McGraw-Hill Higher Education ‘Aino The McGraw Hil Compo DDISIGNING AND MANAGING THE SUPPLY CHAIN ‘Copyright © 2000 by The McGraw-Hill Companies, Inc. All rights reserved. Printed in the ‘Une Stevo Amer, Escep ae permite under the United Sates Copyright Act 1976, ‘opartef us pubbestion nay berspoduced or ited nny for ory an teans ot ‘Sore inadatubenr tetra ystem, wou! te pie wee permission ofthe publisher This hooks printed on aire paper 294567890RGR/FOR909876543210 ssn o2se261687 Vice president Editor Chief Mia! W. Junior Publisher: oy. Saoed Senior sponsoring eon Sit nbs Eston eortinaton late Shaper Senior maetingwanager Collet). Subic Projo manages Kimberly Mona ‘Mattage new book proucon: Mele Slat Desgne: Suzmne Noweaer Supplement corres Fase M. Range Now ida ‘Chole Peto (Coonpontor hats ne. “ypetce 10/2 Paitin Frnt Quo Printing Book Groupee brary of Congress Cataloging -Publeton Data Sich Le, David ‘Designing and managing the supply chain / Davi inc Le, hip Raminey, Bath Sich Lev zen In¢lade bibliographical references and index aN oo7oa42 [hye dstriation of goods—Managerent, 2. Marketing channele-Management. 3 Basins ogsts. 4 Inu Drocurement Kaminsky Philp. Simcht-Lei, Bab. invite BesiIs7S05_ 2000 e85-—del seam p//oneeenhe com "To our children, Sara and Yuval, who have the patience and humor to survive our work logether DSL,ESL ‘Tomy fay, for hele support and encouragementTHE IRWIN/MCGRAW-HILL SERIES Operations and Decision Sciences (ONS MANAGEMENT and Closs, Logistical Management: The Integrated Supply Chain First Edition uilano, and Jacobs, Production and Operations Management, tition onstein, and Greenlavs, PROSIM for Windows, Third Edition \"Apte, Manufacturing Automation, First Eition silano, and Chase, Fundamentals of Operations Management, ition 4 Burt, Purchasing and Supply Management, Sixth Edition slobal Operations Management, Firs! Edition ‘ns and Fitzsimmons, Service Management: Operations, Strategy, tion Technology, Second Edition Larson, Project Management: The Managerial Process, first Edition afacturing Strategy: Text & Cases, Third Edition Spearman, Factory Physics, Second Eaition rd Stock, Strategic Logistics Management, Thind Edition 1nd Fearon, Purchasing and Supply Chain Management, Edition sd Denzler, Operations Management, Second Edition shadsi, and Yakie, HOM Operations Management Software for ¢s, First Edition Production and Operations Analysis, Thind Edition Competitive Manufacturing Management, Firs! Edition d Cho, Operations Scheduling, Fist Edition ‘and Uzumeri, Managing Product Families, Firs! Edition | Operations Management: Contemporary Concepts and Cases, tion ser and Knod, Operations Management: Customer-Focused Principles, ition vi, Kaminsky, Simchi-Levi, Designing and Managing theSupply Chain: ts, Strategies, and Case Studies, First Edition Business Dynamics: Systems Thinking and Modeling for a Complex First Edition 1, Production/Operations Management, Sixth Eiition Beery, and Whybark, Manufacturing Planning & Control Systems, ition sundations of Inventory Management, First Edition ITATIVE METHODS AND MANAGEMENT SCIENCE arraway, Frey, Pfeifer, Quantitative Business Analysis: Casebook, ition arraway, Frey, Pfeifer, Quantitative Business Analysis: Text and Cases, ition fusman, and Bierman, Quantitative Analysis for Business Decisions, dition About the Authors David Simchi-Levi is Professor of Industrial Engineering and Manage- ‘ment Science at Northwestern University. Prior to joining the faculty at Northwestern, he taught at Columbia University. Professor Simchi-Levi received his Ph.D. in operations research from Tel-Aviv University and has won awards for his work in supply chain, logistics, and transportation. He is the co-author, with Julien Bramel, of The Logic of Logistics, published by Springer-Verlag, and the founder and chairman of LogicTools, Inc, a co pany that provides decision support software for supply chain and logis ‘management. Philip Kaminsky is Assistant Professor of Industrial Engineering at the University of California at Berkeley. He received his Pa. in Industrial Engineering from Northwestem University. Prior to his graduate studies, he worked for the production division of Merck & Co,, Ine. He has con- sulted in the areas of supply chain and production management, Edith Simchi-Levi is President of LogicTools, Inc. She received her BS. in Mathematics and Computer Science from Tel-Aviv University. She has extensive experience in software development and numerous consulting, projects in logistics and supply chain management.SEE Foreword In the last few years we have seen an explosion of publications on supply’ chain management; numerous books have been published and many ar ticles have appeared in academic, trade, and popular magazines. These publications are either too technical—and therefore inaccessible to prac- titioners and students—or they lack the breadth and depth that the topic deserves. Certainly, itis difficult to find a book appropriate for teaching supply chain management to business or engineering students. Designing and Managing the Supply Chain solves this problem! ‘The book is an important contribution and major milestone for the supply chain community. Itis the first book that covers a comprehensive breadth ofsupply chain topics in depth, and addresses the major challenges in this area. It was written by experts from academia and industry who have been researching, consulting, and developing software for supply chain management for many years. This book includes many classic and new case studies, numerous ex- amplesas well as in-depth analysis of some of the technical issues involved in inventory management, network design, and strategic partnering, to name a few. It is therefore an ideal textbook for classes on supply chain ‘management at the undergraduate, Master’s, and M.B.A. levels. Since each chapter is self-contained, instructors can pick the chapters they want touse, depending on the length ofthe class and its requirements. The book comes with two computerized games. The Computerized Beer Game pro- vides an excellent instructional fool that engages students in managing a supply chain and provides a starting point for discussing the value of in- formation in the supply chain, strategic partnering, centralized decisionFoner soar sen dn mange le ig Te ruc muy ay mapa co peo Te ee Gee cot cpm mn sry a amg te cepts tet leet FE ig nae empemengoae et Hau L. Lee Kleiner Perkins, Mayfield, Sequoia Capital Professor Director, Stanford Global Supply Chain Forum Stanford University Preface This book grew out of a number of supply chain management courses and Executive Education programs we have taught at Northwestern Univer: sity over the past several years, as well as numerous