the publisher's resulting delivery response function is the same as it would be in an integrated system. Thus the retailer's untruthful behavior actually causes the system to behave optimally. Had the retailer been knowingly truthful, the system would not be coordinated. Inventory Policy Implications of Customer Purchase Behavior in the Face of Stockouts Harry Groenevelt—Simon School of Business, University of Rochester Pranab Majumder—Fuqua School of Business, Duke University The availability of extensive customer purchase behavior data in modern forms of retailing (such as TV shopping channels and on-line retailers) has implications for inventory management. In this paper we first propose and analyze a simple EOQ based model that incorporates customer behavior in the face stockouts. Customers are assumed to maximize their utility, which decreases with the amount of time they have to wait until the item becomes available. This results in a fraction of the demand during a stockout being lost, and this fraction increases with the length of a stockout. In addition, we discuss various ways in which the basic model can be implemented in practice, as well as an extension to incorporate constraints on the number of products that can be displayed on web pages or shown on shopping channels. Parallel Experimentation, Manufacturing Competency and Market Structure Korhan Gürkan, Haim Mendelson—Graduate School of Business, Stanford University In this paper, we study the effects of recent trends in product development and competency and manufacturing competency and their effect on market structure. Firms in an industry compete in designing and manufacturing products. Modeling product design as a parallel search for the best alternative among experiments with probabilistic outcomes, we advance 15 a game-theoretic model of product development, using elements of extreme value theory and the theory of industrial organization. The resulting equilibrium market structure can have either of two distinct forms. In the traditional one-shot market structure, each firm experiments once, and the number of firms in the industry is determined largely by market attributes such as demand elasticity, marginal production cost, and the cost of experimentation. In the innovation-intensive market structure, firms make multiple experiments, and industry structure is substantially affected by the shape parameter of the distribution of R&D outcomes. For distributions with heavy right-hand tails, the industry has fewer firms that conduct a larger number of experiments, which results in higher-quality products and larger R&D spending. As experimentation costs decline, the prevailing market structure changes. Traditional industries shift toward an innovation-intensive structure, and in cases where there was no industry before, new ones emerge. Furthermore, industries with R&D processes that have heavy right-hand tails go through market consolidation as a result of declining experimentation costs, whereas all other industries become more fragmented. In industries with heavy right-hand tails, it becomes more important for firms to differentiate themselves by developing innovative competencies, whereas, in other industries, superior manufacturing becomes more crucial. Economic Lot Scheduling with Non-Identical Resources in Parallel Cagri Haksoz, Michael Pinedo—Stern School of Business, New York University In this paper we consider the economic lot scheduling problem with m machines (or facilities) in parallel. There are n different types of items. Item j has a demand dj per unit time, a holding cost hj per unit time, and a setup cost cj. The m machines have different speeds v1 < v2 < … < vm. The speed of the slowest machine v1 = 1. Machine i can produce item j at rate vipj. We consider three different models. The objective in all three models is to find an assignment of items to machines that minimizes the total cost per unit time. In the first model each machine operates according to a rotation or cyclical schedule and the cycle times of the rotation schedules of the m machines have to be the same. In the second model each machine again operates according to a rotation schedule, but the rotation schedules are allowed to have different cycle times. In the third model the machines do not have to operate according to rotation schedules; each item on each machine may have its own cycle time. For each model we consider a number of special cases that provide insights into the role each parameter plays. Based on the results of the special cases we formulate for each model a heuristic that can be applied to arbitrary instances. In the concluding remarks we discuss the significance of our results for problems that occur in practice. Joint Price and Capacity Choice Under Monopoly Joseph M. Hall We study optimal price and capacity choices in settings where either congestion effects or stockouts may result in service failures and accompanying revenue loss. Congestion effects are studied via loss-type queueing systems and stockouts are studied via newsvendor models. We develop and study a static model in which price and capacity decisions affect only current period revenues and costs. Within the context of this model, we show that the optimal price and capacity choices take on simple forms: optimal capacity is proportional to demand and optimal price follows from a traditional, risk-free monopolistic formulation with an augmented capacity cost. We extend the static model to include the effects of past price and capacity decisions on current period demand. Two numerical examples are studied in the context of these models: retail delivery of a physical good from inventory and delivery of a service via a system that involves physical queueing of customers. Optimizing Supply Flows Using Advance Demand Information Samuel Holler, Fikri Karaesmen, Yves Dallery—Laboratoire Productique Logistique, Ecole Centrale Paris Motivated by manufacturer-supplier relationships in the automotive industry, we investigate the supply _ow optimization problem with advance demand information. In the auto-industry context, the supplier delivers parts to the manufacturer's plant based on information about the manufacturer's assembly schedule. In particular, the manufacturer determines the assembly sequence from end-customers' orders several days in advance, transmits this information to the supplier and requires timely delivery to the assembly line. Within this interval, the assembly sequence is subject to frequent modifcations that cause perturbations of initially announced demand lead times. Our goal is to analyze the impacts of demand visibility and randomness on the supply performance. In order to gain insights into the effects of demand lead time randomness, we analyze a single-stage supply model in discrete time and with unit demand arrivals. The supply lead time, LS, is constant but the demand lead time is composed of two components : the constant nominal demand lead time, ±, and a random variable, X, which can take positive and negative values. The control policy is a two-parameter mechanism which combines the standard base-stock logic (with base-stock level S) with a release timing parameter, L, whose role is to regulate replenishment of parts using advance demand information. To analyze the system with random demand lead times, we exploit a duality property between demand lead time and supply lead time variability. Using this property, we determine the release lead time parameter L and the base stock level S that minimize inventory related costs. Based on this analysis, we present results on the impact of several key factors, such as visibility and demand lead time perturbations, on system performance. Finally, we discuss the relative merits using safety stocks versus safety lead times in the presence of demand lead time randomness. An Agency Model of Buyer-Supplier Design/Production Effort Ananth Iyer, Leroy Schwarz—Purdue University Stefanos Zenios—Stanford University 16 We consider a principal agent model of buyer–supplier effort in the product and process design and production of product when supplier capability is private information. We focus on the difference between black box designs and glass box designs which differ in which the extent of the buyer influences the process of production. The model focuses on role of alternate mechanisms. We solve for the optimal supplier and buyer effort in the presence of a buyer resource constraint under all mechanisms. Managerial implications of the model are explored. The Impact of Sharing Current Lead-Time Information in Decentralized Supply-Chains Apurva Jain, Kamran Moinzadeh—University of Washington Recent explosive growth in supply chain software applications and communication technologies has enabled easy exchange of information across businesses. While the value of information flowing upstream has been the subject of much study recently, we model the impact of a relatively new practice: information flowing downstream from manufacturer/warehouse to retailer. This has been enabled by software such as J D Edwards’ customer relationship management system, that allows up- stream manufacturers or distribution warehouses to authorize a retailer’s access to the real-time information about the congestion levels or the inventory availability. We study the impact of such downstream information exchange on the members of a supply- chain. We model a one warehouse – multiple retailers supply chain in continuous time. The retailers experience Poisson demands and place replenishment orders at the warehouse. Both echelons backorder unsatisfied demands, have fixed lead-times and incur holding and backorder costs. To begin with, our focus is on the warehouse and one retailer while other retailers are considered exogenous to the system. In the base model, there is no information sharing between the supplier and the retailer. Both, the retailer and the warehouse make locally optimal decisions about their ordering policies and we develop the equilibrium solution. The next model allows the retailer to know if there is any on-hand inventory available at the warehouse at the time of placing an order. Retailer’s optimal ordering decision incorporates this information. We characterize the actual information available to both parties as well as their beliefs about the incomplete information. We then develop an analysis for determining the optimal ordering decisions for both parties and for exactly evaluating