SUNIL Chopra_Meindl_SCM

138 Chapter 5 • Network Design in the Supply Chain 8. Hot&Cold and CaldoFreddo are two European manufacturers of home appliances that have merged. Hot&Cold has plants in markets: Northeast, Southeast, Midwest, South, and West. France, Germany, and Finland, whereas CaldoFreddo has Each PC sells for $1,000. The firm anticipates a 50 percent plants in the United Kingdom and Italy. The European market growth in demand (in each region) this year (after which is divided into four regions: North, East, West, and South. demand will stabilize) and wants to build a plant with a Plant capacities (millions of units per year), annual fixed costs capacity of 1.5 million units per year to accommodate the (millions of euros per year), regional demand (millions of growth. Potential sites being considered are in North units), and variable production and shipping costs (euros per Carolina and California. Currently the firm pays federal, unit) are as shown in Table 5-14. state, and local taxes on the income from each plant. Each appliance sells for an average price of 300 euros. All Federal taxes are 20 percent of income, and all state and plants are currently treated as profit centers, and the company local taxes are 7 percent of income in each state. North pays taxes separately for each plant. Tax rates in the various Carolina has offered to reduce taxes for the next 10 years countries are as follows: France, 0.25; Germany, 0.25; from 7 percent to 2 percent. Blue Computers would like to Finland, 0.3; UK, 0.2; and Italy, 0.35. take the tax break into consideration when planning its network. Consider income over the next 10 years in your a. Before the merger, what is the optimal network for each of analysis. Assume that all costs remain unchanged over the the two firms if their goal is to minimize costs? What is the 10 years. Use a discount factor of 0.1 for your analysis. optimal network if the goal is to maximize after-tax profits? Annual fixed costs, production and shipping costs per unit, and current regional demand (before the 50 percent growth) b. After the merger, what is the minimum cost configuration are shown in Table 5-13. if none of the plants is shut down? What is the configura- tion that maximizes after-tax profits if none of the plants a. If Blue Computers sets an objective of minimizing total is shut down? fixed and variable costs, where should it build the new plant? How should the network be structured? c. After the merger, what is the minimum cost configuration if plants can be shut down (assume that a shutdown saves b. If Blue Computers sets an objective of maximizing after- 100 percent of the annual fixed cost of the plant)? What is tax profits, where should it build the new plant? How the configuration that maximizes after-tax profits? should the network be structured? Table 5-13 Variable Production and Shipping Costs for Blue Computers Variable Production and Shipping Cost ($/Unit) Kentucky Northeast Southeast Midwest South West Annual Fixed Pennsylvania Cost (Million $) N. Carolina 185 180 175 175 200 California 170 190 180 200 220 150 Demand (’000 units/month) 180 180 185 185 215 200 220 220 195 195 175 150 700 400 400 300 600 150 Table 5-14 Capacity, Cost, and Demand Data for Hot & Cold and CaldoFreddo Variable Production and Shipping Costs North East South West Capacity Annual Fixed Cost Hot&Cold France 100 110 105 100 50 105 50 1,000 Germany 95 105 110 110 40 1,000 35 Finland 90 100 115 50 850 90 60 Demand 30 20 20 115 1,000 20 1,150 CaldoFreddo U.K. 105 120 110 Italy 110 105 90 Demand 15 20 30

Chapter 5 • Network Design in the Supply Chain 139 Bibliography Anderson, Kenneth E., Daniel P. Murphy, and James M. Reeve. International Journal of Production Economics (2001) 69: “Smart Tax Planning for Supply Chain Facilities.” Supply 193–204. Chain Management Review (November–December 2002): MacCormack, Alan D., Lawrence J. Newman III, and Donald B. 46–52. Rosenfield. “The New Dynamics of Global Manufacturing Site Location.” Sloan Management Review (Summer 1994): Ballou, Ronald H. Business Logistics Management. Upper Saddle 69–79. River, NJ: Prentice Hall, 1999. Mentzer, Joseph. “Seven Keys to Facility Location.” Supply Chain Management Review (May–June 2008): 25–31. Bovet, David. “Good Time to Rethink European Distribution.” Murphy, Sean. “Will Sourcing Come Closer to Home?” Supply Supply Chain Management Review (July–August 2010): 6–7. Chain Management Review (September 2008): 33–37. Note on Facility Location. Harvard Business School Note Daskin, Mark S. Network and Discrete Location. New York: 9–689–059, 1989. Wiley, 1995. Robeson, James F., and William C. Copacino, eds. The Logistics Handbook. New York: Free Press, 1994. Drezner, Z., and H. Hamacher. Facility Location: Applications Tayur, Sridhar, Ram Ganeshan, and Michael Magazine, eds. and Theory. Berlin: Springer Verlag, 2004. Quantitative Models for Supply Chain Management. Boston: Kluwer Academic Publishers, 1999. Ferdows, Kasra. “Making the Most of Foreign Factories.” Tirole, Jean. The Theory of Industrial Organization. Cambridge, Harvard Business Review (March–April 1997): 73–88. MA: The MIT Press, 1997. Harding, Charles F. “Quantifying Abstract Factors in Facility- Location Decisions.” Industrial Development (May–June 1988): 24. Korpela, Jukka, Antti Lehmusvaara, and Markku Tuominen. “Customer Service Based Design of the Supply Chain.” CASE STUDY Managing Growth at SportStuff.com In December 2008, Sanjay Gupta and his management jackets from families and any surplus equipment from team were busy evaluating the performance at manufacturers and retailers and sell these over the Internet. SportStuff.com over the previous year. Demand had The idea was well received in the marketplace, demand grown by 80 percent. This growth, however, was a grew rapidly, and by the end of 2004, the company had mixed blessing. The venture capitalists supporting the sales of $0.8 million. By this time, a variety of new company were very pleased with the growth in sales and and used products were being sold, and the company the resulting increase in revenue. Sanjay and his team, received significant venture capital support. however, could clearly see that costs would grow faster than revenues if demand continued to grow and the In June 2004, Sanjay leased part of a warehouse supply chain network was not redesigned. They decided in the outskirts of St. Louis to manage the large to analyze the performance of the current network to see amount of product being sold. Suppliers sent their how it could be redesigned to best cope with the rapid product to the warehouse. Customer orders were growth anticipated over the next three years. packed and shipped by UPS from there. As demand grew, SportStuff.com leased more space within the SportStuff.com warehouse. By 2007, SportStuff.com leased the entire warehouse and orders were being shipped to Sanjay Gupta founded SportStuff.com in 2004 with a mis- customers all over the United States. Management sion of supplying parents with more affordable sports divided the United States into six customer zones for equipment for their children. Parents complained about planning purposes. Demand from each customer zone having to discard expensive skates, skis, jackets, and shoes in 2007 was as shown in Table 5-15. Sanjay estimated because children outgrew them rapidly. Sanjay’s initial that the next three years would see a growth rate of plan was for the company to purchase used equipment and about 80 percent per year, after which demand would level off. (continued)

140 Chapter 5 • Network Design in the Supply Chain (continued) Table 5-15 Regional Demand at SportStuff.com for 2007 Zone Demand in 2007 Zone Demand in 2007 Northwest 320,000 Lower Midwest 220,000 Southwest 200,000 Northeast 350,000 Upper Midwest 160,000 Southeast 175,000 The Network Options aggregating throughput through a few facilities reduces the inventory held as compared with disaggregating through- Sanjay and his management team could see that they put through many facilities. Thus, a warehouse handling 1 needed more warehouse space to cope with the antici- million units per year incurred an inventory holding cost of pated growth. One option was to lease more warehouse $600,000 in the course of the year. If your version of Excel space in St. Louis itself. Other options included leasing has problems solving the nonlinear objective function, use warehouses all over the country. Leasing a warehouse the following inventory costs: involved fixed costs based on the size of the warehouse and variable costs that depended on the quantity shipped Range of F Inventory Cost through the warehouse. Four potential locations for warehouses were identified in Denver, Seattle, Atlanta, 0–2 million $250,000Y ϩ 0.310F and Philadelphia. Warehouses leased could be either 2–4 million $530,000Y ϩ 0.170F small (about 100,000 sq. ft.) or large (200,000 sq. ft.). 4–6 million $678,000Y ϩ 0.133F Small warehouses could handle a flow of up to 2 million More than 6 million $798,000Y ϩ 0.113F units per year, whereas large warehouses could handle a flow of up to 4 million units per year. The current ware- If you can handle only a single linear inventory house in St. Louis was small. The fixed and variable cost, you should use $475,000Y ϩ 0.165F. For each costs of small and large warehouses in different loca- facility, Y=1 if the facility is used, 0 otherwise. tions are shown in Table 5-16. SportStuff.com charged a flat fee of $3 per shipment Sanjay estimated that the inventory holding costs at sent to a customer. An average customer order contained a warehouse (excluding warehouse expense) was about four units. SportStuff.com in turn contracted with UPS to $600 1F, where F is the number of units flowing through handle all its outbound shipments. UPS charges were the warehouse per year. This relationship is based on the based on both the origin and the destination of the theoretical observation that the inventory held at a facility shipment and are shown in Table 5-17. Management (not across the network) is proportional to the square root of the throughput through the facility. As a result, Table 5-16 Fixed and Variable Costs of Potential Warehouses Small Warehouse Large Warehouse Location Fixed Cost Variable Cost Fixed Cost Variable Cost ($/year) ($/Unit Flow) ($/year) ($/Unit Flow) Seattle 300,000 0.20 500,000 0.20 Denver 250,000 0.20 420,000 0.20 St. Louis 220,000 0.20 375,000 0.20 Atlanta 220,000 0.20 375,000 0.20 Philadelphia 240,000 0.20 400,000 0.20