consulting projects land supply chain decision-support systems we have developed at Logic- ‘Tools. The courses, taught in the Master of Management in Manufacturing (MMM) program, a joint M.A. program between the Kellogg School ‘of Business and McCormick School of Engineering at Northwestern Uni- ‘versity and in Executive Education programs sponsored by Kellogg, have Spawned many innovative and effective supply chain education concepts. ‘The focus in these programs has been on presenting, in an easily accessible manner, recently developed state-of-the-art models and solution methods important in the design, control, and operation of supply chains. The con- sulting projects and decision-support systems developed by LogicTools have focused on applying these advanced techniques to solvespecific prob- Jems faced by our client Titerest in supply chain management, both in industry and in academia, has grown rapidly over the past several years. A number of major forces have contributed to this trend. First, in recent years it has become clear that many companies have reduced manufacturing, costs as much as prac- tically possible. Many of these companies are discovering the magnitude fof savings that can be achieved by planning and managing their supply Chain more effectively. Indeed, a striking example is Wal-Mart's success, ‘which is partly attributed to implementing a new logistics strategy called cross-docking, At the same time, information and communication sys- tems havebeen widely implemented, and provide access tocomprehensive data from all components of the supply chain. Finally, deregulation of thePrice transportation industry has led to the development of a variety of trans- portation modes and reduced transportation costs, while significantly in- creasing the complexity of logistics systems. Itis therefore not surprising that many companies are involved in the analysis of their supply chains. In most cases, however, this analysis is performed based on experience and intuition; very few analytical mod- els or design tools have been used in this process. In contrast, in the last two decades the academic community has developed various models and tools for supply chain management, Unfortunately, the first generation of this technology was not robust or flexible enough to allow industry to use it effectively. This, however, has changed over the last few years during which improved analysis and insight, and effective models and decision: support systems, have been developed, but these are not necessarily famil- iar to industry. Indeed, to our knowledge there is no published work that discusses these problems, models, concepts, and tools at an appropriate level. In this book, we intend to fill this gap by providing state-of-the-art models, concepts, and solution methods important in the design, control, operation, and management of supply chain systems. In particular, we haveattempted to convey both the intuition behind many key supply chain concepts and to provide simple techniques that can be used to analyze various aspects of te supply chain. ‘The emphasis is on a format that will be accessible to executives and. practitioners, as well as students interested in careers in related industries In addition, it wil introduce readers to information systems and decision- support tools that can aid in the design, analysis, and control of supply chains, “The book is written to serve as: + A textbook for M.B.A.-level logistics and supply chain management + A textbook for BS. and MSS. industrial engineering courses on logistics and supply chain management. + A reference for teachers, consultants, and practitioners involved in any one of the processes that make up the supply chain. Ofcourse, supply chain managementisa very broad area, and it would bbe impossible fora single book to cover all of the relevant areas in depth, Indeed, there isconsiderable disagreement in academia and industry about exactly what these relevant areas are, Nevertheless, we have attempted to provide a broad introduction to many critical facets of supply chain man- agement. Although many essential supply chain management issues are interrelated, we have strived wherever possible to make each chapter as self-contained as possible, so that the reader can refer directly to chapters covering topics of interest. The discussion ranges from basic topics of in- Proce xi Pie et can planer ede ens BE ae ls mane col a Cooper ia ae en making ne ln eo Fane ee ta Re eee eae cee nen a a a ore Hon ite Prin in Opts Ma ae Sigh arin Guth Nr ea yAcknowledgments Itis our pleasure to acknowledge all those who helped us with this manu- script. First, we would like to thank Dr. Myron Feinstein, former director ‘of supply chain strategy development at Unilever, New York City, who read through and commented on various chapters. Similarly, we are in- debted to the reviewers, Professors Michael Ball (University of Maryland), Wendell Gilland (University of North Carolina, Chapel Hill), Eric John- son (Vanderbilt University), Douglas Morrie (The University of Texas, “Austin), Michae! Pangburn (Pennsylvania State University), Powell Robin- son (Texas A&M University), Wiliam Tallon (Northern Ilinois University), and Rachel Yang (University of Dinois, Urbana-Champaign). Their com- ‘ments were invaluable in improving the organization and presentation of the book. Weare also grateful to Deniz Caglar, a Ph.D. candidate at North~ ‘western University, for his comments on earlier drafts of the book. In addition, we thank Dr. Kathleen A. Stair and Ms. Ann Stuart for carefully editing and proofreading many chapters. Finally, we wish to thank Ms, Colleen Tuscher, who assisted us in the initial stage of the project, our editor, Mr. Scott Isenberg, and his as- sistant, Ms. Nicolle Schieffer, of Irwin/ McGraw-Hill, who encouraged us throughoutand helped uscomplete thebook. Also, thanks toMs. Kimberly ‘oranda and the production staff at McGraw-Hill for their help. David Simchi-Levi, Evanston, Illinois Philip Kaminsky, Berkeley, Califo Faith Simchi-Levi, Northbrook, IllinoisCase: Case: List of Cases The Bis Corporation JAM Electronics: Service Level Crisis Case: Swimsuit Production Case: Barilla SpA (A) Case: Barilla SpA (B) Barilla SpA (C) Modern Book Distribution, Inc. Audio Duplication Services, Inc. : Wal-Mart Changes Tactics to Meet International Tastes : Hewlett-Packard: DeskJet Printer Supply Chain Dell’s Direct Business Model Backup in the Espresso Lane ERP Brews Instant Success . Supply Chain Management Smooths Production Flow Brief Contents 11 Introduction to Supply Chain Management 1 2 Logistics Network Configuration 15 3. Inventory Management and Risk Pooling 39 4 The Value of Information 67 5 Distribution Strategies 109 6 Strategic Alliances 121 7 International Issues in Supply Chain Management| 145 8 Coordinated Product and Supply Chain Design 167 9 Customer Value and Supply Chain Management 197 10 Information Technology for Supply Chain Management 215 1 Decision-Support Systems for Supply Chain Management 249 Appendix A The Computerized Beer Game 275 ‘Appendix B ‘The Risk Pool Game 293, Bibliography 303 Index 309Contents 1 Introduction to Supply Chain Management 1 1.1 What Is Supply Chain Management? 