their costs. We also characterize the equilibrium solution with downstream information sharing. The comparison between the above two models shows that such downstream information sharing may increase the warehouse costs. We also show that the incomplete information may also lead to retailer seeing an increase in its cost. We propose a simple contract to replicate the supply chain optimal solution in such situations. Managerial implications of our results are discussed. This model is extended in several ways: the retailer is allowed to observe the value of warehouse net inventory rather than just availability; impact on other retailers is considered when information is shared only with selected retailer; and finally, all other retailers are also given access to the up-stream information. We also model the up-stream business as a manufacturer’s queue and analyze the impact of the retailer knowing the number-in-manufacturer’s system at the time of ordering. Inventory Allocation and Order Fulfillment with Future Demand Information M. Eric Johnson—Tuck School of Business, Dartmouth College Sharing future demand information is a key component of collaborative supply chain relationships. In this paper, we examine the order fulfillment process in an inventory system where customers specify future demand information with a desired delivery time. In this multiclass system, customers who place orders to be delivered at a later time, expect the products to be shipped when they were requested – not late and not early! We develop an exact expression for the distribution of order response time and the probability of fulfilling the order by the customer request date. Using our model, we compare order fulfillment performance and inventory investment under different product allocation polices. We develop jointly optimal allocation and inventory policies for a system that serves multiple customer types. Scheduling Multiple Types of Time-Shared Aircraft With Crew Duty-Time Constraints Itir Karaesmen, Pinar Keskinocak, Sridhar Tayur, Wei Yang—Carnegie Mellon University We look at a real life application where multiple types of time-shared aircraft have to be scheduled to satisfy customers’ flight requests. Any request that cannot be met by company aircraft is satisfied using charters. The objective is to minimize charter hours and to maximize revenue generating flight hours for the company while allowing for upgrade /downgrade of customers among multiple types of aircraft. Given a scheduling horizon, aircraft may have scheduled maintenance as well as crew duty-time constraints. Each aircraft has a crew that has daily minimum rest-time and maximum duty-time requirements. The resulting schedule has to satisfy these constraints. We first analyze the problem without the crew constraints. We present different mathematical formulations of the problem and discuss their computational performance for real life instances. We show these models can be used to deal with very special cases of crew duty-time constraints. For general cases, we discuss how the crew constraints though not guaranteed to be satisfied 100% of the time can be incorporated into these models. Next, we present formulations for the problem with crew duty-time constraints. We prove that the problem with hard crew-time constraints is NP complete. We propose a heuristic to solve large instances of this problem. We provide results of an extensive numerical experiment to test the performance of the heuristic. The Multiple-Family ELSP with Safety Stocks 17 Serge M. Karalli, A. Dale Flowers—Case Western Reserve University The ELSP with normally distributed, time-stationary demand is considered in a manufacturing setting where the relevant costs include family setup costs, item setup costs, and inventory holding costs for both cycle and safety stocks. A family is a subset of the items that share a common family setup with its associated setup cost and setup time. Each item within the family may have its own setup time and setup cost. The families form a partition of the set of items manufactured on a single facility. The safety stock level for any item is a function of the time interval between production runs for the item, the service level specified, and the variance of its demand. The Multiple-Family ELSP with safety stocks differs from multi-level inventory models with family setups in that the former assumes non-instantaneous inventory replenishment and considers the cost of holding safety stocks; the latter assumes instantaneous replenishment and does not directly assess the impact of safety stock levels on the total cost. The solution to the mathematical model is comprised of the basic period length, the family multipliers, and the item multipliers that give the lowest total cost of setups and carrying inventory. The family multipliers and items multipliers are restricted to integer powers of two. An efficient solution procedure is developed for this model. Properties of the non-convex feasible space are identified and used in the solution approach. The traditional economic run formula will only attain a sub-optimal solution. For a truly optimal solution, the product volume and product mix decisions must be coordinated with the economic run decisions to ensure that profit is maximized overall. This paper offers a superior solution to the problem of product lot sizing by focusing on the bottom line (profit) rather than on the middle line (cost). Dynamic Portfolio Selection of New Products under Uncertain Market Conditions Stylianos (Stelios) Kavadias, Christoph H. Loch Selecting program portfolios is an important challenge in the management of new product development (NPD). At the aggregate level of business management a key decision is the allocation of resources across product lines or market segments. This article develops a dynamic model of resource allocation, taking into account multiple interacting factors, such as uncertain market payoffs that change over time, increasing or decreasing returns from the NPD investment, carry-over of the investment benefit over multiple periods, and interactions across segments. We characterize optimal policies in closed form and derive qualitative decision rules for managers. In the presence of increasing returns, the whole budget is optimally allocated to one product line, while decreasing returns lead to a split of the budget. The optimal allocation properties are subtle and partly counter-intuitive. For example, neither the longevity of the product line, nor market size in future periods always increase the investment. For a risk-averse decision maker, a higher variance in next period's market potential makes a product line less attractive, but a higher variance in a future period may increase the optimal allocation. If the product lines interact, the complement/substitution effect acts as an additional/ reduced carry-over benefit. Pricing, Production, And Distribution Planning Under Exchange-Rate Uncertainty Burak Kazaz, Haresh Gurnani—University of Miami Global companies face the challenge of coordinating production and allocation decisions with pricing strategies in a fluctuating exchange-rate environment. The natural sequence of decisions requires that production plans be made prior to the selling season. Later, closer to the selling season, when the exchange rate is realized the company adjusts its prices influencing the demand. We consider the case of two production facilities – the first facility is home-based (referred to as the domestic plant) and is not directly exposed to exchange rate fluctuations; the second facility is located in the foreign market (referred to as the foreign plant). Both production facilities supply to meet demand in a foreign market. Our study outlines a mathematical model that provides businesses in choosing the optimal production, allocation, and pricing decisions. Our model is a two-stage stochastic program where production decisions are made at the first stage; based on the realized exchange rate, pricing and allocation decisions are simultaneously made in the second stage. It should be noted here that demand is impacted by the second stage prices, and production decisions of the first stage are made without knowing the exact value of demand. In the analysis of the model, we start at the second stage where for a realized exchange rate and given production quantities at the two facilities, the optimal distribution policy is determined along with the optimal selling price. Using the optimal policies at the second stage and taking expectation with respect to the exchange rate, we determine the optimal production quantities at the first stage. Conditions are derived when it would be optimal to supply the foreign market using domestic or foreign plant only. In an extension to the model, we plan to consider the impact of an additional domestic market on the optimal policies. New Product Introduction: Timing, Design, and Pricing Ted Klastorin, Weiyu Tsai—Department of Management Science, University of Washington Short product life cycles, diverse customer preferences, and rapidly changing technologies have created significant challenges for managers who develop and introduce new products and services. Among other factors, these managers must consider the timing of the introduction of the new product, the design of the product (in terms of features and quality), as well as the price. The tradeoffs among these decisions are further complicated when there is more than one firm entering a market. In this paper, we consider the case when two profit-maximizing firms enter a new market with a competing product that has a finite (and known) life cycle. We assume that both firms have similar product development capabilities and compete on 18 the basis of product design, timing, and price. Product development time is an increasing function of the product’s complexity that is measured by the number of features and the performance level of these features. We also assume that the customers’ willingness to purchase a product in this market is a function of the price of the product, the design of the product, and the customers’ willingness to adopt a new innovation. The latter factor reflects customers’ general reluctance to adopt new technologies as well as changing demands that may result from the impact of network externalities. We consider two cases in this paper. In the first case, we assume that both firms make design decisions simultaneously without information