Chapter 5 • Network Design in the Supply Chain 141 estimated that inbound transportation costs for shipments 2. What supply chain network configuration do you recom- from suppliers were likely to remain unchanged, no matter mend for SportStuff.com? Why? what warehouse configuration was selected. 3. How would your recommendation change if transportation Questions costs were twice those shown in Table 5-17? 1. What is the cost SportStuff.com incurs if all warehouses leased are in St. Louis? Table 5-17 UPS Charges per Shipment (Four Units) Northwest Southwest Upper Midwest Lower Midwest Northeast Southeast Seattle $2.00 $2.50 $3.50 $4.00 $5.00 $5.50 Denver $2.50 $2.50 $2.50 $3.00 $4.00 $4.50 St. Louis $3.50 $3.50 $2.50 $2.50 $3.00 $3.50 Atlanta $4.00 $4.00 $3.00 $2.50 $3.00 $2.50 Philadelphia $4.50 $5.00 $3.00 $3.50 $2.50 $4.00 CASE STUDY Designing the Production Network at CoolWipes Matt O’Grady, vice president of supply chain at CoolWipes currently had one factory in Chicago that produced both thought that his current production and distribution net- products for the entire country. The wipes line in the work was not appropriate given the significant increase in Chicago facility had a capacity of 5 million units, an transportation costs over the past few years. Compared annualized fixed cost of $5 million a year, and a variable to when the company had set up its production facility in cost of $10 per unit. The ointment line in the Chicago Chicago, transportation costs had increased by a factor of facility had a capacity of 1 million units, an annualized more than four and were expected to continue growing in fixed cost of $1.5 million a year, and a variable cost of the next few years. A quick decision on building one or $20 per unit. The current transportation costs per unit more new plants could save the company significant (for both wipes and ointment) are shown in Table 5-19. amounts in transportation expense in the future. CoolWipes New Network Options CoolWipes was founded in the late 1980s and produced Matt had identified Princeton, New Jersey; Atlanta; and baby wipes and diaper ointment. Demand for the two Los Angeles as potential sites for new plants. Each new products was as shown in Table 5-18. The company plant could have a wipes line, an ointment line, or both. A new wipes line had a capacity of 2 million units, an Table 5-18 Regional Demand at CoolWipes (in ’000s) Wipes Ointment Wipes Ointment Demand Demand Zone Demand Demand Zone 65 Northwest 500 50 Lower Midwest 800 120 Southwest 700 70 Upper Midwest 900 90 Northeast 1,000 120 Southeast 600 (continued)

142 Chapter 5 • Network Design in the Supply Chain (continued) Table 5-19 Transportation Costs per Unit Upper Lower Northeast Southeast Northwest Southwest Midwest Midwest $5.76 $5.96 Chicago $6.32 $6.32 $3.68 $4.04 $3.68 $4.08 Princeton $6.60 $6.60 $5.76 $5.92 $4.04 $3.64 Atlanta $6.72 $6.48 $5.92 $4.08 $6.72 $6.60 Los Angeles $4.36 $3.68 $6.32 $6.32 annual fixed cost of $2.2 million, and a variable produc- included? Assume that the Chicago plant will be main- tion cost of $10 per unit. A new ointment line had a tained at its current capacity but could be run at lower capacity of 1 million units, an annual fixed cost of $1.5 utilization. Would your decision be different if transporta- million, and a variable cost of $20 per unit. The current tion costs are half of their current value? What if they were transportation costs per unit are shown in Table 5-19. double their current value? Matt had to decide whether to build a new plant and if 3. If Matt could design a new network from scratch (assume so, which production lines to put into the new plant. he did not have the Chicago plant but could build it at the cost and capacity specified in the case), what production Questions network would you recommend? Assume that any new plants built besides Chicago would be at the cost and 1. What is the annual cost of serving the entire nation from capacity specified under the new network options. Would Chicago? your decision be different if transportation costs were half of their current value? What if they were double their 2. Do you recommend adding any plant(s)? If so, where current value? should the plant(s) be built and what lines should be

6 {{{ Designing Global Supply Chain Networks LEARNING OBJECTIVES After reading this chapter, you will be able to 1. Identify factors that need to be included in total cost when making global sourcing decisions. 2. Define uncertainties that are particularly relevant when designing global supply chains. 3. Explain different strategies that may be used to mitigate risk in global supply chains. 4. Understand decision tree methodologies used to evaluate supply chain design decisions under uncertainty. Globalization has offered tremendous opportunity as well as increased risk in the development of supply chains. High-performance supply chains such as Nokia and Zara have taken full advantage of globalization. In contrast, several supply chains have found themselves unprepared for the increased risk that has accompanied globalization. As a result, managers must account for both opportunities and uncertainties over the long term when designing a global supply chain network. In this chapter, we identify sources of risk for global supply chains, discuss risk mitigation strategies, detail the methodologies used to evaluate network design decisions under uncertainty, and show how they improve global supply chain decisions. 6.1 THE IMPACT OF GLOBALIZATION ON SUPPLY CHAIN NETWORKS Globalization offers companies opportunities to simultaneously grow revenues and decrease costs. In its 2008 annual report, P&G reported that more than a third of the company sales growth was from developing markets with a profit margin that was comparable to developed market margins. By 2010, sales for the company in developing markets represented almost 34 percent of global sales. Similarly, Nokia’s two largest global markets in 2009, in terms of net sales, were China and India. Sales in these two countries represented almost 21.5 percent, and sales in the BRIC countries (Brazil, Russia, India, and China) represented more than 28 percent of Nokia’s global sales in 2009. Clearly, globalization has offered both P&G and Nokia a significant revenue enhancement opportunity. Apparel and consumer electronics are two areas in which globalization has offered significant cost reduction opportunities. Consumer electronics focuses on small, lightweight, high-value items that are relatively easy and inexpensive to ship. Companies have exploited large economies of scale by consolidating production of standardized electronics components in a single location for use in multiple products across the globe. Contract manufacturers like Foxconn and Flextronics have become giants with facilities in low-cost countries. Apparel manufacture has high labor content and the product is relatively lightweight and cost effective to transport. Companies have exploited globalization by shifting much apparel manufacturing to low-labor-cost countries, especially China. In the first half of 2009, about 33 percent of U.S. imports of apparel were from China. The net result is that both industries have benefited tremendously from cost reduction as a result of globalization. 143

144 Chapter 6 • Designing Global Supply Chain Networks One must keep in mind, however, that the opportunities from globalization are often accompanied by significant additional risk. In a survey conducted by the consulting company Accenture in 2006, more than 50 percent of the executives surveyed believed that supply chain risk had increased as a result of their global operations strategy. For example, in 2005, hurricane damage to 40,000 acres of plantations decreased Dole’s global banana production by about 25 percent. Component shortage when Sony introduced the PlayStation 3 game console hurt revenues and Sony’s stock price. The ability to incorporate suitable risk mitigation into supply chain design has often been the difference between global supply chains that have succeeded and those that have not. The Accenture survey categorized risk in global supply chains as shown in Table 6-1 and asked respondents to indicate the factors that affected them. More than a third of the respondents were impacted by natural disasters, volatility of fuel prices, and the performance of supply chain partners. Crude oil spot price and exchange rate fluctuations in 2008 illustrate the extreme volatility that global supply chains must deal with. Crude started 2008 at about $90 per barrel, peaked in July at more than $140 per barrel, and plummeted to below $40 per barrel in December. The euro started 2008 at about $1.47, peaked in July at almost $1.60, dropped to about $1.25 at the end of October, and then rose back to $1.46 toward the end of December. One can only imagine the havoc such fluctuation played on supply chain performance in 2008! Similar fluctuations in exchange rates and crude prices have continued since then. The only constant in global supply chain management seems to be uncertainty. Over the life of a supply chain network, a company experiences fluctuations in demand, prices, exchange rates, and the competitive environment. A decision that looks good under the current environment may be quite poor if the situation changes. Between 2000 and 2008, the euro fluctuated from a low of $0.84 to a high of almost $1.60. Clearly, supply chains optimized to $0.84 per euro would have difficulty performing well when the euro reached $1.60. Uncertainty of demand and price drives the value of building flexible production capacity at a plant. If price and demand do vary over time in a global network, flexible production capacity can be reconfigured to maximize profits in the new environment. Between 2007 and 2008, auto sales in the United States dropped by more than 30 percent. Whereas all Table 6-1 Results of Accenture Survey on Sources of Risk That Impact Global Supply Chain Performance Risk Factors Percentage of Supply Chains Impacted Natural disasters 35 Shortage of skilled resources 24 Geopolitical uncertainty 20 Terrorist infiltration of cargo 13 Volatility of fuel prices 37 Currency fluctuation 29 Port operations/custom delays 23 Customer/consumer preference shifts 23 Performance of supply chain partners 38 Logistics capacity/complexity 33 Forecasting/planning accuracy 30 Supplier planning/communication issues 27 Inflexible supply chain technology 21 Source: Adapted from “Integration: The Key to Global Success.” Jaume Ferre, Johann Karlberg, and Jamie Hintlian, Supply Chain Management Review (March 2007): 24–30.