1 12 Why Supply Chain Management? 5 13. TheComplexity 7 14. Key Issues in Supply Chain Management 8 15 Book Objectives and Overview 11 2 Logistics Network Configuration 15 Case: The Bis Corporation 15 21 Introduction 17 22 DataCollection 20 221 Data Aggregation 20 222 Transportation Rates 23 223 Mileage Estimation 24 224 Warehouse Costs 26 225 Warehouse Capacities 27 226 Potential Warehouse Locations 28 22.7 Service Level Requirements 28 228 FutureDemand 29 23 Modeland Data Validation 29 24 Solution Techniques 30 24.1 Heuristics and the Need for Exact Algorithms 30 242 Simulation Models and Optimization Techniques 33 25 Key Features of a Network Configuration DSS 34 ‘orporation Distribution Problem 36Cone 3 Inventory Management and Risk Pooling 39 Case: JAM Electronics: Service Level Crisis 39 3.1 Introduction 41 32_A Single Warehouse Inventory Example 43, 321 The Economic Lot Size Model 43 322. The Effect of Demand Uncertainty 46 Case: Swimsuit Production 46 323 Multiple Order Opportunities 51 324 No Fixed Order Costs 51 325 Fixed Onder Costs 54 326 Variable Lead Times 56 33° RiskPooling 56 34 Centralized versus Decentralized Systems 60) 35. Managing Inventory in the Supply Chain 61 36. Practical Issues 63 Summary 65 4 The Value of Information 67 Case: Barilla SpA (A). 67 41 Introduction. 81 42. ‘The Bullwhip Effect 82 421 Quantifying the Bullwhip Eifect 86 422 The impact of Centralized Information on the Bullichip fect 88 423 Methods for Coping with the Bullwhip Effect 91 ase: Barilla SpA (8). 93. Case: Barilla Sp (C) 95 43> Effective Forecasts 99 44 Information forthe Coordination of Systems 100 45. Locating Desired Products 101 46 Lead TimeReduction 101 47 Integrating the Supply Chain 102 47. Conflicting Objectives in the Supply Chain 108, 472 Designing the Supply Chain for Conflicting Goals 103 Summary 107 5 Distribution Strategies 109 Case: Modern Book Distribution, Inc. 108 5.1 Introduction 111 Conents xix 52 Contratized versus Decentralized Control 111 5.3. Distribution Strategies 112 53.1 Direct Shipment 113 53.2 Cross-Docking 113, 54° Transshipment 116 55° Central versus Local Facilities 116 56 Push versus Pull Systems 117 561 Push-Based Supply Chain 118 562 Pull-Based Supply Chain 119 Summary 120 Strategic Alliances 121 Case: Audio Duplication Services, Inc. (ADS) 121 64 Introduction 122 62 A Framework for Strategic Alliances 123 63. Third-Party Logistics 126 63.1 WhatIs3PL? 126 632 Advantages and Disadvantages of IPL 126 633 3PLIssuesand Requirements 129 634 3PL Implementation Issues 130 64 Retailer-Supplier Partnerships 131 64.1 Typesof RSP 132 64.2 Requirements for RSP_ 133 643 Inventory Ownership in RSP 134 644 Issuesin RSP Implementation 135 645 Steps in RSP Implementation 136 646 Advantages and Disadvantages of RSP_ 136 647 Successes and Failures 138 65 Distributor Integration 140 65:1 Types of Distributor Integration 141 652. Issues in Distributor Integration 142 Summary 143, Intemational Issues in Supply Chain Management 145 (Case: Wal-Mart Changes Tactics to Mest International Tastes 145 71 Introduction 49 72.2 Global Market Forces 150 71.2 Technological Forces 151 713 Global Cost Forces 152 744 Political and Economic Forces 152Contents 7.2__ Risks and Advantages of International Supply Chains 721 Risks 154 72.2 Addressing Global Risks 155 723 Requirements for Global Strategy Implementation 157 153 73 Issues in International Supply Chain Management 158 7.3.1 International versus Regional Products 158 732 Local Autonomy versus Central Control 159 73.3 Miscellaneous Dangers 160 74 Regional Differences in Logistics 161 7.4.1 Cultural Differences 161 10 742 Infrastructure 162 7.43. Performance Expectation and Evaluation 163 7.4.4 Information System Availability 164 74.5 Human Resources 164 Summary 165 8 Coordinated Product and Supply Chain Design 167 Case: Hewlett-Packard: DeskJet Printer Supply Chain 167 811 Introduction 177 BL Design for Logistics: Overview 177 81.2 Economic Packaging and Transportation 177 843 Concurrent and Parallel Processing 179 B14 Postponement 181 815 important Considerations B16 ThePush-Pull Boundary B17 Case Analysis 186 82. Supplier Integration into New Product Development 189 ‘82.1 The Spectrum of Supplier Integration 189 822 Keys to Effective Supplier Integration 190 823 A “Bookshelf” of Technologies and Suppliers 191 83 Mass Customization 19) 83.1 What Is Mass Customization? 191 832 Making Mass Customization Work 192 n 833 Mass Customization and Supply Chain Management 194 Summary 195 184 185 9 Customer Value and Supply Chain Management 197 Case: Dell's Direct Business Model 197 9.1 Introduction 198 92. The Dimensions of Customer Value 200 9.2.1 Conformance to Requirements 201 92.2 Product Selection 202 conten xxi 923 Priceand Brand 204 924 Value-Added Services 205 925 Relationships and Experiences 206 93°” Customer Value Measures 208, 9.4 Information Technology and Customer Value 211 Summary 213, Information Technology for Supply Chain Management 215 Case: Backup in the Espresso Lane 215 Case: ERP Brews Instant Success 219 10.1 Introduction 221 102 Goals of Supply Chain Information Technology 223, 103 Standardization 227 104 Information Technology Infrastructure 229 104.1 Interface Devices 229 1042 Communications 230 1043 Databases 231 1044 System Architecture 232 105 Electronic Commerce 234 105.1 Electronic Commerce Levels 236 10.6 Supply Chain Management System Components 239 107 Integrating Supply Chain Information Technology 242 1071 Stages of Development 243 1072 Implementation of ERP and DSS. 245 10.73. “Best-of Breed” versus Single Vendor ERP Solutions 246 Summary 247 Decision-Support Systems for Supply Chain Management 249 Case: Supply Chain Management Smooths Production Flow 249) ML Introduction 251 112. Understanding Decision Support Systems 253 1121 Input Data 254 3122 Analytical Tools 256 1123 Presentation Tools 260 113 Supply Chain Decision Support Systems 264 114 Selecting A Supply Chain DSS 272 Summary 274Contents Appendix A The Computerized Beer Game 275 Ad Introduction 275 A2. The Traditional Beer Game 275 ‘A21 The Difficulties with the Traditional Beer Game 276 A3 TheScenario 277 ‘Aa Playing aRound 278 Ad.l Introducing the Game 279 AA2 Understanding the Screen 281 A43 Playing the Game 281 A44 Other Features 284 AS Options and Settings 285 ‘ASA FileCommands 285 AS2_ OptionsCommands 285 A53. ThePlayCommands 289 AS4 TheGraphs Commands 289 ASS. The Reports Commands 291 Appendix B. The Risk Pool Game 293 BA Introduction 293 B2 TheScenario 293 B3_ Playing Several Rounds 294 B31 Introducing the Game 294 B32 Understanding the Seren 296 B33 Playing theGame 297 B34 Other Features 297 BA Options and Settings 297 B41 FileCommands 297 B42 PlayCommands 298 B43. The Reports Commands 301 Bibliography 303, Index. 