about the other firm’s decisions. The firm that designs its product with the fewer number of features enters the market first; at that time, the ―leader‖ firm sets a price for its product and enjoys a monopoly situation until the second firm (the ―follower‖) enters the market. When the second firm enters the market, both firms simultaneously set (or reset) their product prices knowing the design of both products at that time (which we assume are fixed for the remainder of the product’s life). In the second case, we assume that one firm is a leader who independently begins development of the product. At some later point in time, the second firm (the follower) becomes aware of this development effort and begins to develop its own product. At this point, the follower can choose to crash his product development time by allocating additional resources to this effort. Again, firms set (or reset) prices when either firm enters the market. We develop two models that represent these new product introduction processes. Using results from these models, we show that each firm would like to differentiate its product from its rival’s (which may not occur in the first case). Specific decisions of each firm, however, depend on the distribution of customer preferences, the innovation diffusion rate and the length of product life cycle. A numerical example illustrates our models and resulting managerial implications. Competitive Location and Capacity Decisions for Facilities Serving Time Sensitive Customers Anthony Kwasnica, Euthemia Stavrulaki—Penn State University Locating facilities around a customer base has received considerable attention in both the Economics and Management Science literature. The Economics literature has explored equilibrium location strategies in a competitive environment, and the Management Science literature has studied location strategies that minimize travel times and/or transportation costs in a monopolistic environment. The novelty of our paper is that it blends the Economics and Management Science research streams to investigate a competitive location problem that also considers travel times, capacity setting, and time sensitive customers. Our work aims to provide insights with respect to interactions between competitive location, capacity and delivery performance. Based on Hotelling’s linear city, we consider two firms offering a single product to time-sensitive customers who are located uniformly on a line segment. Consumers send their orders to a distribution facility, then the firm processes (customizes) these orders on a first-come, first-served queue, and subsequently ships them to customers. Thus, consumers incur delays due to both shipping leadtimes and queuing congestion within the facility, and decide to purchase from the firm that offers them the highest (non-negative) expected surplus. Given a fixed price and a fixed interval of customers, we explicitly model consumers’ waiting costs in the profit- maximizing objective function of each firm. We identify subgame perfect equilibria when firms first choose capacities and then locations. We prove that depending on the problem parameters, three types of equilibria are possible; local monopoly (in which each customer is served by a single firm, and some customers are left unserved), constrained local monopoly (in which firms are generally located at different points but serve the entire interval of consumers), and duopoly (in which firms are generally located at different points and compete for some consumers). We also provide comparative statics results using numerical examples. Principal-Agent Problems When Choosing Scale for Operations Phillip J. Lederer, Tushar Mehta—University of Rochester In many real operations settings, managers must choose the production technology and the scale of output using the technology. This paper shows that traditional NPV analysis where a "hurdle" rate is specified to the decision making manager generates suboptimal technology and scale choices. Further, if the discount rate specified by management is chosen to induce an optimal scale choice, the manager will systematically undervalue the project. The paper show that if the hurdle rate is chosen to reflect the risk of the optimally designed project, the manager will make suboptimal decisions but we show that the manager's design error and its underestimation of the project is generally small. Scheduling Parallel Machines that are Subject to Breakdown and Repair Joseph Y-T. Leung—Department of Information Systems, New Jersey Institute of Technology Michael Pinedo—Stern School of Business, New York University We consider a number of machines in parallel and n jobs. The machines are identical and subject to breakdown and repair. The number may therefore vary over time and is at time t equal to m(t). Preemptions are allowed. We consider three objectives, namely the total completion time Cj, the maximum lateness Lmax and the makespan Cmax. We study the conditions on m(t) under which various rules minimize the objective functions under consideration. We also analyze the cases when the jobs have deadlines to meet and when the jobs are subject to precedence constraints. The rules that are optimal are variations of the classical rules, such as the Shortest Processing Time first (SPT), the Longest Remaining Processing Time first (LRPT), and the Earliest Due Date (EDD). Performance Analysis of Production Systems with Rework Loops Jingshan Li—Enterprise Systems Lab, GM Research & Development Center 19 This paper is devoted to the study of production systems with rework loops. Such systems are often encountered in manufacturing plants and accurate performance analysis is needed for design and continuous improvement. However, no analytical methods to evaluate its production rate are available in the current literature. In this paper, we present an aggregation procedure using overlapping decomposition approach to approximate the production rate for systems with rework loops. The idea is to decompose the system into four serial production lines, with machines modified to accommodate the interactions with other machines and buffers. A recursive procedure which analyzes serial lines during each iteration is presented. The convergence of the iterative procedure and the uniqueness of the solution are justified analytically. The accuracy of the estimate is evaluated numerically with good results. Selling a Fixed Amount of Inventory: Why More Inventories May Be Sold at a Higher Price Qing Li, Hongtao Zhang—Hong Kong University of Science and Technology A retailer faces the problem of selling a given stock of stems in one time period. Demand is price sensitive and random. Expected demand is downward sloping. Will it always be optimal for the retailer to choose a lower price if there are more initial inventories? This paper shows that the answer is no and explains why. Switch-over Policies for the Stochastic Knapsack Grace Y. Lin, Yingdong Lu—IBM T.J. Watson Research Center David D. Yao—IEOR Department, Columbia University The stochastic knapsack used to be a model for studying access control in telecommunication networks. Recently it has found new applications in diverse areas such as revenue management and dynamic pricing. We first formulate the stochastic knapsack as a dynamic programming problem, and establish several key properties of the value function, such as concavity and submodularity. These properties point to a lower-orthant structure of the optimal policy. Beyond this structure, however, the optimal policy in general can only be derived in the form of a decision table, and is hence difficult to implement. Based on the lower-orthant structure, we focus on a class of so-called "switch-over" policies. For instance, a policy can be characterized by a set of time epochs, the switch-over points, with the control switching to a different action at each epoch. The switch-over points are determined over the time horizon so as to optimize the objective function. A more detailed, two- dimensional switch-over policy involves switch-over points in terms of the available resource, as well the switch-over times. We characterize the dynamics under such switch-over policies by Markov processes, derive the corresponding objective functions, and solve the resulting optimization problems. We also demonstrate the performance of the switch-over policies through comparisons against the optimal policies. Revenue Management for Systems with Shared Resources: Pricing, Capacity Sizing, and Service Differentiation Constantinos Maglaras, Assaf Zeevi—Graduate School of Business, Columbia University This talk considers revenue management for service systems that are characterized by three salient features: capacity is large, processing resources are shared between the users, and demand is elastic. These features are intrinsic in modern information and communication systems. The service provider is assumed to operate a finite set of processing resources that can be shared among users, however, this shared mode of operation results in a service rate degradation. This mode of service provision is dubbed ―best effort.‖ Users, in turn, are sensitive to the delay implied by the potential degradation in service rate, and to the usage fee charged for accessing the system. We study the equilibrium behavior of such systems in the specific context of pricing, capacity sizing and service level differentiation, under revenue and social optimization objectives. Exact solutions to these problems can only be obtained via exhaustive simulations. In contrast, we pursue approximate solutions that exploit large capacity asymptotics. Economic considerations and natural scaling relations demonstrate that the optimal operational mode for the system is close to ―heavy traffic.‖ This, in turn, supports the derivation of simple approximate solutions to economic optimization problems, via asymptotic methods that completely alleviate the need for simulation. These approximations seem to be extremely accurate. The main insights that are gleaned from the analysis are the following: Congestion costs are ―small‖, the optimal price admits an intuitively appealing decomposition, and the joint capacity sizing and pricing problem decouples and admits simple analytical solutions that are asymptotically optimal. In addition, the problem of optimal design of a differentiated service menu offering ―guaranteed‖ and ―best effort‖ services, the former involves allocating dedicated capacity, admits a simple and illuminating solution. All of the above phenomena are intimately related to statistical economies of scale that are an intrinsic part of these systems. Product Life Cycle Considerations in Remanufacturing Competition Pranab Majumder—Fuqua School of Business, Duke University Harry Groenevelt—Simon School of Business, University of Rochester We study the problem of an OEM considering life-cycle design issues for a remanufacturable product in the context of a diffusion model of product adoption. The OEM faces competition from small remanufacturers, has to pay for the returns and does not recover all the remanufacturable items from the customers. 