Chapter 6 • Designing Global Supply Chain Networks 145 vehicle categories were affected, the drop in SUV sales was much more significant than the drop in sales of small cars and hybrids. SUV sales dropped by almost 35 percent, but small car sales actually increased by about 1 percent. Honda dealt with this fluctuation more effectively than its competitors because its plants were flexible enough to produce both vehicle types. This flexibility to produce both SUVs and cars in the same facility kept Honda plants operating at reasonably high levels of utilization. In contrast, companies with plants dedicated to SUV production had no option but to leave a lot of idle capacity. In the late 1990s, Toyota made its global assembly plants more flexible so that each plant could supply multiple markets. One of the main benefits of this flexibility is that it allows Toyota to react to fluctuations in demand, exchange rates, and local prices by altering production to maximize profits. Thus, supply, demand, and financial uncertainty must be considered when making global network design decisions. 6.2 THE OFFSHORING DECISION: TOTAL COST This importance of comparative advantage in global supply chains was recognized by Adam Smith in The Wealth of Nations when he stated, “If a foreign country can supply us with a commodity cheaper than we ourselves can make it, better buy it of them with some part of the produce of our own industry, employed in a way in which we have some advantage.” Cost reduction by moving production to low-cost countries is typically mentioned among the top reasons for a supply chain to become global. The challenge, however, is to quantify the benefits (or comparative advantage) of offshore production along with the reasons for this comparative advantage. Whereas many companies have taken advantage of cost reduction through offshoring, others have found the benefits of offshoring to low-cost countries to be far less than anticipated and in some cases nonexistent. The increases in transportation costs between 2000 and 2011 have had a significant negative impact on the perceived benefits of offshoring. Companies have failed to gain from offshoring for two primary reasons—(1) focusing exclusively on unit cost rather than total cost when making the offshoring decision and (2) ignoring critical risk factors. In this section, we focus on dimensions along which total landed cost needs to be evaluated when making an offshoring decision. The significant dimensions of total cost can be identified by focusing on the complete sourcing process when offshoring. It is important to keep in mind that a global supply chain with offshoring increases the length and duration of information, product, and cash flows. As a result, the complexity and cost of managing the supply chain can be significantly higher than anticipated. Table 6-2 identifies dimensions along which each of the three flows should be analyzed for the impact on cost and product availability. Ferreira and Prokopets (2009) suggest that companies should evaluate the impact of off-shoring on the following key elements of total cost: 1. Supplier price: should link to costs from direct materials, direct labor, indirect labor, management, overhead, capital amortization, local taxes, manufacturing costs, and local regulatory compliance costs. 2. Terms: costs are affected by net payment terms and any volume discounts. 3. Delivery costs: include in-country transportation, ocean/air freight, destination transport, and packaging. 4. Inventory and warehousing: include in-plant inventories, in-plant handling, plant warehouse costs, supply chain inventories, and supply chain warehousing costs. 5. Cost of quality: includes cost of validation, cost of performance drop due to poorer quality, and cost of incremental remedies to combat quality drop. 6. Customer duties, value added-taxes, local tax incentives 7. Cost of risk, procurement staff, broker fees, infrastructure (IT and facilities), and tooling and mold costs. 8. Exchange rate trends and their impact on cost.

146 Chapter 6 • Designing Global Supply Chain Networks Table 6-2 Dimensions to Consider When Evaluating Total Cost from Offshoring Performance Dimension Activity Impacting Performance Impact of Offshoring Order communication Order placement More difficult communication Supply chain visibility Scheduling and expediting Poorer visibility Raw material costs Sourcing of raw material Could go either way depending on raw Unit cost Production, quality (production and material sourcing Freight costs transportation) Labor/fixed costs decrease; Taxes and tariffs Transportation modes and quantity quality may suffer Supply lead time Border crossing Higher freight costs Order communication, supplier Could go either way On-time delivery/lead production scheduling, production time, Lead time increase results time uncertainty customs, transportation, receiving in poorer forecasts and Production, quality, customs, higher inventories Minimum order quantity transportation, receiving Poorer on-time delivery Product returns and increased uncertainty Inventories Production, transportation resulting in higher Working capital inventory and lower Hidden costs Quality product availability Stock-outs Lead times, inventory in transit and Larger minimum quantities production increase inventory Inventories and financial reconciliation Increased returns likely Order communication, invoicing Increase errors, managing exchange rate risk Ordering, production, transportation Increase with poorer visibility Higher hidden costs Increase It is important to quantify these factors carefully when making the offshoring decision and to track them over time. As Table 6-2 indicates, unit cost reduction from low labor and fixed costs along with possible tax advantages are likely to be the major benefit from offshoring, with almost every other factor getting worse. In some instances, the substitution of labor for capital can provide a benefit when offshoring. For example, auto and auto parts plants in India are designed with much greater labor content than similar manufacturing in developed countries to lower fixed costs. The benefit of lower labor cost, however, is unlikely to be significant for a manufactured product if labor cost is a small fraction of total cost. It is also the case that in several low-cost countries such as China and India, labor costs have escalated significantly. As mentioned by Goel et al. (2008), wage inflation in China averaged 19 percent in dollar terms between 2003 and 2008 compared to around 3 percent in the United States. During the same period, transportation costs increased by a significant amount (ocean freight costs increased 135 percent between 2005 and 2008) and the Chinese yuan strengthened relative to the dollar (by about 18 percent between 2005 and 2008). The net result was that offshoring manufactured products from the United States to China looked much less attractive in 2008 than in 2003.

Chapter 6 • Designing Global Supply Chain Networks 147 In general, offshoring to low-cost countries is likely to be most attractive for products with high labor content, large production volumes, relatively low variety, and low transporta- tion costs relative to product value. For example, a company producing a large variety of pumps is likely to find that offshoring the production of castings for a common part across many pumps is likely to be much more attractive than the offshoring of highly specialized engineered parts. Given that global sourcing tends to increase transportation costs, it is important to focus on reducing transportation content for successful global sourcing. Suitably designed components can facilitate much greater density when transporting products. IKEA has designed modular products that are assembled by the customer. This allows the modules to be shipped flat in high density, lowering transportation costs. Similarly, Nissan redesigned its globally sourced components so that they can be packed tighter when shipping. The use of supplier hubs can be effective if several components are being sourced globally from different locations. Many manufacturers have created supplier hubs in Asia that are fed by each of their Asian suppliers. This allows for a consolidated shipment to be sent from the hub rather than several smaller shipments from each supplier. More sophisticated flexible policies that allow for direct shipping from the supplier when volumes are high, coupled with consolidated shipping through a hub when volumes are low, can be effective in lowering transportation content. It is also important to perform a careful review of the production process to decide which parts are to be offshored. For example, a small American jewelry manufacturer wanted to offshore manufacturing for a piece of jewelry to Hong Kong. Raw material in the form of gold sheet was sourced in the United States. The first step in the manufacturing process was the stamping of the gold sheet into a suitable-sized blank. This process generated about 40 percent waste, which could be recycled to produce more gold sheet. The manufacturer faced the choice of stamping in the United States or Hong Kong. Stamping in Hong Kong would incur lower labor cost but higher transportation cost and would require more working capital because of the delay before the waste gold could be recycled. A careful analysis indicated that it was cheaper for the stamping tools to be installed at the gold sheet supplier in the United States. Stamping at the gold sheet supplier reduced transportation cost because only usable material was shipped to Hong Kong. More importantly, this decision reduced working capital requirement because the waste gold during stamping was recycled within two days. One of the biggest challenges with offshoring is the increased risk and its potential impact on cost. This challenge gets exacerbated if a company uses an offshore location that is primarily targeting low costs to absorb all the uncertainties in its supply chain. In such a context, it is often much more effective to use a combination of an offshore facility that is given predictable, high-volume work along with an onshore or near-shore facility that is specifically designed to handle most of the fluctuation. Companies solely using an offshore facility often find themselves carrying extra inventory and resorting to air freight because of the long and variable lead times. The presence of a flexible onshore facility that absorbs all the variation can often lower total landed cost by eliminating expensive outbound freight and significantly reducing the amount of inventory carried in the supply chain. Key Point It is critical that offshoring decisions be made accounting for total cost. Offshoring typically lowers labor and fixed costs but increases risk, freight costs, and working capital. Before offshoring, product design and process design should be carefully evaluated to identify steps that may lower freight content and the need for working capital. Including an onshore option can lower the cost associated with covering risk from an offshore facility.

148 Chapter 6 • Designing Global Supply Chain Networks 6.3 RISK MANAGEMENT IN GLOBAL SUPPLY CHAINS Global supply chains today are subject to more risk factors than localized supply chains of the past. These risks include supply disruption, supply delays, demand fluctuations, price fluctuations, and exchange-rate fluctuations. As was evident in the financial crisis of 2008, underestimating risks in global supply chains and not having suitable mitigation strategies in place can result in painful outcomes. For example, contamination at one of the two suppliers of flu vaccine to the United States led to a severe shortage at the beginning of the 2004 flu season. This shortage led to rationing in most states and severe price gouging in some cases. Similarly, the significant strengthening of the euro in 2008 hurt firms that had most of their supply sources located in Western Europe. In another instance, failure to buffer supply uncertainty with sufficient inventory resulted in high costs rather than savings. An automotive component manufacturer had hoped to save $4 to $5 million a year by sourcing from Asia instead of Mexico. As a result of port congestion in Los Angeles–Long Beach, the company had to charter aircraft to fly the parts in from Asia because it did not have sufficient inventory to cover the delays. A charter that would have cost $20,000 per aircraft from Mexico ended up costing the company $750,000. The anticipated savings turned into a $20 million loss. It is thus critical for global supply chains to be aware of the relevant risk factors and build in suitable mitigation strategies. Table 6-3 contains a categorization of supply chain risks and their drivers that must be considered during network design. Table 6-3 Supply Chain Risks to Be Considered During Network Design Category Risk Drivers Disruptions Delays Natural disaster, war, terrorism Systems risk Labor disputes Forecast risk Supplier bankruptcy Intellectual property risk Procurement risk High capacity utilization at supply source Inflexibility of supply source Receivables risk Poor quality or yield at supply source Inventory risk Information infrastructure breakdown Capacity risk System integration or extent of systems being networked Inaccurate forecasts due to long lead times, seasonality, product variety, short life cycles, small customer base Information distortion Vertical integration of supply chain Global outsourcing and markets Exchange-rate risk Price of inputs Fraction purchased from a single source Industry-wide capacity utilization Number of customers Financial strength of customers Rate of product obsolescence Inventory holding cost Product value Demand and supply uncertainty Cost of capacity Capacity flexibility Source: Adapted from “Managing Risk to Avoid Supply Chain Breakdown.” Sunil Chopra and Manmohan S. Sodhi, Sloan Management Review (Fall 2004): 53–61.