309 comms Introduction to Supply Chain Management 1.1 What Is Supply Chain Management? Fiesce competition in today’s global markets, the introduction of products with shot life cycles, and the heightened expectations of customers have forced business enterprises toinvestin, and focusattention on, theirsupply chains. This, together with continuing advances in communications and transportation technologies (eg, mobile communication and overnight delivery) has motivated the continuous evolution ofthe supply chain and of the techniques to manag: Ina typical supply chain, raw materials are procured, items are pros duced at one oF more factories, shipped to warehouses for intermediate Storage, and then shipped to retailers or customers. Consequently, tore- pany's performance and market share. Of course, one can argue that these three examples are associated with companies that are among, the biggest ‘companies in their espective industries; these companies can implement technologies and strategies that very few others can afford, However, in today’s competitive markets, most companies have no choice; they are forced to integrate their supply chain and engage in strategic partnering. ‘This pressure stems from both their customers and their supply chain part- ners. How can integration be achieved successfully? Clearly, information sharing and operational planning are the keys to a Suezessfully integrated. supply chain, But what information should be shared? How should it be used? How does information affect the design and operation of the sup- ply chain? What level of integration is needed within the organization and ‘with external partners? Finally, what types of partnerships can be imple- ‘mented, and which type should be implemented for a given situation? Product Design Effective design plays several critical roles in the supply chain, Most ob- ‘viously certain product designs may increase inventory holding or trans: portation costs relative to other designs, while other designs may facilitate fa shorter manufacturing lead time. Unfortunately, product redesign isof- ten expensive. When is it worthwhile to redesign products soas to reduce logistics costs or supply chain lead times? Ist possible to leverage product Seton 15. Book Obie and Oseraew UL design to compensate for uncertainty in customer demand? Can one guan- tify the amount of savings resulting from such a strategy? What changes Should be made in the supply chain o take advantage ofthe new product design? Finally, new concepts sich as mass customization are increasingly popular. What role docs supply chain management play inthe successful {Implementation ofthese concepts? Information Technology and Decision-Support Systems Information technology isa critical enabler of effective supply chain man- agement. Indeed, muchof the current interest insupply chain management is motivated by the opportunities that appeared due fo the abundance of data and the savings that can be achieved by sophisticated analysis of these data. The primary issue in supply chain management is not whether data ‘can be received, but what data should be transferred; that is, which data, fae significant for supply chain management and which data can safely be ignored? How should the data be analyzed and used? What is the impact of the Internet? What i the role of electronic commerce? What infrastructure is required both internally and between supply chain part- rners? Finally, since information technology and decision-support systems fare both available, can these technologies be viewed as the main tools used to achieve competitive advantage in the market? If they can, then what is preventing others ‘rom using the same technology? Customer Value ‘Customer value is the measure ofa company’s contribution toits customer, ‘based on the entire range of products, services, and intangibles that consti- tute the company’s offerings. In recent years this measure has superseded ‘measures such as quality and customer satisfaction. Obviously, effective supply chain management is critical if a firm wishes to fulfill customer ‘needs and provide value. But what determines customer value in differ- tent industries? How is customer value measured? How is information technology used to enhance customer value in the supply chain? How does supply chain management contribute to customer value? How do ‘emerging trends in customer value, such as development of relationships {and experiences, affect supply chain management? 15 Book Objectives and Overview For many reasons, interest in logistics and supply chain management has grown explosively inthe last few years. This interest has led many compa- hes to analyze their supply chains, In most cases, however, this has been ‘done based on experience and intuition; very few analytical models or de~ sign tools have been used in this process. On the other hand, in the last(Copter Introduction to Suply Cha Management two decades the academic community has developed various models and tools to assist with the management of the supply chain. Unfortunately, the first generation of this technology was not robust or flexible enough to be effectively utilized by industry. This, however, has changed in the last few years. Analysis and insight have improved, and eftective models and decision-support systems have been developed-—but these may not be familiar to industry. This book aims to fill this gap by presenting state-of-the-azt models and solution methods important in the design, conteol, operation, and ‘management of supply chain systems. We intend this book to be useful both as a textbook for MBA-level logistics and supply chain courses, and {asa reference for teachers, consultants, and managers involved in any one of the processes that make up the supply chain, Each chapter includes case studies and numerous examples. In addi tion, each chapter is mostly self-contained, and mathematical and technical sections can be skipped without loss of continuity. Therefore, we believe the bookis accessible to anyone with an interest in some of the many aspects of supply chain management. For example, transportation managers de- 1). This factor depends on the specific application and allows us to assess the amount of space available