20 We establish the existence of a sub-game perfect Nash Equilibrium, and show that the remanufacturing game is myopic. This allows us to study the strategic decisions by the OEM with respect to parameters representing remanufacturability, remanufacturing costs and competition. Optimal Stocking in Reparable Parts Networks with Repair Capacity and Inventory Pooling John A. Muckstadt, Peter L. Jackson—School of Operations Research and Industrial Engineering, Cornell University Kathryn E. Caggiano, James A. Rappold—School of Business, University of Wisconsin This paper addresses a tactical planning problem for a three-echelon reparable-parts service network characterized by a central repair facility and local opportunities for inventory pooling. The objective of the paper is to develop a model that can be used to find the optimal total system stock and to set target base stocking levels for each location in the system. The model we propose can be solved in time that is linear in the number of part number- location combinations, for large networks. To achieve this linear time performance, several approximations are employed. One implication of the approach used in this paper is to emphasize the importance of highly effective stock allocation routines in inventory management execution systems. Intertemporal Pricing in Retail and Services Serguei Netessine—The Wharton School, University of Pennsylvania One of the challenges that companies practicing price discrimination face is customer segmentation. While in some situations customers can be segmented based on some exogenous category (for example, travel preferences), often segmentation has to be based on endogenous parameter such as time of purchase (i.e., the product is priced differently throughout the sales period and company controls pricing policy). In practice, due to high administrative costs associated with price changes, there is a limit on the number of times the price can be adjusted. If that is the case, the company must select, in addition to prices, optimal points of time to switch from one price to another. These decisions are the focus of this paper. We analyze price discrimination problem with endogenized customer segmentation in a dynamic, deterministic environment where customer arrival rate depends on current price and time, limited number of price changes is allowed and there is a capacity constraint that is either exogenously given or set optimally. In our model, arrival rate can evolve in time arbitrarily allowing us to model situations where price decrease or increase is optimal. It is demonstrated that, while pricing decision is well behaved, timing of price changes is much more ambiguous, i.e., multiple local minima and maxima may exist. Analytical insights obtained directly from the solution as well as from specific examples provide intuition behind the optimal pricing and timing decisions. We obtain simple condition that guarantees optimality of low-to-high and high-to-low pricing. Further, we demonstrate that charging a single price instead of price discriminating is not only sub-optimal but may actually result in the worst possible performance. Day-of-Week Pricing for Express Package Delivery Services Alexandra Newman—Division of Business and Economics, Colorado School of Mines Candace A. Yano—IEOR Dept/Haas School of Business, UC Berkeley Most package delivery services charge a premium for faster delivery, but the practice of pricing by day of week is very limited. In the absence of this type of price differentiation, shipment volumes exhibit strong day-of-week patterns, especially in the express package delivery market. Although vehicle schedules may be adjusted to account for this day-of-week seasonality, it is rarely possible to exactly match shipping capacity to demand. Consequently, the excess shipping capacity varies by day of week and by route. When negotiating with potential high- volume contract customers, it may be advantageous to offer the customer an incentive to make shipments counter-cyclical to the overall demand pattern. Such a counter-cyclical shipping pattern would improve profit in two ways. First, revenue is generated using available excess capacity for which the incremental operating costs are quite small. Second, by smoothing the overall demand pattern, requirements for additional transport capacity are minimized. In this paper, we focus on the flow of a class of similar packages from a single shipper (typically a manufacturer) to a single consignee (a downstream user of the manufactured parts), as packages of different types or with different destinations can be priced differently. Even this simple version of the problem entails three levels of decision-making. First, the package delivery service sets prices by day of week and speed of service (or priority class) considering its own package handling and shipping constraints. Then, given the delivery service's price structure, the shipper (customer) decides which packages to ship each day via each priority class, considering its own processing constraints and the cost of holding packages in inventory (to take advantage of lower shipping costs on another day). Finally, the package delivery service must determine how to transport the induced demand while satisfying speed-of-service requirements. We formulate the problem as a Stackelberg game and present a solution approach that can provide guidance to a package delivery service in its negotiations with customers. Dynamic Quotation of Prices and Lead-Times Tava Olsen—Olin School of Business, Washington University in St. Louis We consider a retailer who has the ability to offer a menu of lead-times and prices to its customers. In other words, customers seeking a particular product will choose from a menu of prices and lead-times the lead-time that maximizes the value of their given lead-time utility minus the price for that lead-time. Lead-time sensitive customers will assign a high utility to short lead-times; whereas, customers with less urgency for the product will be will have a flatter utility and therefore be more price 21 sensitive. The retailer can change its menu dynamically; thus, the most applicable industry for this work is e-commerce. Dynamic control allows the retailer to charge higher prices when capacity is tight and lower prices when there is little backlog. We discuss how such dynamic menus should be chosen and present relevant structural results for this problem. We also numerically compare such dynamic control to static pricing. Quantity Flexibility Contracts with Renegotiation Erica L. Plambeck—Stanford Graduate School of Business Terry A. Taylor In the emerging industry of biologics (pharmaceuticals produced by microbial or mammalian cell culture), production capacity is expensive and inflexible; building a new plant takes at least four years and expansion of an existing plant may require FDA approval. Furthermore, the development and clinical testing of a new biopharmaceutical is an expensive and risky process that evolves over many years. Right now, production capacity is in such short supply that small biotech companies and pharmaceutical giants are choosing to scale back or terminate drug development projects. In this environment, the Swiss firm Lonza has been wildly successful as a pioneer in contract manufacturing. Today, biotechnology and pharmaceutical companies are contracting with Lonza for production that will occur three years from now. The contracts specify a minimum and maximum volume of fermentation capacity per unit time, and a unit price. With assured capacity, a biotech company will make ongoing investments in development, testing, and marketing which increase the likelihood of having a successful drug. Nevertheless, at the time of contracting, the biotech company is uncertain about the amount of production capacity that will ultimately be needed. A drug may fail in clinical trials, and even if it does succeed, the recommended dosage may vary by as much as a factor of ten from the initial target. Therefore, Lonza contracts with multiple buyers. When demand is realized, contracts are renegotiated to allow for an efficient allocation of capacity. A buyer with an unsuccessful drug may take less than the minimum contracted volume, whereas a buyer with a successful drug may exceeds his maximum; transfer payments ensure that all parties benefit from the new plan. A priori, in the design of a quantity flexibility contract, Lonza must anticipate the potential for renegotiation. We develop a stylized model of the biologics industry and use cooperative game theory to address the following questions: If the buyer anticipates contractual renegotiation, how will this influence his incentives for investing in drug development? How does renegotiation affect the amount of "flexibility" in contracts and the overall level of capacity investment? What is the impact of renegotiation on aggregate expected profit? Sell the Plant? The Impact of Contract Manufacturing on Innovation, Capacity and Profitability Erica L. Plambeck—Stanford Graduate School of Business Terry A. Taylor In the electronics industry and others, Original Equipment Manufacturers (OEMs) are selling their production facilities to contract manufacturers (CMs). The CMs achieve high capacity utilization through pooling (supplying many different OEMs). Meanwhile, the OEMs focus on innovation: research and development, product design, and marketing. We use cooperative game theory to investigate how this new industry structure affects investment in innovation and capacity, and thus profitability. In particular, because innovation is noncontractible, we find that OEMs will invest less in innovation than would be ideal for the industry as a whole. Hence, although contract manufacturing can increase profit through more efficient capacity