Chapter 6 • Designing Global Supply Chain Networks 149 Good network design can play a significant role in mitigating supply chain risk. For instance, having multiple suppliers mitigates the risk of disruption from any one supply source. An excellent example is the difference in impact on Nokia and Ericsson when a plant owned by Royal Philips Electronics, located in Albuquerque, New Mexico, caught fire in March 2000. Nokia adjusted to the disruption quickly, using several other supply plants in its network. In con- trast, Ericsson had no backup source in its network and was unable to react. Ericsson estimated that it lost revenues of $400 million as a result. Similarly, having flexible capacity mitigates the risks of global demand, price, and exchange-rate fluctuations. For example, Hino Trucks uses flexible capacity at its plants to change production levels for different products by shifting work- force between lines. As a result, the company keeps a constant workforce in the plant even though the production at each line varies to best match supply and demand. As illustrated by these examples, designing mitigation strategies into the network significantly improves a supply chain’s ability to deal with risk. Every mitigation strategy, however, comes at a price and may increase other risks. For example, increasing inventory mitigates the risk of delays but increases the risk of obsolescence. Acquiring multiple suppliers mitigates the risk of disruption but increases costs because each supplier may have difficulty achieving economies of scale. Thus, it is important to develop tailored mitigation strategies during network design that achieve a good balance between the amount of risk mitigated and the increase in cost. Some tailored mitigation strategies are outlined in Table 6-4. Most of these strategies are discussed in greater detail later in the book. Global supply chains should generally use a combination of mitigation strategies designed into the supply chain along with financial strategies to hedge uncovered risks. A global supply chain strategy focused on efficiency and low cost may concentrate global production in a few low-cost countries. Such a supply chain design is vulnerable to the risk of supply disruption along with fluctuations in transportation prices and exchange rates. In such a setting, it is crucial that the firm hedge fuel costs and exchange rates because the supply chain design itself has no built-in mechanisms to deal with these fluctuations. In contrast, a global supply chain designed with excess, flexible capacity allows production to be shifted to whatever location is most Table 6-4 Tailored Risk Mitigation Strategies During Network Design Risk Mitigation Strategy Tailored Strategies Increase capacity Focus on low-cost, decentralized capacity for predictable demand. Build centralized capacity for unpredictable demand. Increase Get redundant suppliers decentralization as cost of capacity drops. Increase responsiveness More redundant supply for high-volume products, less Increase inventory redundancy for low-volume products. Centralize redundancy for Increase flexibility low-volume products in a few flexible suppliers. Pool or aggregate demand Favor cost over responsiveness for commodity products. Favor Increase source capability responsiveness over cost for short–life cycle products. Decentralize inventory of predictable, lower value products. Centralize inventory of less predictable, higher value products. Favor cost over flexibility for predictable, high-volume products. Favor flexibility for unpredictable, low-volume products. Centralize flexibility in a few locations if it is expensive. Increase aggregation as unpredictability grows. Prefer capability over cost for high-value, high-risk products. Favor cost over capability for low-value commodity products. Centralize high capability in flexible source if possible. Source: Adapted from “Managing Risk to Avoid Supply Chain Breakdown.” Sunil Chopra and Manmohan S. Sodhi, Sloan Management Review (Fall 2004): 53–61.

150 Chapter 6 • Designing Global Supply Chain Networks effective in a given set of macroeconomic conditions. The ability of such a flexible design to react to fluctuations decreases the need for financial hedges. Operational hedges such as flexibility are more complex to execute than financial hedges, but they have the advantage of being reactive because the supply chain can be reconfigured to best react to the macroeconomic state of the world. It is important to keep in mind that any risk mitigation strategy is not always “in the money.” For example, flexibility built into Honda plants proved effective only when demand for vehicles shifted in an unpredictable manner in 2008. If there had been no fluctuation in demand, the flexibility would have gone unutilized. Flexibility in the form of the intelligent body assembly system (IBAS) built by Nissan in the early 1990s almost bankrupted the company because the state of the automotive markets was relatively stable at that time. Similarly, the use of fuel hedges that made billions for Southwest Airlines cost it money toward the end of 2008 when crude oil prices dropped significantly. It is thus critical that risk mitigation strategies be evaluated rigorously as real options in terms of their expected long-term value before they are implemented. In the following sections, we discuss methodologies that allow for the financial evaluation of risk mitigation strategies designed into a global supply chain. Flexibility, Chaining, and Containment Flexibility plays an important role in mitigating different risks and uncertainties faced by a global supply chain. Flexibility can be divided into three broad categories—new product flexibility, mix flexibility, and volume flexibility. New product flexibility refers to a firm’s ability to introduce new products into the market at a rapid rate. New product flexibility is critical in a competitive environment wherein technology is evolving and customer demand is fickle. New product flexibility may result from the use of common architectures and product platforms with the goal of providing a large number of distinct models using as few unique platforms as possible. The PC industry has historically followed this approach to introduce a continuous stream of new products. New product flexibility may also result if a fraction of the production capacity is flexible enough to be able to produce any product. This approach has been used in the pharmaceutical industry in which a fraction of the capacity is very flexible with all new products first manufactured there. Only once the product takes off is it moved to a dedicated capacity with lower variable costs. Mix flexibility refers to the ability to produce a variety of products within a short period of time. Mix flexibility is critical in an environment wherein demand for individual products is small or highly unpredictable, supply of raw materials is uncertain, and technology is evolving rapidly. The consumer electronics industry is a good example in which mix flexibility is essential in production environments, especially as more production has moved to contract manufacturers. Modular design and common components facilitate mix flexibility. Zara’s European facilities have significant mix flexibility, allowing the company to provide trendy apparel with highly unpredictable demand. Volume flexibility refers to a firm’s ability to operate profitably at different levels of output. Volume flexibility is critical in cyclical industries. Firms in the automotive industry that lacked volume flexibility were badly hurt in 2008 when demand for automobiles in the United States shrank significantly. The steel industry is an example in which some volume flexibility and consolidation have helped performance. Prior to 2000, firms had limited volume flexibility and did not adjust production volumes when demand started to fall. The result was a buildup of inventories and a significant drop in the price of steel. In the early 2000s, a few large firms consolidated and developed some volume flexibility. As a result, they were able to cut production as demand fell. The result has been less buildup of inventory and smaller drops in price during downturns, followed by a quicker recovery for the steel industry.

Chapter 6 • Designing Global Supply Chain Networks 151 Dedicated Fully Flexible Chained Network Chained Network Network Network with One Long with Two Short Chain Chains FIGURE 6-1 Different Flexibility Configurations in Network Given that some form of flexibility is often used to mitigate risks in global supply chains, it is important to understand the benefits and limitations of this approach. When dealing with demand uncertainty, Jordan and Graves (1995) make the important observation that as flexibility is increased, the marginal benefit derived from the increased flexibility decreases. They suggest operationalizing this idea in the concept of chaining, which is illustrated as follows. Consider a firm that sells four distinct products. A dedicated supply network with no flexibility would have four plants, each dedicated to producing a single product, as shown in Figure 6-1. A fully flexible network configuration would have each plant capable of producing all four products. The production flexibility of plants is beneficial when demand for each of the four products is unpredictable. With dedicated plants, the firm is not able to meet demand in excess of plant capacity. With flexible plants, the firm is able to shift excess demand for a product to a plant with excess capacity. Jordan and Graves define a chained network with one long chain (limited flexibility), configured as shown in Figure 6-1. In a chained configuration, each plant is capable of producing two products with the flexibility organized so that the plants and their products form a chain. Jordan and Graves show that a chained network mitigates the risk of demand fluctuation almost as effectively as a fully flexible network. Given the higher cost of full flexibility, the results of Jordan and Graves indicate that chaining is an excellent strategy to lower cost while gaining most of the benefits of flexibility. The desired length of chains is an important question to be addressed when designing chained networks. When dealing with demand uncertainty, longer chains have the advantage of effectively pooling available capacity to a greater extent. Long chains, however, do have a few disadvantages. The fixed cost of building a single long chain can be higher than the cost of multiple smaller chains. With a single long chain, the effect of any fluctuation ripples to all facilities in the chain, making coordination more difficult across the network. It has also been observed by several researchers that flexibility and chaining are effective when dealing with demand fluctuation but less effective when dealing with supply disruption. In the presence of supply disruption, Lim et al. (2008) have observed that designing smaller chains that contain or limit the impact of a disruption can be more effective than designing a network with one long chain. An example of containment is shown in the last example in Figure 6-1, which shows four plants with the flexibility to produce the four products in the form of two short chains. In this design, any disruption in one of the chains does not impact the other chain. A simple example of containment is hog farming: The farms are large to gain economies of scale, but the hogs are kept separated in small groups to ensure that the risk of disease is contained within a group and does not spread to the entire farm. Key Point Appropriate flexibility is an effective approach for a global supply chain to deal with a variety of risks and uncertainties. Whereas some flexibility is valuable, too much flexibility may not be worth the cost. Strategies like chaining and containment should be used to maximize the benefit from flexibility while keeping costs low.