in the warehouse more accurately. A typical factor used in practice is theee. ThisChapter 2 Legitis Network Congeraton FIGURE 26 Inventory level as a function of tne average Inventory Time factor would be used in the following way. Consider a situation where the annual flow through the warehouse is 1,000 units and the inventory turnover ratios 10.0. This implies that the average inventory level isabout 100 units and, hence, if each unit takes 10 square feet of floor space, the required space for the products is 2,000 square feet. ‘Therefore, the total space required for the warehouse is about 6,000 square feet 2.2.6 Potential Warehouse Locations Itis also important to effectively identify potential locations for new ware- houses. Typically, these locations must satisfy a variety of conditions: + Geographical and infrastructure conditions. + Natural resources and labor availability. + Local industry and tax regulations. + Public interest. As a result, only a limited number of locations would meet all the requirements. These are the potential location sites for the new facilities. 227 Service Level Requirements There are various way’ to define service levelsin thiscontext. For example, we might specify a maximum distance between each customer and the warehouse serving it. This ensures that a warehouse will be able to serve its customers within a reasonable time. Sometimes we must recognize that for some customers, such as those in rural or isolated areas, itis harder to satisfy the same level of service that most other customers receive. In this case, itis often helpful to define the service level as the proportion of customers whose distance to their assigned warehouse is no more than ‘Section 2.3 Made and Data Vato 29 a given distance. For instance, we might require that 95 percent of the ‘customers be situated within 200 miles of the warehouses serving them. 2.2.8 Future Demand Asobserved in Chapter 1, decisionsat the strategic level, which include dis- tribution network design, have along-lasting effect on the firm. In particu- lar, decisions regarding the number, location, and size of warehouses have fan impact on the firm for at least the next three to five years. This implies that changes in customer demand over the next few years should be taken. into account when designing the network. This is most commonly ad- dressed using a scenario-based approach incorporating net present value calculations. For example, various possible scenarios representing a va- riety of possible future demand patterns over the planning horizon can be generated. These scenarios can then be directly incorporated into the ‘model to determine the best distribution strategy. 2.3 Model and Data Validation ‘The previous section documents the difficulties in collecting, tabulating, and Cleaning the data for a network configuration model. Once this is done, how do we ensure that the data and model accurately reflect the netivork design problem? “The process used to address this issue is known as model and data validation. This is typically dane by reconstructing the existing ne-work Configuration using the model and collected data, and comparing the out- putof the modelo existing data. (The importance of validation cannot be overstated. Valuable output of the model configured to duplicate current eperatingconditionsincludesall costs—warehousing, inventory, production, and ransportation—gener- Sted unor the eunrent network configuration. These data can be com pared tothe company’s accounting information, This soften the best way {o identify errors inthe data, problematic assumptions, modeling faws, and so forth (In one project, for example, the transportation cos's cal ulated during the validation process were consistently underestimating the costs suggested by the accounting data. After a careful review ofthe distrbution practices, the consultants concluded thatthe effective truck capacity was only about 30 percent ofthe truck's physical capacity that iss trucks were being sent out with very little lead. Thus, the validation processnot only helped calibrate some of the parameters used inthe model batalso suggested potential improvement inthe utilization of the existing network30 Example 241 | Consider the following distribution system: Caper Lois Network Congurtion {itis often also helpful to make local or small changes in the network configuration to see how the system estimates their impact on costs and service levels.) Specifically, this step involves positing a variety of what- if questions. This includes estimating the impact of closing an existing ‘warehouse on system performance. Or, to give another example, itallows the user to change the flow of material Unrough the existing network and see the changes in the costs, Often, managers have good intuition about ‘what the effect of these small-sale changes on the system should be, so they can more easly identify errors inthe model. Intuition about the effect of radical redesign ofthe entire system is often much less reliable. ‘Validation iscritieal for determining the validity of the model and data, but the process has other benefits, In particular, it helps the user make the connection between the current operations, which were modeled during the validation process, and possible improvement after optimization. 2.4 Solution Techniques ‘Once the data are collected, tabulated, and verified, the next step isto opti- mize the configuration of the logistics network. In practice, two techniques are employed: @ Mathematical optimization techniques that include: + Exact algorithms that are guaranteed to find optimal solutions, that is least-cost solutions, + Heuristic algorithms that find good solutions, not necessarily ‘optimal solutions. 