utilization, it may instead reduce profit by weakening the incentives for innovation. Contract manufacturing improves profitability for the industry as a whole if and only if OEMs are in a strong bargaining position vis-à-vis the CM. As an alternative to dealing with the CM, OEMs may retain their production facilities and pool capacity with other OEMs through supply contracts or a joint venture. We show that this can cause over investment in innovation and capacity. Nevertheless, weak OEMs would do better to outsource among themselves rather than sell the plant to a CM. Dynamic Control of an Assemble-to-Order System with Multiple Products and Common Components Erica Plambeck—Stanford Graduate School of Business Amy Ward We consider an assemble-to-order system with a variety of different products, some of which have common components. The customers for these various products differ in their price-sensitivity and delay tolerance. We assume that customer orders for each product arrive according to a renewal process whose rate is a decreasing linear function of both the price and lead-time for the product. We also assume components arrive according to a renewal process whose rate is negotiable. The manager of this system must undertake the following control activities. Set a static price for each product. Set the static production capacity (delivery rate) and maximum inventory level for each component. Quote a dynamic lead-time to potential customers for each product. Schedule customer orders for assembly dynamically. In dynamic lead-time quotation and scheduling, the system manager takes into account both component availability and order backlog. His objective is to maximize the long run average system profit rate, subject to the constraint that lead-time quotations are accurate. By assuming the time scale on which customers measure delay is large compared to the inter-arrival times of potential customers, we show that an optimal pricing scheme will place the system in "heavy traffic", where utilization of the production capacity for each component is close to 100%. Hence, the scheduling problem corresponds to a diffusion control problem. We characterize this control problem and find its solution. We then reinterpret the solution to the diffusion control problem in terms of scheduling rules for the assemble-to-order system. In particular, we give conditions under which a static priority policy is optimal. 22 Managing Product Variety: Does Sharing Components across Products have Different Consequences for Consumer Perceptions and Consumer Choice? Kamalini Ramdas—University of Virginia While introducing new products can increase sales revenues, product variety can increase the costs of product design, manufacturing, distribution and after-sales support. Many assembled product manufacturers view component sharing – using the same version of a component across multiple products - as a means to achieve high variety at low cost. While recent research has examined the impact of components sharing on costs, the market impact of components sharing is not well understood. Yet conventional wisdom suggests that components sharing can have significant detrimental market impact. In this paper, we will empirically examine the impact of the sharing of different types of components on A) consumer choice and B) perceived similarity across products. To do this, we conducted a simulated test market (STM) experiment at a major multinational wristwatch manufacturer. New wristwatches may share both visual and functional components with existing watches. For our study, the company selected 13 market-ready prototypes, together with corresponding ―baseline sets‖ of 4-5 existing models that were deemed most likely to lose share to each prototype. In the STM experiment, we measured consumers’ preferences first over each baseline set, and then over augmented sets comprised in each case of a baseline set and its corresponding prototype, in order to estimate cannibalization. Consumers were also asked to rate similarity between each prototype and existing models in its baseline set. Further, we collected information on exactly what components were shared between each prototype and each existing model. By regressing the share of points lost, and rated similarity, against actual similarity, we estimate the impact of different types of components sharing on consumer choice and perceived similarity. We hypothesize that the specific components which, when shared, increase perceived similarity between models may differ from those which, when shared, impact consumer choice. We also examine the impact of brand sharing and price differentials on both consumer choice and perceived similarity. Examining these issues is important to managerial practice because it will allow assembled product manufacturers to focus their components sharing efforts on those components that have the least detrimental market impact. Choice of Inventory Structure by Internet Retailers: An Empirical Examination of the Role of Inventory Ownership Taylor Randall—University of Utah Serguei Netessine—University of Pennsylvania Nils Rudi—University of Rochester Should Internet retailers integrate a virtual storefront with in-house fulfillment capabilities or should they de-couple the sales process from physical assets by utilizing the inventory of wholesalers and distributors? This paper uses data from a sample of 55 Internet retailers to examine the role of inventory in electronic retailing. We aim to empirically inform the following questions. First, what factors influence firms to de-couple inventory and virtual storefronts in Internet retailing? Second, what effect did the inventory choice have on the success of each electronic retailer? We find that while Internet technology allows de- coupling of the information intensive sales process and the physical inventory management process to a higher extent than traditional settings, there are many circumstances where integration is still prudent. Further, empirical data shows that older firms delivering smaller products with lower amounts of product variety tend to hold inventory. Finally, we examine whether our findings represent logical inventory choice by creating a measure of inventory irrationality. We find that firms making irrational inventory decisions tend to go bankrupt more likely than those making rational inventory choices. The Choice of Warranty Cost Charge-Back Mechanisms to Minimize Supply Chain Quality Costs Canan Savaskan—Kellogg School of Management, Northwestern University In the automotive Industry, manufacturers have primarily been responsible for warranty management, which included financing of part costs, labor costs and sublet charges for replaced product parts/systems by the dealers. Today, as suppliers assume more of the system integration responsibilities in products, they are put under pressure to help improve customer satisfaction and profitability by reducing warranty costs through investing in design and manufacturing process quality and sharing the risks/reward with the manufacturers. For an OEM, the challenge is to find proper incentive mechanisms to share warranty costs with their suppliers, which would induce optimal participation of the suppliers to warranty cost reduction related activities such as, investment in product design and manufacturing process quality improvements. This paper examines the impact of warranty charge-back mechanisms on supplier's incentives to invest in quality improvement. Using the principal-agent framework, under different contractual environments, we compare three types of cost sharing mechanisms which are currently being used in practice. i) fixed share rate, ii) incentive (target) system, and iii) supplier responsible rate. Contractual environments are characterized with respect to the uncertainty in outcome of the supplier's quality improvement efforts, risk aversion of the supplier, the uncertainty in the responsibility determination process and the role of the supplier in the overall product development process. The analysis provides insights as to the use of different schemes under different contractual environments. Online Haggling and Price-Discrimination in a Name-Your-Own-Price Channel: Theory and Application Sergei Savin—Columbia Business School Il-Horn Hann—GSIA, Carnegie Mellon University Christian Terwiesch—The Wharton School, University of Pennsylvania 23 The emergence of the Internet and its extensive usage for electronic commerce has given companies the opportunity to experiment with a number of innovative pricing models. A well-known example of these new pricing models is the concept of name-your-own-price models and, more generally, the concept of online haggling. In this setting, the potential buyer suggests a price, which can then be either accepted or rejected by the seller. In this article, we present a formal model of the haggling between a seller and a set of individual buyers in a name-your-own-price setting. The seller's problem is to choose a decision rule determining which consumer bids to accept. This decision rule can be static (a fixed threshold price) or dynamic (different thresholds for different consumers and/or bidding rounds). Our study is motivated by several name-your-own-price companies that have recently emerged on the Internet. Based on detailed transaction data of a large German reverse buying site, we compare different decision rules for the seller and compare their financial implications, both relative to each other as well as relative to the actual threshold prices chosen by the company we study. Inventory Models with Regular and Expedited Sourcing: Single Index Policies Alan Scheller-Wolf—Carnegie Mellon University Geert-Jan van Houtum—Eindhoven University of Technology We examine a single stage periodic-review inventory system with backorders, in which replenishment can be obtained either through a regular channel or, for a premium, via an expedited channel with a smaller lead-time. Such dual-channel systems have become more common recently, being used for example by The American Red Cross delivering blood to hospitals, Caterpillar Inc. sourcing compact machines and work tools, and Hewlett Packard shipping server components. Our objective is to minimize the undiscounted infinite horizon average cost subject to either a minimum service level constraint or a linear penalty cost for unfilled orders. As optimal policies for these problems are complex, we propose an order-up-to