152 Chapter 6 • Designing Global Supply Chain Networks 6.4 DISCOUNTED CASH FLOWS Global supply chain design decisions should be evaluated as a sequence of cash flows over the duration of time they will be in place. This requires the evaluation of future cash flows accounting for risks and uncertainties likely to arise in the global supply chain. In this section, we discuss the basics of analysis to evaluate future cash flows before introducing uncertainty in the next section. The present value of a stream of cash flows is what that stream is worth in today’s dollars. Discounted cash flow (DCF) analysis evaluates the present value of any stream of future cash flows and allows management to compare two streams of cash flows in terms of their financial value. DCF analysis is based on the fundamental premise that “a dollar today is worth more than a dollar tomorrow” because a dollar today may be invested and earn a return in addition to the dollar invested. This premise provides the basic tool for comparing the relative value of future cash flows that will arrive during different time periods. The present value of future cash flow is found by using a discount factor. If a dollar today can be invested and earn a rate of return k over the next period, an investment of $1 today will result in 1 ϩ k dollars in the next period. An investor would therefore be indifferent between obtaining $1 in the next period or $1/(1 ϩ k) in the current period. Thus, $1 in the next period is discounted by the discount factor = 1 (6.1) 1+k to obtain its present value. The rate of return k is also referred to as the discount rate, hurdle rate, or opportunity cost of capital. Given a stream of cash flows C0, C1, . . . , CT over the next T periods, and a rate of return k, the net present value (NPV) of this cash flow stream is given by NPV = C0 T 1 k t (6.2) 1+ +a a b Ct t=1 The NPV of different options should be compared when making supply chain decisions. A negative NPV for an option indicates that the option will lose money for the supply chain. The decision with the highest NPV will provide a supply chain with the highest financial return. EXAMPLE 6-1 Trips Logistics, a third-party logistics firm that provides warehousing and other logistics services, is facing a decision regarding the amount of space to lease for the upcoming three-year period. The general manager has forecast that Trips Logistics will need to handle a demand of 100,000 units for each of the next three years. Historically, Trips Logistics has required 1,000 sq. ft. of warehouse space for every 1,000 units of demand. For the purposes of this discussion, the only cost Trips Logistics faces is the cost for the warehouse. Trips Logistics receives revenue of $1.22 for each unit of demand. The general manager must decide whether to sign a three-year lease or obtain warehousing space on the spot market each year. The three-year lease will cost $1 per square foot per year, and the spot market rate is expected to be $1.20 per square foot per year for each of the three years. Trips Logistics has a discount rate of k ϭ 0.1. Analysis: The general manager decides to compare the NPV of signing a three-year lease for 100,000 sq. ft. of warehouse space with obtaining the space from the spot market each year. If the general manager obtains warehousing space from the spot market each year, Trips Logistics will earn $1.22 for each

Chapter 6 • Designing Global Supply Chain Networks 153 unit and pay $1.20 for one square foot of warehouse space required. The expected annual profit for Trips Logistics in this case is given by the following: Expected annual profit if warehousing = 100,000 * $1.22 space is obtained from spot market - 100,000 * $1.20 = $2,000. Obtaining warehouse space from the spot market provides Trips Logistics with an expected positive cash flow of $2,000 in each of the three years. The NPV may be evaluated as follows: NPV1No lease2 = C0 + C1 + C2 = 2,000 + 2,000 + 2,000 = $5,471 1+k 11 + k22 1.1 1.12 If the general manager leases 100,000 sq. ft. of warehouse space for the next three years, Trips Logistics pays $1 per square foot of space leased each year. The expected annual profit for Trips Logistics in this case is given by the following: Expected annual profit with three-year lease = 100,000 * $1.22 - 100,000 * $1.00 = $22,000. Signing a lease for three years provides Trips Logistics with a positive cash flow of $22,000 in each of the three years. The NPV may be evaluated as NPV1Lease2 = C0 + C1 + 11 C2 = 22,000 + 22,000 + 22,000 = $60,182 1+k + k22 1.1 1.12 The NPV of signing the lease is $60,182 Ϫ $5,471 ϭ $54,711 higher than obtaining warehousing space on the spot market. Based on this simple analysis, a manager may opt to sign the lease. However, this does not tell the whole story because we have not yet included the uncertainty in spot prices and valued the greater flexibility to adjust to uncertainty that the spot market provides the manager. In the next section, we introduce methodology that allows for uncertainty and discuss how the inclusion of uncertainty of future demand and costs may cause the manager to rethink the decision. 6.5 EVALUATING NETWORK DESIGN DECISIONS USING DECISION TREES In any global supply chain, demand, prices, exchange rates, and several other factors are highly uncertain and are likely to fluctuate during the life of any supply chain decision. In an uncertain environment, the problem with using a simple DCF analysis is that it typically undervalues flexibility. The result is often a supply chain that performs well if everything goes according to plan but becomes terribly expensive if something unexpected happens. A manager makes several different decisions when designing a supply chain network. For instance: • Should the firm sign a long-term contract for warehousing space or get space from the spot market as needed? • What should the firm’s mix of long-term and spot market be in the portfolio of transportation capacity? • How much capacity should various facilities have? What fraction of this capacity should be flexible? If uncertainty is ignored, a manager will always sign long-term contracts (because they are typically cheaper) and avoid all flexible capacity (because it is more expensive). Such decisions, however, can hurt the firm if future demand or prices are not as forecast at the time of the decision.

154 Chapter 6 • Designing Global Supply Chain Networks For example, until around 1990, all production capacity in the pharmaceutical industry was dedicated. Dedicated capacity was cheaper than flexible capacity but could be used only for the drug it was designed for. Pharmaceutical companies, however, found it difficult to forecast the demand and price for drugs in the marketplace. Thus, a large fraction of the dedicated capacity could go unused if the forecast demand did not materialize. Today, pharmaceutical companies have a strategy of carrying a portfolio of dedicated and flexible capacity. Most products are introduced in a flexible facility and are moved to a dedicated facility only when a reasonably accurate forecast of future demand is available. During network design, managers thus need a methodology that allows them to estimate the uncertainty in their forecast of demand and price and then incorporate this uncertainty in the decision-making process. Such a methodology is most important for network design decisions because these decisions are hard to change in the short term. In this section, we describe such a methodology and show that accounting for uncertainty can have a significant impact on the value of network design decisions. The Basics of Decision Tree Analysis A decision tree is a graphic device used to evaluate decisions under uncertainty. Decision trees with DCFs can be used to evaluate supply chain design decisions given uncertainty in prices, demand, exchange rates, and inflation. The first step in setting up a decision tree is to identify the number of time periods into the future that will be considered when making the decision. The decision maker should also identify the duration of a period—which could be a day, a month, a quarter, or any other time period. The duration of a period should be the minimum period of time over which factors affecting supply chain decisions may change by a significant amount. “Significant” is hard to define, but in most cases it is appropriate to use as a period the duration over which an aggregate plan holds. If planning is done monthly, we set the duration of a period at a month. In the following discussion, T will represent the number of time periods over which the supply chain decision is to be evaluated. The next step is to identify factors that will affect the value of the decision and are likely to fluctuate over the next T periods. These factors include demand, price, exchange rate, and inflation, among others. Having identified the key factors, the next step is to identify probability distributions that define the fluctuation of each factor from one period to the next. If, for instance, demand and price are identified as the two key factors that affect the decision, the probability of moving from a given value of demand and price in one period to any other value of demand and price in the next period must be defined. The next step is to identify a periodic discount rate k to be applied to future cash flows. It is not essential that the same discount rate apply to each period or even at every node in a period. The discount rate should take into account the inherent risk associated with the investment. In general, a higher discount rate should apply to investments with higher risk. The decision is now evaluated using a decision tree, which contains the present and T future periods. Within each period, a node must be defined for every possible combination of factor values (say, demand and price) that can be achieved. Arrows are drawn from origin nodes in Period i to end nodes in Period i ϩ 1. The probability on an arrow is referred to as the transition probability and is the probability of transitioning from the origin node in Period i to the end node in Period i ϩ 1. The decision tree is evaluated starting from nodes in Period T and working back to Period 0. For each node, the decision is optimized taking into account current and future values of various factors. The analysis is based on Bellman’s principle, which states that for any choice of strategy in a given state, the optimal strategy in the next period is the one that is selected if the entire analysis is assumed to begin in the next period. This principle allows the optimal strategy to be solved in a backward fashion starting at the last period. Expected future cash flows are discounted back and included in the decision currently under consideration. The value of the