5 Simulation models that provide a mechanism to evaluate specified design alternatives created by the designe: | 24.1 Heuristics and the Need for Exact Algorithms We will start our discussion by considering mathematical optimization techniques. In order to understand the effectiveness of heursticalgorithms and the need for exact algorithms, consider the following example devel- ‘oped by Geoffrion and Van Roy in (41). + Single product + Two plants referred toas plant pl and plant p + Plant y2 hasan annual capacity of 60000 unit Sesion 24 Sluion Ties 31 +The two plants have the same production costs + Two existing warchouses, refered as warehouse wand warehouse w2, have Mdentcal warehouse handling ost. + Thrve markets areas, 1,2, and c3, with demands of 50,000, 100,000 and 5000, respectively, + Table 24 provides distribution cost per unit, For instance, distributing one “unit fom plant pl to warehouse 2 costs $5. TABLEZA Distribution costs per unit Factty Werowe ol a a 8 a ° 4a @ 8 2 3 202 1 2 (Our objective isto find distribution strategy tha {rom the suppliers through the warehouses tothe market areas without violating plant pl production capacity constraint, that satisfies market area demands, and that minimizes total distribution costs. Observe that this problem is sigriicanty teaser to solve than the logistics network configuration problem discassed eariet. ae we assume that falitylestion abt an issue and we merely attempt to find an effective distribution strategy. For this purpose, consider the following two intutie heuristics: Hourstic. Foreachmertet wechoose the cheepest warchouse to source demand. Thus, cl,c2,and c3 would be supplied by 2 Then, for this warehouse choose the cheapest plants thats, distribute 0,000 arts fom p2 and the remaining 140,000 from pl. The total costs 2 5,005 +1 > 100,000 +2 » 50,000, “+2 60,0004 5» 140,000 = 1,120,000 LHeurstic2.Foceach market are, choose the warehouse where the total delivery ets to and from the warehouse are the lowest; that i, consider inbound and foutbound distribution costs Thus, for market area cl, consider the paths pl —> Gh chspl os w2 = cL po» wl cl p> wd —> cl. Among all these lternatives the cheapest ipl > sol + el, 60 choose wl for cl. Using similar Analysis, we choose w2 fr and w2for <3. "Thisimples that warehouse wi delivers total of 50,00 units while warehouse 12 delivers total of150,000units. The best inbound flow pattern stosuppy 50000 from plant to warehouse, supply 7000 units from plant p2 to warehouse 2, tnd supply 9000 from plant pl to warehouse u2. The total cost for this strategy 19320000, ‘Unfortunately the two heuristics described earlier do not produce the best leastcost, strategy. To find the Best distribution strategy, consider the followingChapter? Logistics Network Configuration Optimization model. Indeed, the distribution problem described earlier can be framed as the following linear programming problem! For this purpose, let + x(p1, wb, x, u2), x92, wl) and x(p2, 2) be the flows from the plants to the warehouses. + x(, el), x€01,c2),x(1, <3) be the flows from warehouse w1 to customer zones 1, e2, and + x(02, el), x(a2, 2), x(02, 3) be the flows from warehouse 12 to customer ‘zones-l, 2, and 3, “The linear programming problem we need to solves: Minimize (Ox(pl. 21) 45x(p1, 2) + 4x(92, 11) + 2x2, n2) 4 SeC@T cD) + 4xGw, 2) + Sx(w, 9) + 2e(w2, cl) + 2x(U2, 3) Subject tothe following constraints (92, 01) +292, 02) = 60,000, (pl. s01) + x(92, 001) = x(@1, €1) + 2(01,<2) +x(01,€3) x(02,e1) +2102, 2) + 402,08) x(q, e1) +(02, e1) = 50,000 (wl, €2) + x(u2, 2) x(wl, 3) + x(a, €3) = 50,000 All flows are greater than or equal to ze. ‘One can easly construct an Excel model for this problem and use the Excel linear programming solver to find the optimal strategy. For more information on how to construct the Excel model, see [58]. This strategy is described in Table 25. 100,000 TAOLE2S Optimal Distribution Strategy Fecty Waveowe yl roa a 38 wt 14900 | 0 saoo0 4000 50.000 = © eape0 05000 ° ‘The total cot forthe optimal strategy is $740,000. This par ofthe section requires abasic know edge of near programing, Itcan be skipped without los of contin mene Section 24 Soltion Techies 33 [This example clearly ilustrates the value of optimization based tech- niques. These fol ean determine strategies thal wil sgnfcnlly reduce the tol syste cost (Of course, the logistics network configuration model that we would like to analyze and solve i typically more complex than the simple exam ple desenbed above. One key difference isthe need to establish optimal Focations for warehouses, distribution ceners, and cross-dock faites. Unfortunately, these decisions render linear programming inappropriate tnd rege the use of techie called integer programming. This is trae ipetwtoe linear programing deals with continuous variables whale a de Coion on whether or not fo open a warehouse ata specific city ia binary Variable Oifwedo notopena warehouse in that locaton and Lothe ‘This, the logis network configuration zodel san integer program ming model, Unfortunately integer propramaing models are significantly ‘mor difficult to solve. Te interested reader is eerred to [14] fora discus, Sion on exact algorithms forthe logistics network configuration problem 24.2 Simulation Models and Optimization Techniques ‘The mathematical optimization techniques described earlier have some important limitations. They deal with static models—typically by con- sidering annual, or average, demand—and they do not take into account changes overtime, Simulation-based tools take into account the dynamics of the system and are capable of characterizing system performance for a ‘given design, Thus, itis up to the user to provide the simulation model with ‘a number of design alternatives. ‘This implies that simulation models allow the user to perform a mi- crolevel analysis. Indeed, the simulation model may include, see [46]: 1, Individual ordering patter. 2. Specific inventory policies. 3, Inventory movements inside the warehouse. Unfortunately, simulation models only model a prespecified logistics network design. In other words, given a particular configuration of ware- hhouses, retailers, and so forth, a simulation model can be used to help testimate the costs associated with operating that configuration. Ifa dif- ferent configuration is considered (e.g. a few of the customers are to be served by a different warehouse), the model has to be rerun. As you will see in more detail in Chapter 11, simulation isnot an optimization tool. Itis "useful in characterizing the performance of a particular configuration, but not in determining an effective configuration from a large set of potential configurations. Tn addition, a detailed simulation model that incorporates informa- tion about individual customer ordering patterns, specific inventory andPy 25 Key Features of a Network Configuration DSS < «2. Chap? Loic NetwekCoguratin production policies, daily distribution strategies, and so on may require ‘enormous computational time to achieve a desired level of accuracy in system performance. This implies that typically one can consider very few alternatives using a simulation tool Thus, if system dynamics is not a key issue, a static model is appro- priate and mathematical optimization techniques can be applied. In our ‘experience, this type of model accounts for almost all the network con- figuration models used in practice. When detailed systom dynamics is an {importantissue,itmakes sense to utilize the following two-stage approach, suggested by Hax and Candea [46], which takes advantage of the strengths ‘of both simullation- and optimization-based approaches: 1 Use an optimization model to generate a number of least cont solutions atthe macrolevel, taking into account the most, important cost components. 