policy with regular and expedited base stock levels as a heuristic, mirroring common industry behavior. For our single index policy, we derive simple expressions for the optimal regular base stock level for each difference between the regular and expedited levels. We then present an efficient procedure utilizing mixtures of Erlang distributions, fit to the first two moments of demand, to calculate this optimal regular level given a parameter difference, as well as each parameter pair's cost. This enables us to find the optimal single index policy for any given grid search tolerance within a few seconds. We include a computational section investigating the behavior of our policy as problem parameters change, and comparing our solutions with optimal single sourcing costs. Further we demonstrate the accuracy of our mixed Erlang approximation via comparisons with a common, but analytically difficult, distribution used to model demand in these systems, the truncated normal. Performance Improvement Paths: Empirical Evidence from the U.S. Airline Industry Gary D. Scudder, Michael A. Lapré—Owen Graduate School of Management, Vanderbilt University Several articles have been written during the past few years examining performance improvement paths and operating frontiers in operations. In a series of articles in the mid 1990’s, Kim Clark, Bob Hayes and Gary Pisano discussed current and desired operating frontiers. The question raised was how should a company move from one operating frontier to another, presumable better operating frontier. Or, as Schmmener and Swink and Vastag would state it, how to improve the operating frontier and move nearer the asset frontier. A ―better‖ operating frontier would be one where a company could, for example, achieve lower cost AND higher quality relative to its previous frontier. Should a company attempt to improve on both (or more) performance measures at one time? If not, which one should be addressed first? Does the sequence matter? In this paper, we attempt to shed some light on these questions using a database for 10 companies from the U.S. airline industry for a period of 11 years. The results show that better performing airlines (in terms of cost-quality position) seem to follow different improvement paths from the others. The Impact of Cost-Information Sharing in a Two-Echelon Supply Chain Yuelin Shen, Sean P. Willems—Boston University As companies move forward creating outsourced and virtual supply chains, understanding the linkages and interdependencies between participants becomes even more critical. One of the most critical dimensions to understand is the role that cost information plays in determining optimal supply chain performance. This paper revisits the classic buyback paper by Pasternack (1985) in the presence of incremental cost at the retailer and different information sharing mechanisms. Under an unlimited return policy at a predetermined buyback price, we propose a Stackelberg game where the manufacturer leads by setting the buyback price and production quantity while the retailer follows by choosing the order quantity. The manufacturer s decisions critically depend on the manufacturer s knowledge of the cost structure at the retailer. In the full information case, the buyback price monotonically increases as a function of the retailer s incremental cost, and can exceed the wholesale price to maintain system-optimal profits. For the no-information case, the manufacturer orders too much and sets the buyback too low, thereby creating a supply chain that can not achieve system-optimal profitability. For the asymmetric information case, an optimization problem is suggested for the manufacturer, and a heuristic solution is developed, which performs comparably to the full information case. Managing Supply Chain Risks with Derivatives Dailun Shi—IBM T. J. Watson Research Center 24 Richard L. Daniels—University of Georgia William Grey—IBM T. J. Watson Research Center This paper discusses the value and impact of derivatives for managing supply chain risks. Our analysis is based on a simple two-stage supply chain in which a retailer can either buy product directly, or purchase options on product. We first derive optimal replenishment policies for the retailer, optimal production policies for the supplier, as well as close-form solutions for optimal expected profits. We then show how options enhance information flows, encourage risk sharing, and improve supply chain efficiency. The paper includes a discussion of how options can be used to align the incentives of supply chain partners, and to improve supply chain responsiveness to changes in the business environment. We also derive analytic and numerical results that provide insight into how supply chain participants could most effectively utilize options to enhance their profitability. Gatekeepers and Referrals in Services Robert A. Shumsky, Edieal J. Pinker—Simon School, University of Rochester We examine services in which customers encounter a gatekeeper who makes an initial diagnosis of the customer's problem and then may refer the customer to a specialist. The gatekeeper may also attempt to solve the problem, but the probability of treatment success decreases as the problem's complexity increases. Given the costs of treatment by the gatekeeper and the specialist, we find the firm's optimal referral rate from a particular gatekeeper to the specialists. We then consider the principal-agent problem that arises when the gatekeeper, but not the firm, observes the gatekeeper's treatment ability as well as the complexity of each customer's problem. We examine the relative benefits of compensation systems designed to overcome the effects of this information asymmetry and identify when bonuses based solely on referral rates do not ensure first-best system performance. We also consider the value of such output-based contracts when gatekeepers are heterogeneous in ability, so that two gatekeepers face different probabilities of treatment success when given the same problem. Finally, we compare environments in which the gatekeeper is, and is not, faced with risk in the form of significant variance in compensation. A Conceptual Model for Maximizing Equipment Operating Time Donald E. Simmons—Ithaca College This paper makes use of classical economic theory to establish a model for analyzing the effects of mixing specialized and flexible servers when providing after-sales support of purchased equipment. The objective in this environment is to maximize equipment operating time, while balancing the customer’s delay cost and the firm’s cost of providing service. The model development starts with the simulation of a queuing system, which includes equipment failures (arrivals) and various combinations of specialized and flexible servers. Flexible servers reduce delay time by being able to respond to multiple types of problems, but cost more per unit time due to the amortization of required training costs. Specialized servers have lower training costs, but may cause higher customer delay costs due to possible mismatches between their skills and the requirements of arriving jobs. This system will be conceptually modeled by using the concept of iso-profit curves to represent levels of machine ―uptime‖ (based on reciprocals of expected downtime costs). Each point on an iso-profit curve represents the same positive measure of performance, determined by varying the mix of servers. A rational firm will want to be on the iso-profit curve farthest away from the origin, subject to a budget constraint. It will be shown that the slope of the budget line is –(1 + p), where p is the premium paid for flexibility. As p increases, the slope of the budget line becomes more negative and the line rotates inward toward the origin, creating a tangency point with a lower value iso-profit curve. Examining the effect of changes in the premium leads to the classic derivation of a demand curve for flexible servers. Assessing the elasticity of the demand curve for a specific range of p can provide useful insight for managers regarding the substitution rates of flexible and specialized servers. Dynamic Yield-Based Dispatching for a Multi-Product Reentrant Flowline Thomas W. Sloan—University of Miami In many manufacturing environments, equipment condition has a significant impact on product quality and yield. Examples range from the wear of tool bits in a machine shop to the contamination of ultra-clean production environments needed for the manufacture of pharmaceuticals and integrated circuits. However, previous research on production scheduling problems with random yield has primarily focused on single-product, single-stage systems and has treated equipment condition as a variable that is beyond the decision maker’s control (see Yano & Lee 1995 for a review). Thus, production decisions are reduced to determining input quantities given a particular pattern of process/equipment deterioration. The purpose of this paper is to explore how expanding the decision process to incorporate information about equipment condition and yield can improve scheduling policies. Building on recent work by Sloan & Shanthikumar (2002), we develop a stochastic dynamic programming model of a multi-product, reentrant flowline with multiple machines at each stage, where each machine deteriorates as more units are processed. The machine condition affects the yield of different products differently, and the equipment can be returned to like- new condition by stopping production and performing a maintenance or cleaning action. At each stage and each decision epoch, one must decide whether to clean the equipment, thus improving the yield, or to continue producing. If one chooses to produce, then one must also select the product type. The objective is to maximize expected average profit while still meeting output targets. 