Chapter 6 • Designing Global Supply Chain Networks 155 node in Period 0 gives the value of the investment as well as the decisions made during each time period. Tools such as Treeplan are available that help solve decision trees on spreadsheets. The decision tree analysis methodology is summarized as follows: 1. Identify the duration of each period (month, quarter, etc.) and the number of periods T over which the decision is to be evaluated. 2. Identify factors such as demand, price, and exchange rate whose fluctuation will be considered over the next T periods. 3. Identify representations of uncertainty for each factor; that is, determine what distribution to use to model the uncertainty. 4. Identify the periodic discount rate k for each period. 5. Represent the decision tree with defined states in each period as well as the transition probabilities between states in successive periods. 6. Starting at period T, work back to Period 0, identifying the optimal decision and the expected cash flows at each step. Expected cash flows at each state in a given period should be discounted back when included in the previous period. Evaluating Flexibility at Trips Logistics We illustrate the decision tree analysis methodology by using the lease decision facing the general manager at Trips Logistics. The manager must decide whether to lease warehouse space for the coming three years and the quantity to lease. The long-term lease is currently cheaper than the spot market rate for warehouse space. The manager anticipates uncertainty in demand and spot prices for warehouse space over the coming three years. The long-term lease is cheaper but could go unused if demand is lower than anticipated. The long-term lease may also end up being more expensive if future spot market prices come down. In contrast, spot market rates are high and warehouse space from the spot market will be expensive if future demand is high. The manager is considering three options: 1. Get all warehousing space from the spot market as needed. 2. Sign a three-year lease for a fixed amount of warehouse space and get additional requirements from the spot market. 3. Sign a flexible lease with a minimum charge that allows variable usage of warehouse space up to a limit with additional requirement from the spot market. We now discuss how the manager can make the appropriate decision, taking uncertainty into account. One thousand square feet of warehouse space is required for every 1,000 units of demand, and the current demand at Trips Logistics is for 100,000 units per year. The manager forecasts that from one year to the next, demand may go up by 20 percent with a probability of 0.5 or go down by 20 percent with a probability of 0.5. The probabilities of the two outcomes are independent and unchanged from one year to the next. The general manager can sign a three-year lease at a price of $1 per square foot per year. Warehouse space is currently available on the spot market for $1.20 per square foot per year. From one year to the next, spot prices for warehouse space may go up by 10 percent with probability 0.5 or go down by 10 percent with probability 0.5, according to a binomial process. The probabilities of the two outcomes are independent and unchanged from one year to the next. The general manager believes that prices of warehouse space and demand for the product fluctuate independently. Each unit Trips Logistics handles results in revenue of $1.22, and Trips Logistics is committed to handling all demand that arises. Trips Logistics uses a discount rate of k ϭ 0.1 for each of the three years. The general manager assumes that all costs are incurred at the beginning of each year and thus constructs a decision tree with T ϭ 2. The decision tree is shown in Figure 6-2, with each node representing demand (D) in thousands of units and price (p) in dollars. The probability of each transition is 0.25 because price and demand fluctuate independently.

156 Chapter 6 • Designing Global Supply Chain Networks Period 1 Period 2 Period 0 D = 100 D = 120 0.25 D = 144 p = $1.20 p = $1.32 0.25 p = $1.45 0.25 0.25 D = 144 0.25 D = 120 p = $1.19 p = $1.08 0.25 D = 96 0.25 D = 80 p = $1.45 0.25 p = $1.32 D = 144 p = $0.97 D = 80 p = $1.08 D = 96 p = $1.19 D = 96 p = $0.97 D = 64 p = $1.45 D = 64 p = $1.19 D = 64 p = $0.97 FIGURE 6-2 Decision Tree for Trips Logistics Considering Demand and Price Fluctuation Evaluating the Spot Market Option The manager first analyzes the option of not signing a lease and obtaining all warehouse space from the spot market. He starts with Period 2 and evaluates the profit for Trips Logistics at each node. At the node D ϭ 144, p ϭ $1.45, Trips Logistics must satisfy a demand of 144,000 and faces a spot price of $1.45 per square foot for warehouse space in Period 2. The cost incurred by Trips Logistics in Period 2 at the node D ϭ 144, p ϭ $1.45 is represented by C(D ϭ 144, p ϭ1.45, 2) and is given by C1D = 144, p = 1.45, 22 = 144,000 * 1.45 = $208,800 The profit at Trips Logistics in Period 2 at the node D ϭ 144, p ϭ $1.45 is represented by P(D ϭ 144, p ϭ 1.45, 2) and is given by P1D = 144, p = $1.45, 22 = 144,000 * 1.22 - C1D = 144, p = 1.45, 22 = 175,680 - 208,800 = - $33,120

Chapter 6 • Designing Global Supply Chain Networks 157 Table 6-5 Period 2 Calculations for Spot Market Option Revenue Cost Profit C(D ‫؍‬, p ‫؍‬, 2) P(D ‫؍‬, p ‫؍‬, 2) D ϭ 144, p ϭ 1.45 144,000 × 1.22 144,000 × 1.45 Ϫ$33,120 D ϭ 144, p ϭ 1.19 144,000 × 1.22 144,000 × 1.19 $4,320 D ϭ 144, p ϭ 0.97 144,000 × 1.22 144,000 × 0.97 D ϭ 96, p ϭ 1.45 $36,000 D ϭ 96, p ϭ 1.19 96,000 × 1.22 96,000 × 1.45 D ϭ 96, p ϭ 0.97 96,000 × 1.22 96,000 × 1.19 Ϫ$22,080 D ϭ 64, p ϭ 1.45 96,000 × 1.22 96,000 × 0.97 $2,880 D ϭ 64, p ϭ 1.19 64,000 × 1.22 64,000 × 1.45 D ϭ 64, p ϭ 0.97 64,000 × 1.22 64,000 × 1.19 $24,000 64,000 × 1.22 64,000 × 0.97 Ϫ$14,720 $1,920 $16,000 The profit for Trips Logistics at each of the other nodes in Period 2 is evaluated similarly, as shown in Table 6-5. The manager next evaluates the expected profit at each node in Period 1 to be the profit during Period 1 plus the present value (in Period 1) of the expected profit in Period 2. The expected profit EP(D ϭ, p ϭ, 1) at a node is the expected profit over all four nodes in Period 2 that may result from this node. PVEP(D ϭ, p ϭ, 1) represents the present value of this expected profit; P(D ϭ, p ϭ, 1), the total expected profit, is the sum of the profit in Period 1 and the present value of the expected profit in Period 2. From the node D ϭ 120, p ϭ $1.32 in Period 1, there are four possible states in Period 2. The manager thus evaluates the expected profit in Period 2 over all four states possible from the node D ϭ 120, p ϭ $1.32 in Period 1 to be EP(D ϭ 120, p ϭ 1.32, 1), where EP1D = 120, p = 1.32, 12 = 0.25 * 3P1D = 144, p = 1.45, 22 + P1D = 144, p = 1.19,2) + P1D = 96, p = 1.45, 22 + P1D = 96, p = 1.19, 224 = 0.25 * 3-33,120 + 4,320 - 22,080 + 2,8804 = - $12,000 The present value of this expected value in Period 1 is given by PVEP1D = 120, p = 1.32, 12 = EP1D = 120, p = 1.32, 12>11 + k2 = -12,000>1.1 = - $10,909 The manager obtains the total expected profit P(D ϭ 120, p ϭ 1.32, 1) at node D ϭ 120, p ϭ 1.32 in Period 1 to be the sum of the profit in Period 1 at this node and the present value of future expected profits. P1D = 120, p = 1.32, 12 = 120,000 * 1.22 - 120,000 * 1.32 + PVEP1D = 120, p = 1.32, 12 = - $12,000 - $10,909 = -$22,909 The total expected profit for all other nodes in Period 1 is evaluated as shown in Table 6-6. For Period 0, the total profit P(D ϭ 100, p ϭ 1.20, 0) is the sum of the profit at Period 0 and the present value of the expected profit over the four nodes in Period 1. EP1D = 100, p = 1.20, 02 = 0.25 * 3P1D = 120, p = 1.32, 12 + P1D = 120, p = 1.08, 12 + P1D = 96, p = 1.32, 12 + P1D = 96, p = 1.08, 124 = 0.25 * [-22,909 + 32,073 - 15,273 + 21,382] = $3,818

158 Chapter 6 • Designing Global Supply Chain Networks Table 6-6 Period 1 Calculations for Spot Market Option Node EP(D ‫؍‬, p ‫؍‬, 1) p (D ‫؍‬, p ‫؍‬, 1) ‫ ؍‬D ؋ 1.22 ؊ D ؋ p ؉ Ϫ$12,000 EP(D ‫؍‬, p ‫؍‬, 1)/(1 ؉ k) D ϭ 120, p ϭ 1.32 $16,800 D ϭ 120, p ϭ 1.08 Ϫ$22,909 D ϭ 80, p ϭ 1.32 Ϫ$8,000 $32,073 D ϭ 80, p ϭ 1.08 $11,200 Ϫ$15,273 $21,382 PVEP1D = 100, p = 1.20, 12 = EP1D = 100, p = 1.20, 02>(1 + k) = 3,818>1.1 = $3,471 P1D = 100, p = 1.20, 02 = 100,000 * 1.22 - 100,000 * 1.20 + PVEP1D = 100, p = 1.20, 02 = $2,000 + $3,471 = $5,471 Thus, the expected NPV of not signing the lease and obtaining all warehousing space from the spot market is given by NPV1Spot Market2 = $5,471 Evaluating the Fixed Lease Option The manager next evaluates the alternative whereby the lease for 100,000 sq. ft. of warehouse space is signed. The evaluation procedure is very similar to that for the previous case, but the outcome in terms of profit changes. For example, at the node D ϭ 144, p ϭ 1.45, the manager will require 44,000 sq. ft. of warehouse space from the spot market at $1.45 per square foot because only 100,000 sq. ft. have been leased at $1 per square foot. If demand happens to be less than 100,000 units, Trips Logistics still has to pay for the entire 100,000 sq. ft. of leased space. For Period 2, the manager obtains the profit at each of the nine nodes as shown in Table 6-7. The manager next evaluates the total expected profit for each node in Period 1. Again, the expected profit EP(D ϭ, p ϭ, 1) at a node is the expected profit of all four nodes in Period 2 that Table 6-7 Period 2 Profit Calculations at Trips Logistics for Fixed Lease Option Node Leased Space Warehouse Space Profit P (D ‫؍‬, p ‫؍‬, 2) ‫ ؍‬D ؋ 1.22 at Spot Price (S) ؊ (100,000 ؋ 1 ؉ S × p) D ϭ 144, p ϭ 1.45 100,000 sq. ft. D ϭ 144, p ϭ 1.19 100,000 sq. ft. 44,000 sq. ft. $11,880 D ϭ 144, p ϭ 0.97 100,000 sq. ft. 44,000 sq. ft. $23,320 D ϭ 96, p ϭ 1.45 100,000 sq. ft. D ϭ 96, p ϭ 1.19 100,000 sq. ft. 44,000 sq. ft. $33,000 D ϭ 96, p ϭ 0.97 100,000 sq. ft. 0 sq. ft. $17,120 D ϭ 64, p ϭ 1.45 100,000 sq. ft. 0 sq. ft. $17,120 D ϭ 64, p ϭ 1.19 100,000 sq. ft. 0 sq. ft. $17,120 D ϭ 64, p ϭ 0.97 100,000 sq. ft. 0 sq. ft. Ϫ$21,920 0 sq. ft. Ϫ$21,920 0 sq. ft. Ϫ$21,920