2. Usea simulation model to evaluate the solutions generated in the first phase ‘Gre of the key requirements of any decision-support system for network “design is Bex. In this context, we define flxity as the ability of the system to incorporate a lange set of preexisting network characteris- tics Indeed, depending on the particular application, a whole spectrum of design options may be appropriate. At one end of this spectrum isthe complete reoptimization ofthe existing network. This means that each ‘warehouse can be either opened or closed anda transportation flows can be redirected, At the other end of the spectrum, it may be necessary 0 incorporate the following features in the optimization model 1. Customer specific seroc level requirements, 2. Exiting warehouses. In most case, warehouses already exist and theirlenses have not yet expited. Therefore, the model should not permit the closing ofthe warehouse 2. Expansion of existing warehouses. Existing warehouses may be expandable. 4. Speci low pater. In a variety of situations, specific ow pattems (eg, from a particular warehouse to a set of customers) Should not be changed, o perhaps more likely a certain ‘manufacturing location doesnot or cannot produce certain SKUs. 5. Wrtouse-oswurchouse fle. In some cases, material may fw fromone warehouse to another warehouse. £ Billofmateras. In some cases, inal assembly is done atthe warehouse and this needs o be captured by the model. Fo this oD 1 vedo Section 25 Key Fatres of Network Confit DSS 35 purpose, the user needs to provide information on the components used to assemble finished goods. Its not enough forthe decision-support system to incorporate all of the features described above. Italso must have the capability to deal with all these issues with little or no reduction in its requirement is directly related to the so-called rabusiness ofthe system. ‘This stipulates thatthe relative quality of the solution generated by the system (Le, cost and service level) should be independent of the specific environment, the variability of the data, or the particular setting. If a particular decision-support system snot robust iis dificult to determine how effective it will be fr a particular problem. | Tt is also essential that the system running time be reasonable. Of course, the term reasonable depends on the particular problem at hand. ‘The running time in seconds of one popular DSS, running on a Pentium- 200 personal computer, i listed in Table 26. The table provides running, time for variety of problems. For each instance the table provides the num ber of aggregated customers (Num Customers), the number of products (Num Products), the number of suppliers (Num Suppliers), the number ‘of potential locations for warehouses (Num Potential) and the number of existing warehouses (Num Exist). Inaddition, the table provides the range of values (Min-Max) ofthe number of new warehouses considered by the DSS. This implies that she number of warehouses considered by the DSS ‘was no less than the Min value and no more than the Max value. In each case, the model includes a service level requirement, define as a limit on the distance between a customer and a warehouse Serving it. This value is provided in the Distance column. The decision-support system found solutions guaranteed to be within 1.0 percent and 0.5 percent ofthe opti- ‘mal solution, and the running times to get each of the solutions is listed, All ofthe problems listed were based on real-world data for a number of Companies in the United States. TaBLER6 Running Times (Seconds) = = = 28 2006 Nan Pract Num Num Nem i umting Rung Suppers Powisl Ene! Mat__ Diente ‘Time 1% Time OS% 2 2 2 2 wom 10 2 woo 184 2» 4 m0 GR 25 30 a 2 0 3mm 2. S0036 SO Chapter? Laie Network Confartion 26 Solving the Bis Corporation Distribution Problem SUMMARY ‘We now go back to the Bis Corporation case described at the beginning of thechapter. An important issuein the analysis is an effective way to predict theimpact of the estimated increase in demand on thenetworkdesign. One ‘way of doing that is to design the network based on the current demand ‘and then evaluate the impact of future demand on the total cost; thats, fix the nuinber and location of warehouses based on current (1997) demands and calculate total cost for 1997, 1998, and 1999. Compare this, for instance, to designing the logistics network based on, say, 1998 demands. Specifically, consider the following three options. Option I: Find optimal network design based on 1997 demands, Option Il: Find optimal network design based on 1998 demands. Option II: Find optimal network design based on 1999 demands, Consider Option I. In this case, we need to calculate total cost for 1998 and 1999, assuming thatthe network configuration does not change. We fox the network configuration according to this design option and evaluate the impact of the increase in demand on total cost. By calculating net present ‘value of total logistics costs for 1997, 1998, and 1999, we get the total cost associated with this design option. Note that during this process, we are able to address capacity issues. For instance, the analysis of 1998 demand. data will allow us toevaluate whether there is enough production capacity ‘ot whether produetion capacity needs to be increased to satisfy customer demands. ‘The same analysis is also applied to the other two design options, and the three net present values are used in making the final decision. [Tithe chapter we examined isu important the design ofthe logics etwork, Ong question ten raced ite efcency of demand agar: tn Shavinfovmation savaiableon individual eet demands, isnot Sear why the arlyso combines groupe of customer and treats therm aa 4 Single ageegrecastomer, Ave have seen, tera vo man fessont focoggiegnng demand dan The fit ithe sizeof the model hat ess fromthe pur data, Indeed he imei takes fo solve a network design problem gown esponenaly asthe numberof usomers creases Even [Foptintztion tone not an sue aggregating demand is imporant be. cate t improves tie accuracy offoreeet nat Thisistruteause Our tility Yo orca’ customer demand atthe acount and product levels is tually poor By conta beeuse ofthe reduction n varity achieved Summary 37 through aggregation, forecast demand isignificantly more accurate atthe aggregated level ‘A second issue often raised in practice is the need for a decsion- support system to optimize the logistics network. The question is whether ‘sophisticated tools required of spreadsheets alone ae sufficient. The Chapter argues that a thorough logstis network analysis should consider Complex transportation cost structures, warehouse sies, manufacturing