25 The convoluted, reentrant process flows associated with semiconductor manufacturing and the dependency between current decisions and future yield make this a difficult problem to solve analytically. Furthermore, even if an optimal policy is determined, it is not clear how it would be implemented in a real production facility since not all products would be available at a particular stage at any given point in time. Thus, we propose a decomposition approach that treats each stage separately. The production and maintenance policies are updated dynamically based on the actual attained yield and changes in equipment condition. Using a discrete-event simulation model, we compare the dynamic policy to static, heuristic policies as well as other policies traditionally used in practice. Preliminary results indicate that using the dynamic policy can substantially improve yield. An important goal of the study is also to determine when the dynamic policy is not significantly better—in other words, to identify situations in which simple policies perform well. By doing this, we hope not only to extend the theory of production scheduling with random yield but also to provide insights that will be of use to practitioners. Inventory Replenishment with a Financial Criterion Matthew J. Sobel, Jiang Zhang—Weatherhead School of Management, Case Western Reserve University We consider a setup cost inventory model that would yield an optimal (s,S) policy if the criterion were minimization of the expected present value of the time stream of costs. However, the financial nature of a firm suggests other criteria; our criterion is the market value of a single-product firm that is entirely equity financed and which does business at a single location. So we maximize the expected present value of the time stream of dividends received by the owners. We discuss the optimality of (s,S) policies with this criterion. The firm reviews inventories periodically, experiences a sequence of nonnegative demands that are independent and identically distributed random variables, has convex inventory-related costs each period, incurs a setup cost each time that replenishment occurs, receives revenues for ordered goods, and excess demand is backlogged. Liquidity limits the dividend that can be issued and obliges the firm to monitor its working capital as well as its inventory level. That is, replenishment and dividend decisions directly and indirectly affect both working capital and inventory level. This leads to a dynamic programming formulation in which both the state and action are vectors. The analogous problem without a setup cost and a criterion of cost minimization or profit maximization is the standard dynamic newsvendor model, of course, which has an optimal base-stock-level policy. When the criterion is the expected present value of dividends, standard methods yield the same kind of optimal policy for this criterion as for the common criteria (Li, Shubik and Sobel, 1997). However, when a setup cost augments the model, existing methods of proving that an (s,S) policy is optimal for a cost or profit criterion do not seem useful when the criterion is the expected present value of the dividends. We describe a proof technique that circumvents the obstacles. More generally, our results suggest that standard methods of analysis may often not suffice for inventory and production models with financial criteria. Adaptive Inventory Control for Partially Observed Markov-Modulated Demand Charles R. Sox—Department of Industrial & Systems Engineering, Auburn University Some organizations must manage inventories in environments where the demand process is both dynamic and random. The dynamic nature of the demand may also be characterized by large shifts in the demand distribution that occur suddenly at discrete points in time. These shifts can result from scheduling, marketing, or management policy changes made by a major customer or by product introductions and marketing promotions by a major competitor. We model this problem as an inventory control problem with Markov-modulated demand. The demand that occurs in each time period is generated by one of a finite number of demand distributions. Each of these distributions is associated with a state of a discrete time Markov chain. At the beginning of each time period, this Markov chain may transition to a new state, and the state of the Markov chain determines which of the distributions will generate the actual demand observation for that time period. Furthermore, the decision maker does not know if a transition has occurred but must infer it from the demand observations. There are some interesting computational issues that arise in the application of these models to actual industrial inventory control problems. First, there is the problem of parameter estimation. The transition probability matrix and demand distributions may need to be estimated from historical demand observations. Second, the calculation of the probability distribution of lead-time demand is non-trivial and potentially computationally challenging. Finally, we present some computational results that compare the performance of several different heuristic policies for this problem. Biform Analysis of Competitive Inventory Decisions Harborne W. Stuart, Jr.—Columbia Business School This paper provides a model of the competitive newsvendor problem in which there is price competition following the inventory decisions. Using the biform game formalism of Brandenburger and Stuart (2001), the price competition is modeled by considering the core of the induced cooperative game. Even without the introduction of uncertainty, this type of analysis shows that existence of a stable, price equilibrium can be problematic. In particular, there are situations in which an (non-trivial) equilibrium does not exist, and there are situations in which the only equilibria involve some form of buyer-specific pricing. Quality Controls in Make-to-Order Queues Xuanming Su, Stefanos Zenios—Stanford University 26 Consider a make-to-order production system in which products are of variable quality. Production is modeled by a single server M/M/1 queuing system. In the order specified by the queue, customers may choose to accept the product and leave the system, or to pass it on to the next customer and retain their position in the queue. We wish to maximize the expected utility of each customer by choosing dynamic quality controls - only products with quality exceeding a certain threshold are offered to customers. Higher quality controls imply lower service rates, and we need to ensure that all products offered will be accepted by some customer in the queue. In this way, this model captures two major issues in production planning: (1) tradeoff between quality and service rates, and (2) incentive effects between participating customers. For both FCFS and LCFS queue disciplines, we show that it is optimal to always operate at capacity, i.e. impose no quality controls at all. In fact, under LCFS, this attains the first-best outcome obtained when customers are contracted to accept products of any quality. However, in both cases, quality controls are implicitly enforced by the customers, because not all customers are willing to accept all product types. Case Studies on Implementing POLCA — A New Card-based Control System for High Variety or Custom Engineered Products Rajan Suri—University of Wisconsin-Madison Companies with high-variety or custom-engineered products are struggling to implement effective material control strategies on the shop floor, and finding that Pull/Kanban systems are not meeting their needs in such environments. POLCA (Paired-cell Overlapping Loops of Cards with Authorization) is a quick response manufacturing strategy designed with these situations in mind. POLCA is a hybrid push-pull strategy that combines the best features of card-based control systems and MRP systems. At the same time POLCA gets around the limitations of card-based systems in a high-variety or custom products environment. In partnership with our member companies, the Center for Quick Response Manufacturing (QRM) has implemented POLCA at several factories over the past year, including at three Rockwell Automation factories in the US and Canada. This talk will consist of three main parts. First we will give an overview of the POLCA system, the basis for its design, and how it works. Next we will present results of theoretical research that compares the performance of push systems, pull systems and POLCA systems. This will help us to see, from a theoretical viewpoint, the situations where POLCA can be most effective. Finally we will present practical case studies on the implementation of POLCA at several factories. For this we will discuss why the companies chose to go with POLCA after evaluating all the alternatives, and then we will show how POLCA has been used and the resulting performance improvements obtained at those companies. These improvements include reductions in lead time, employee satisfaction with the system, and dramatic increases in percentage of on-time deliveries. Maximizing Project Cash Availability Subject to Activity-Dependent Resource Constraints Joseph G. Szmerekovsky, George L. Vairaktarakis—Case Western Reserve University This paper extends current work in the area of allocating activities across partners in a project network so as to minimize activity-in-process costs. The paper considers the problem of minimizing a cost function expressed as a linear combination of completion times of activities. The activities are subject to precedence constraints and resource constraints. The resource constraints represent partners in a virtual manufacturing environment, each possessing different manufacturing capabilities. That is, not all partners can perform all tasks and the processing times of activities are partner-dependent. Combined with the precedence constraints the partners compose a project network. In addition to extending the project management literature this problem subsumes the majority of classical scheduling problems. Among the environments included as a special case of this model are the Flow-Shop, Job-Shop, Open-Shop and Unrelated Machine environments. Since the problem considered is very general the methodology employed is effective heuristics accompanied by lower bounding schemes used for performance evaluation. The heuristic procedures developed are used to determine the impact of flexibility on work-in-process costs. Flexibility in the model comes from the freedom to assign an activity to one of a variety of partners and in the speed with which different partners are willing to complete the activity. Understanding the value of this flexibility plays an important role in evaluating contractual agreements in which the project manager agrees to assign a certain amount of work to a partner and in return the partner guarantees a reduced processing time for the work. The heuristic procedures are used to identify the reduction in work- in-process costs for increased levels of processor flexibility, helping to identify the optimal level of flexibility for the system. Equilibrium Analysis of a Natural Gas Supply Chain Sridhar Tayur, Wei Yang—GSIA, Carnegie Mellon University Motivated by the senior leadership and the risk management group at Equitable Resources Incorporated for a model based on fundamental economic principles, we develop a discrete time, infinite horizon and equilibrium-based model for a typical natural gas supply chain where there is a perfect market (a source that represents Henry Hub in Louisiana), a market with a small number of producers (a sink which represents the Pennsylvania, East Ohio and other northeastern regions) and a pipeline connecting them. We extend the available literature by first developing a stationary rational expectations equilibrium (SREE) model for the market with a small number of firms. We show that the value of incorporating entry and exit is significant. A general SREE model is then developed for the entire supply chain. An optimal delivery and storage solution is presented for the pipeline based on the equilibrium values. The features of spot and forward prices in both markets are compared and analyzed. We test a commonly used approximation by comparing with the exact solution and find that the former is not satisfactory. 