Chapter 6 • Designing Global Supply Chain Networks 159 Table 6-8 Period 1 Profit Calculations at Trips Logistics for Fixed Lease Option Node EP(D ‫؍‬, p ‫؍‬, 1) Warehouse Space P(D ‫؍‬, p ‫؍‬, 1) ‫ ؍‬D ؋ 1.22 D ϭ 120, p ϭ 1.32 at Spot Price (S) ؊ (100,000 ؋ 1 ؉ S ؋ p) 0.25 × [P(D ϭ 144, p ϭ 1.45, 2) ϩ P(D ϭ 144, ؉ EP(D ‫؍‬, p ‫؍‬, 1)(1؉ k) D ϭ 120, p ϭ 1.08 p ϭ 1.19, 2) ϩ P(D ϭ 96, p ϭ 1.45, 2) 20,000 D ϭ 80, p ϭ 1.32 ϩ P(D ϭ 96, p ϭ 1.19, 2)] ϭ 0.25 × $35,782 D ϭ 80, p ϭ 1.08 (11,880 ϩ 23,320 ϩ 17,120 ϩ 20,000 17,120) ϭ $17,360 0 $45,382 0 0.25 × [23,320 ϩ 33,000 ϩ 17,120 Ϫ$4,582 ϩ 17,120] ϭ $22,640 Ϫ$4,582 0.25 × [17,120 ϩ 17,120 Ϫ 21,920 Ϫ 21,920] ϭ Ϫ$2,400 0.25 × [17,120 ϩ 17,120 Ϫ21,920 Ϫ21,920] ϭ Ϫ$2,400 may result from this node (see Figure 6-2), and P(D ϭ, p ϭ, 1) is the total expected profit from both Periods 1 and 2. The manager thus obtains the results in Table 6-8. For Period 0, the expected profit EP(D ϭ 100, p ϭ 1.20, 0) over the four nodes in Period 1 is given by EP1D = 100, p = 1.20, 02 = 0.25 * 3P1D = 120, p = 1.32, 12 + P1D = 120, p = 1.08, 1) + P1D = 96, p = 1.32, 12 + P1D = 96, p = 1.08, 124 = 0.25 * [35,782 + 45,382 - 4,582 - 4,582] = $18,000 The present value of the expected profit in Period 0 is given by PVEP1D = 100, p = 1.20, 02 = EP1D = 100, p = 1.20, 02>11 + k2 = 18,000>1.1 = $16,364 The total expected profit is obtained as the sum of the profit in Period 0 and the present value of the expected profit over all four nodes in Period 1. It is P1D = 100, p = 1.20, 02 = 100,000 * 1.22 - 100,000 * 1 + PVEP1D = 100, p = 1.20, 02 = $22,000 + $16,364 = $38,364 The NPV of signing a three-year lease for 100,000 sq. ft. of warehouse space is thus NPV1Lease2 = $38,364 Observe that the NPV of the lease option under uncertainty is considerably less compared to when uncertainty is ignored ($60,182 from Example 6-1). This is because the lease is a fixed decision, and Trips Logistics is unable to react to market conditions by leasing less space if demand is lower. Rigid contracts are less attractive in the presence of uncertainty. The presence of uncertainty in demand and price reduces the value of the lease but does not affect the value of the spot market option. The manager, however, still prefers to sign the three-year lease for 100,000 sq. ft. because this option has a higher expected profit.

160 Chapter 6 • Designing Global Supply Chain Networks Key Point Uncertainty in demand and economic factors should be included in the financial evaluation of supply chain design decisions. The inclusion of uncertainty typically decreases the value of rigidity and increases the value of flexibility. Evaluating the Flexible Lease Option The decision tree analysis methodology is useful when evaluating flexibility within a supply chain. We now consider the evaluation of flexibility with decision trees in the context of warehousing choices for Trips Logistics. The general manager at Trips Logistics has been offered a contract in which, for an up-front payment of $10,000, Trips Logistics will have the flexibility of using between 60,000 sq. ft. and 100,000 sq. ft. of warehouse space at $1 per square foot per year. Trips Logistics must pay $60,000 per year for the first 60,000 sq. ft. and can then use up to another 40,000 sq. ft. on demand at $1 per square foot. The general manager decides to use decision trees to evaluate whether this flexible contract is preferable to a fixed contract for 100,000 sq. ft. The underlying decision tree for evaluating the flexible contract is exactly as in Figure 6-2. The profit at each node, however, changes because of the flexibility in space used. If demand is larger than 100,000 units, Trips Logistics uses all 100,000 sq. ft. of warehouse space even under the flexible contract. If demand is between 60,000 and 100,000 units, Trips Logistics pays only for the exact amount of warehouse space used rather than the entire 100,000 sq. ft. under the contract without flexibility. The profit at all nodes where demand is 100,000 or higher remains the same as in Table 6-7. The profit in Period 2 at all nodes where demand is less than 100,000 units increases as shown in Table 6-9. The general manager evaluates the expected profit EP(D ϭ, p ϭ, 1) from Period 2 and the total expected profit for each node in Period 1 as discussed earlier. The results are shown in Table 6-10. The total expected profit in Period 0 is the sum of the profit in Period 0 and the present value of the expected profit in Period 1. The manager thus obtains EP1D = 100, p = 1.20, 02 = 0.25 * 3P1D = 120, p = 1.32, 12 + P1D = 120, p = 1.08,1) + P1D = 96, p = 1.32, 12 + P1D = 96, p = 1.08, 124 = 0.25 * [37,600 + 47,200 + 33,600 + 33,600] = $38,000 Table 6-9 Period 2 Profit Calculations at Trips Logistics with Flexible Lease Contract Warehouse Space Warehouse Space Profit P(D ‫؍‬, p ‫؍‬, 2) ‫ ؍‬D ؋ 1.22 ؊ (W ؋ 1 ؉ S ؋ p) Node at $1 (W) at Spot Price (S) $11,880 D ϭ 144, p ϭ 1.45 100,000 sq. ft. 44,000 sq. ft. $23,320 D ϭ 144, p ϭ 1.19 100,000 sq. ft. 44,000 sq. ft. $33,000 D ϭ 144, p ϭ 0.97 100,000 sq. ft. 44,000 sq. ft. $21,120 D ϭ 96, p ϭ 1.45 $21,120 D ϭ 96, p ϭ 1.19 96,000 sq. ft. 0 sq. ft. $21,120 D ϭ 96, p ϭ 0.97 96,000 sq. ft. 0 sq. ft. $14,080 D ϭ 64, p ϭ 1.45 96,000 sq. ft. 0 sq. ft. $14,080 D ϭ 64, p ϭ 1.19 64,000 sq. ft. 0 sq. ft. $14,080 D ϭ 64, p ϭ 0.97 64,000 sq. ft. 0 sq. ft. 64,000 sq. ft. 0 sq. ft.

Chapter 6 • Designing Global Supply Chain Networks 161 Table 6-10 Period 1 Profit Calculations at Trips Logistics with Flexible Lease Contract Node EP(D‫؍‬, p ‫؍‬, 1) Warehouse Warehouse P(D ‫؍‬, p ‫؍‬, 1) ‫ ؍‬D ؋ 1.22 ؊ D ϭ 120, p ϭ 1.32 Space at Space at (W ؋ 1 ؉ S ؋ p) ؉ D ϭ 120, p ϭ 1.08 0.25 × [11,880 ϩ 23,320 $1 (W) D ϭ 80, p ϭ 1.32 ϩ 21,120 ϩ 21,120] ϭ $19,360 100,000 Spot Price (S) EP(D ‫؍‬, p ‫؍‬, 1) (1 ؉ k) D ϭ 80, p ϭ 1.08 20,000 $37,600 0.25 × [23,320 ϩ 33,000 100,000 ϩ 21,120 + 21,120] ϭ $24,640 20,000 $47,200 80,000 0.25 × [21,120 ϩ 21,120 0 $33,600 ϩ 14,080 ϩ 14,080] ϭ $17,600 80,000 0 $33,600 0.25 × [21,120 ϩ 21,120 ϩ 14,080 ϩ 14,200] ϭ $17,600 Table 6-11 Comparison of Different Lease Options for Trips Logistics Option Value All warehouse space from the spot market $5,471 Lease 100,000 sq. ft. for three years $38,364 Flexible lease to use between 60,000 and 100,000 sq. ft. $46,545 PVEP1D = 100, p = 1.20, 12 = EP1D = 100, p = 1.20, 02>11 + k2 = 38,000 >1.1 = $34,545 P1D = 100, p = 1.20, 02 = 100,000 * 1.22 - 100,000 * 1 + PVEP1D = 100, p = 1.20, 02 = $22,000 + $34,545 = $56,545 With an up-front payment of $10,000, the net expected profit is $46,545 under the flexible lease. The value of flexibility may now be obtained as the difference between the expected present values of the two contracts. Accounting for uncertainty, the manager at Trips Logistics values the three options as shown in Table 6-11. The flexible contract is thus beneficial for Trips Logistics because it is $8,181 more valuable than the rigid contract for three years. 6.6 TO ONSHORE OR OFFSHORE: EVALUATION OF GLOBAL SUPPLY CHAIN DESIGN DECISIONS UNDER UNCERTAINTY In this section, we discuss a supply chain design decision at D-Solar, a German manufacturer of solar panels, to illustrate the power of the decision tree analysis methodology for designing global supply chain networks while accounting for uncertainty. D-Solar faces a plant location decision in a global network with fluctuating exchange rates and demand uncertainty. Key Point Flexibility should be valued by taking into account uncertainty in demand and economic factors. In general, the value of flexibility increases with an increase in uncertainty.