limitations, inventory tumover rates, inventory cost, and service levels These issues requice the une of eptimization-based decision-support sys tems that can slve langeseale problems efficiently. Fora detailed dscus- sion, see Chapter If(CHAPTER 8 oO Inventory Management and Risk Pooling Case: JAM ELECTRONICS: SERVICE LeveL CRISIS JAM Electronics is a Korean manufacturer of products such as industrial relays. The company has five manufacturing facilities in diferent countries in the Far East ‘with headquarters in Seou), South Korea JAM USA is a subsidiary of JAM Electronics and was established in the United States in 1978, The US. Subsidiary provides distribution and servee functions in this county. [thas a central warehouse in Chicago thet serves to types of cus- tomers: distributors and original equipment manufacturers (OEMs). Distributors typically keep inventory of JAM’ products and supply them to their clients as needed. OEMs use JAMS products to produce different types of items such as ‘automatic gazage door openers. JAM Electronics produces about 2,500 diferent products all of them manufac- tured in the Far East Finished products are stored ina central warehouse in Korea and are shipped from there to different countries. In particulay items sold in the United States are transported by ship tothe warehouse in Chicago. Inzecent years, JAMUSA has seen asignificantinereae in competition and huge pressure from its clients and distributors to improve service evelsand reduce costs Unfortunately, as Al Rub, theinventory manager, points out, "the ervice level right ‘now jsatan all-time low. Only about 70 percent ofall ondrs are delivered on tim. ‘On the other hand, inventory keeps pling up, mostly of products tat are not in demand.” In a recent meeting with Chan Moon, the president of JAM USA; Ken Hall, ‘ice president and general manager; and Li Chan, a representative of the Korean headquarters, Al pointed out several reasons for the low service level. 1 Difeuty forecasting customer demand. Indeed, changes in the economy, customer behavior, and other factors have a major impact on demand, and these are quite difficult to predict, Sours: JAM Flector acolo. The mall ahs ct oe base one Speer wahscer compan: 39ca EY Chapter’ Inortory Management and Risk Pooing 2. Long lead tne i the supply cain. Atypical oder placed by the warehouse in linois will arrive in about six to seven weeks. There ar two primary reasons for the long lead time. First there is a week of processing time in the central distribution center in Korea; second, the ocean transit time is very long, ‘3. The large numberof SKUs handled by JAM USA. As pointed out earlie, JAM. USA distributes about 2,50 diferent products, ranging from small relays to large programmable controllers, to its customers. 4 The loo priority given the U.S. subsidiary by headquarters Seoul. An order that arrives at the Korean headquarter from Japanese or Korean clients typically receives a higher priority over an order eceived from the United States, Thus, the lead time tothe United Stats is ocasionally much longer than seven weeks. Toillustrate the difculty in forecasting customer demand, Al provided a graph that showed monthly demand information fr item xx-1534, relay product used In the manufacture of refrigerators (sce Figure 3.1). As the graph demonstrates, variability in customer demandis very high: demand changesfrom monthtomonth and itis difficult o predict what customer demand is going to be even when so- phisticated forecasting techniques are used. ‘Ken, the general manager, was very critical of Al's analysis. He pointed out that if long lead time isan issue then JAM USA needs to reduce it, perhaps by using air carriers, which would probably reduce lead time to about two weeks. Ken, however, was unsure ofthe impact ofthis change on the overall supply chain, He pointed out that “transportation costs are certainly going to increase, but where fre the savings?” “The meeting ended with a decision to establish a task force headed by Ken to address the service level crisis. 'As Al left the meeting, he could not help but think that many companies must bbe facing similar problems-high demand variability, long lead times, unreliable supply processes, and a large number of SKUs. What is the competition doing to cope with these problems? FIGURE 3.1 ‘Monthly demand fr ite xx-1534 Demand Ain trousands) g 8 ° : sw sce 8Ml0 775 a wy fsbo sats swe a1 De Jan Fab Ma wo Bae 10840 488 Months, 197-38 Seton 31 Irtodection 4 By the end of this chapter, you should be able to understand the following + How afi can cope with huge variability in customer demand. + What the relationship is between service and inventory levels + What the impact of lead time and lead time variability has on inventory levels. + What an efecive inventory management policy is. 3.1 Introduction ‘The importance of inventory management and the need for the coordina- tion of inventory decisions and transportation policies has been evident for a long time. Unfortunately, managing inventory in complex supply chains is typically quite difficult and may Rave a significant impact on the customer service level and supply chain systemwide cost. ‘As we discussed in Chapter 1, a typical supply chain consist of suppli- cers and manufacturers, who convert raw materials inte finished products, and distribution centers and warehouses, from which finished products are distributed to customers. This implies that inventory appears in the supply chain in several forms: + Raw material inventory. + Work-in-process (WIP) inventory. + Finished product inventory. Each of these needs its own inventory control mechanism. The difficulty in determining these mechanisms is that efficlent production, distribution, and inventory control strategies that reduce systemwide costsand improve service levels must take into account the interactions ofthe various levels in the supply chain. Although determining these inventory control mech- anisms may be challenging, the benefits can be enormous: Example3.1 | General Motors (GM) has one of the largest production and distribution networks Inthe world. In 1984GM sdistribution network consisted of 20,000 supplier plants, 133 parts plants, 1 assembly plants, and 11,000 dealers. Freight transportation ‘costs were about $4.1 billion with 60 percent for material shipments. in addition, {GM inventory was valued at $7.4bilion, of which 70 percent was WIP and the rest ‘as finished Vehicles. GM has implemented a decision tool capable of reducing, the combined corporate cost of inventory and transportation, Indeed, by adjusting shipment sizes (Le, inventory policy)and routes (Le, transportation strategy), costs ‘could be reduced by about 6 percent annually [10}