27 Dynamic Lot Sizing with Learning and Forgetting In Production Sunantha Teyarachakul, Suresh Chand, Jame Ward—Krannert Business School, Purdue University We consider a discrete-time, finite-horizon, dynamic lot-sizing problem with known dynamic demands. In our analysis, Wright’s learning function and Globerson’s exponential forgetting function have been used. We present an optimal lot sizing algorithm under the assumption that inventory is zero when an order arrives. We propose a dominance property to reduce the computational effort. Optimal lot sizing under total transmission of learning can be obtained by Wagner-Whithin Algorithm. We solve several numerical examples to understand the effect of learning and forgetting on the batch sizing decisions. The results include; 1) for any given level of forgetting, as the ability to learn increases, the optimal lot size decreases, 2) for any given level of learning, as the level of forgetting increases, the optimal lot size increases and 3) as planning horizon (t) increases, the efficiency of dominance property also increases. A Periodic Inventory Model for Modular Service Parts Doug Thomas, Don Warsing—Smeal College of Business Administration, The Pennsylvania State University Modular product design has received a great deal of attention in recent years. Such design can enable mass customization and postponement strategies and simplify the servicing and repair process. In this work, we model a service parts inventory system where there is exogenous demand for both a complete assembly and its modular components. Our goal is to reduce overall service costs when assembly and/or disassembly can occur at a distribution point by explicitly considering the relationship between an assembly at its components. Demands (field failures) for the assembly and its components are assumed to be independent. In the event of any product shortage, a backorder cost is incurred. This cost is associated with a field service failure and does not depend on the desired assembly or component. For a periodic review inventory system with no ordering cost, we find stocking policies to minimize holding, backorder, assembly and disassembly costs. Since finding optimal policies is computationally intensive, we present and evaluate a heuristic simple enough to be embedded in a spreadsheet. Optimal Auctioning and Ordering in an Infinite Horizon Inventory-Pricing System Garrett van Ryzin, Gustavo Vulcano—Columbia University We consider a joint inventory-pricing problem in which buyers act strategically and bid for units of a firm's product over an infinite horizon. The number of bidders in each period, as well as the individual bidders' valuations, are random but stationary over time. There is a holding cost for inventory and a unit cost for ordering more stock from an outside supplier. The firm must decide how to conduct its auctions and how to replenish its stock over time to maximize its profits. We show that the optimal auction and replenishment policy for this problem is quite simple, consisting of running a standard first-price or second-price auction with a fixed reserve price in each period and following an order-up-to (basestock) policy for replenishing inventory at the end of each period. Moreover, the optimal basestock level can be easily computed. We then compare this optimal auction mechanism and replenishment policy to a traditional fixed price, basestock policy. We prove that in the limiting case of one buyer per period and the limiting case of a large number of buyers per period, list pricing is optimal. List pricing is also optimal as the holding cost tends to zero. Numerical comparisons confirm these theoretical results and show that auctions only provide a significant benefit when: 1) the number of buyers is moderate, 2) holding costs are high or 3) there is high variability in the number of buyers per period. Indeed, one can argue that most retail and industrial trade corresponds to cases where our results suggest auctions are of minimal benefit. However, under the right conditions, our analysis shows an optimal auction and replenishment policy can provide significant improvements in profits over list pricing. Information Sharing and Competition in a Supply Chain for Short Life-Cycle Product Scott Webster—Syracuse University Z. Kevin Weng—University of Wisconsin-Madison We consider a supply chain comprised of one or more suppliers and a single manufacturer. The manufacturer is planning a one-time production run of a new product with a short life-cycle. The suppliers produce a key component. The manufacturer knows his cost structure and how demand depends on the product price, but knows nothing about the supplier cost structure. Each supplier knows her cost structure, but nothing about the market for the new product or the manufacturer’s cost structure. In this setting, the manufacturer may issue a request-for-quote containing component specifications. Each supplier responds with a price/quantity schedule for the component, and the buyer uses this information to determine a production quantity, a product price, and how the component quantity will be allocated among the suppliers. This paper investigates the viability of sharing market information with suppliers. In particular, should the manufacturer share information on how component prices will influence purchase quantities? Returns Policies and Incentives for Collaborative Forecasting Prashant Yadav, Charles P. Schmidt—Information Systems, Management Science and Statistics, University of Alabama In the distribution of seasonal products, returns policies (Pasternack 1985) offered by the manufacturer have gained wide acceptance both in industry and academic literature. Numerous firms are also implementing some form of a Collaborative 28 Forecasting and Replenishment strategy. Literature on the strategic role of collaborative forecasting and coordination issues in a supply chain is rich, but so far offers little treatment of the incentives to the retailer or manufacturer to engage in joint forecasting. A retailer has knowledge of local variables affecting demand for a product in his region and hence can effectively contribute to forecast development. Likewise a manufacturer is aware of the overall marketing strategy for his product and has more resources to use sophisticated techniques for forecasting. In a joint forecasting initiative both parties work together to develop a sales forecast for the season. Using a stylized newsboy model we show that a retailer’s incentive to participate in joint forecasting depends upon the nature of the buyback contract offered by the manufacturer. We consider both parties to have equal knowledge of the nature of the stochastic demand; however, their knowledge of the parameters of the demand distribution depends upon costly efforts they exert on forecasting. The forecasting effort of each party is assumed to be unobservable and non-contractible. We characterize the Nash equilibrium and develop insights into the forms of transfer payment schemes that could lead to the manufacturer and the retailer exerting optimal efforts on forecasting. We extend our analysis to relate to the double moral hazard framework in economics and analyze different functional forms of the effect of effort on forecast accuracy. Inventory Sharing and Rationing in Decentralized Dealer Systems Hui Zhao, Vinayak Deshpande, Jennifer K. Ryan—Purdue University We consider a dealer inventory sharing problem in a continuous review, infinite horizon, decentralized system. Each dealer in the system faces multiple priority classes of demand (regular customers and inventory sharing requests from other dealers) and makes his own inventory control and inventory sharing decisions with the objective of minimizing his individual cost. We start by analyzing the inventory sharing problem in a make-to-stock system in which production/replenishment takes an exponential lead time. We develop dynamic programming formulations for the individual dealer problem under two different inventory sharing schemes (LS and BO). We prove that the optimal policy is a stationary base-stock policy with a fixed rationing level for the sharing scheme LS. We then expand our research to consider a multi-dealer problem. We develop dynamic programming formulations for centralized two-dealer inventory sharing problem under the two sharing schemes. We use the formulations to study the structure of the optimal policy. Queuing models based on a stationary base-stock and rationing policy are developed for both the individual- and the two-dealer problems in order to obtain expressions for the long run average cost. We further our research by analyzing the behavior of a decentralized inventory sharing system in which dealers choose their inventory control and inventory sharing parameters in anticipation of other players’ strategies (Nash Equilibrium). Computational studies of the decentralized inventory sharing system demonstrate the benefits of inventory sharing under different cost parameters and system incentives for sharing. We believe that this research is a first attempt to simultaneously consider inventory sharing and inventory rationing for multiple classes of demand in a decentralized system. 29