162 Chapter 6 • Designing Global Supply Chain Networks Table 6-12 Fixed and Variable Production Costs for D-Solar European Plant Chinese Plant Fixed Cost (euro) Variable Cost (euro) Fixed Cost (yuan) Variable Cost (yuan) 1 million/year 40/panel 8 million/year 340/panel D-Solar sells its products primarily in Europe. Demand in the Europe market is currently 100,000 panels per year and each panel sells for €70. While panel demand is expected to grow, there are some downside risks if the economy slides. From one year to the next, demand may increase by 20 percent with probability 0.8 or decrease by 20 percent with probability 0.2. D-Solar has to decide whether to build a plant in Europe or China. In either case, D-Solar plans to build a plant with a rated capacity of 120,000 panels. The fixed and variable costs of the two plants are shown in Table 6-12. Observe that the fixed costs are given per year rather than as a one-time investment. The European plant is more expensive but will also have greater volume flexibility. The plant will be able to increase or decrease production anywhere in the range of 60,000 to 150,000 panels while maintaining its variable cost. In contrast, the Chinese plant is cheaper (at the current exchange rate of 9 yuan/euro) but will have limited volume flexibility and can produce only between 100,000 and 130,000 panels. If the Chinese plant is built, D-Solar will have to incur variable cost for 100,000 panels even if demand drops below that level and will lose sales if demand increases above 130,000 panels. Exchange rates are volatile, and each year the yuan is expected to rise 10 percent with a probability of 0.7 or drop 10 percent with a probability of 0.3. We assume that the sourcing decision will be in place over the next three years and the discount rate used by D-Solar is k ϭ 0.1. All costs and revenues are assumed to accrue at the beginning of the year, allowing us to consider the first year as period 0 and the following two years as periods 1 and 2. Evaluating the Options Using DCF and Expected Demand and Exchange Rate A simplistic approach often taken is to consider the expected movement of demand and exchange rates in future periods when evaluating discounted cash flows. The weakness of such an approach is that it averages the trends while ignoring the uncertainty. We start by considering such a simplistic approach for the onshoring and offshoring options. On average, demand is expected to increase by 12 percent (20 × 0.8 Ϫ 20 × 0.2 ϭ 12), while the yuan is expected to strengthen by 4 percent (10 × 0.7 – 10 × 0.2 ϭ 4) each year. In this case, the expected demand and exchange rates in the two future periods are shown in Table 6-13. We now evaluate the discounted cash flows for both options assuming the average expected change in demand and exchange rates over the next two periods. For the on-shoring option, we have the following: Period 0 profits = 100,000 * 70 - 1,000,000 - 100,000 * 40 = €2,000,000 Period 1 profits = 112,000 * 70 - 1,000,000 - 112,000 * 40 = €2,360,000 Period 2 profits = 125,440 * 70 - 1,000,000 - 125,440 * 40 = € 2,763,200 Table 6-13 Expected Future Demand and Exchange Rate Period 1 Period 2 Demand Exchange Rate Demand Exchange Rate 112,000 8.64 yuan/ euro 125,440 8.2944 yuan /euro

Chapter 6 • Designing Global Supply Chain Networks 163 Thus, the DCF for the onshoring option is obtained as follows: Expected profit from onshoring = 2,000,000 + 2,360,000/1.1 + 2,763,200/1.21 = €6,429,091 For the off-shoring option we have the following: Period 0 profits = 100,000 * 70 - 8,000,000/9 - 100,000 * 340/9 = €2,333,333 Period 1 profits = 112,000 * 70 - 8,000,000/8.64 - 112,000 * 340/8.64 = €2,506,667 Period 2 profits = 125,440 * 70 - 8,000,000/7.9524 - 125,440 * 340/7.9524 = €2,674,319 Thus, the DCF for the off-shoring option is obtained as follows: Expected profit from off-shoring = 2,333,333 + 2,506,667/1.1 + 2,674,319/1.21 = €6,822,302 Based on performing a simple DCF analysis and assuming the expected trend of demand and exchange rates over the next two periods, it seems that offshoring should be preferred to onshoring because it is expected to provide additional profits of almost €393,000. The problem with the above analysis is that it has ignored uncertainty. For example, even though the demand is expected to grow, there is some probability that it will decrease. If demand drops below 100,000 panels, the offshore option could end up costing more because of the lack of flexibility. Similarly, if demand increases more than expected (grows by 20 percent in each of the two years), the offshore facility will not be able to keep up with the increase. An accurate analysis must reflect the uncertainties and should ideally be performed using a decision tree. Evaluating the Options Using Decision Trees For this analysis we construct a decision tree as shown in Figure 6-3. Each node in a given period leads to four possible nodes in the next period because demand and the exchange rate may go up or down. The detailed links and transition probabilities are shown in Figure 6-3. Demand is in thousands and is represented by D. The exchange rate is represented by E, where E is the number of yuan to a euro. For example, starting with the node D ϭ 100, E ϭ 9.00 in Period 0, one can transition to any of four nodes in Period 1. The transition to the node D ϭ 120, E ϭ 9.90 in Period 1 occurs if demand increases (probability of 0.8) and the yuan weakens (probability of 0.3). Thus, the transition from node D ϭ 100, E ϭ 9.00 in Period 0 to node D ϭ 120, E ϭ 9.90 in Period 1 occurs with probability 0.8 × 0.3 ϭ 0.24. All other transition probabilities in Figure 6-3 are calculated in a similar manner. The main advantage of using a decision tree is that it allows for the true evaluation of profits in each scenario that D-Solar may find itself. Evaluating the Onshore Option Recall that the onshore option is flexible and can change production levels (and thus variable costs) to match demand levels between 60,000 and 150,000. In the following analysis, we calculate the expected profits at each node in the decision tree (represented by the correspon- ding values of D and E) starting in Period 2 and working back to the present (Period 0). With the onshore option, exchange rates do not affect profits in euro because both revenue and costs are in euro. PERIOD 2 EVALUATION We provide a detailed analysis for the node D ϭ 144 (solar panel demand of 144,000), E ϭ 10.89 (exchange rate of 10.89 yuan per euro). Given its flexibility, the onshore facility is able to produce the entire demand of 144,000 panels at

164 Chapter 6 • Designing Global Supply Chain Networks Period 1 Period 2 Period 0 0.8 × 0.3 D = 144 E = 10.89 0.8 × 0.7 D = 144 E = 8.91 D = 120 0.2 × 0.3 D = 96 E = 9.90 E = 10.89 0.8 × 0.3 D = 120 0.8 × 0.3 0.2 0.8 × 0.3 D = 96 0.8 × 0.3E = 8.10 0.2 × 0.3 × 0.7 E = 8.91 D = 80 0.8 × 0.7 E = 9.90 0.8 × 0.7 0.2 × 0.3 D = 80 0.7 D = 100 E = 8.10 0.8 × D = 144 E = 9.00 0.2 × 0.7 E = 7.29 D = 96 E = 7.29 0.2 × 0.70.8 × 0.7 0.2 × 0.7 0.2 × 0.3 0.2 × 0.3 D = 64 E = 10.89 0.2 × 0.7 D = 64 E = 8.91 D = 64 E = 7.29 FIGURE 6-3 Decision Tree for D-Solar a variable cost of €40 and sell each panel for revenue of €70. Revenues and costs are evaluated as follows: Revenue from the manufacture and sale of 144,000 panels = 144,000 * 70 = €10,080,000 Fixed + variable cost of onshore plant = 1,000,000 + 144,000 * 40 = €6,760,000 In Period 2, the total profit for D-Solar at the node D ϭ 144, E ϭ 10.89 for the onshore option is thus given by P1D = 144, E = 10.89, 22 = 10,080,000 - 6,760,000 = €3,320,000 Using the same approach, we can evaluate the profit in each of the nine states (represented by the corresponding value of D and E) in Period 2 as shown in Table 6-14.

Chapter 6 • Designing Global Supply Chain Networks 165 Table 6-14 Period 2 Profits for Onshore Option Production Cost (euro) Profit (euro) D E Sales Cost Quantity Revenue (euro) 6,760,000 3,320,000 144 10.89 144,000 144,000 10,080,000 6,760,000 3,320,000 144 8.91 144,000 144,000 10,080,000 4,840,000 1,880,000 4,840,000 1,880,000 96 10.89 96,000 96,000 6,720,000 6,760,000 3,320,000 96 8.91 96,000 96,000 6,720,000 4,840,000 1,880,000 144 7.29 144,000 144,000 10,080,000 3,560,000 96 7.29 96,000 96,000 6,720,000 3,560,000 920,000 64 10.89 64,000 64,000 4,480,000 3,560,000 920,000 64 8.91 64,000 64,000 4,480,000 920,000 64 7.29 64,000 64,000 4,480,000 PERIOD 1 EVALUATION Period 1 contains four outcome nodes to be analyzed. A detailed analysis for one of the nodes, D ϭ 120, E ϭ 9.90, is presented here. In addition to the revenue and cost at this node, we also need to consider the present value of the expected profit in Period 2 from the four nodes that may result. The transition probability into each of the four nodes is as shown in Figure 6-3. The expected profit in Period 2 for the four potential outcomes resulting from the node D ϭ 120, E ϭ 9.90 is thus given by EP(D = 120, E = 9.90, 1) = 0.24 * P(D = 144, E = 10.89, 2) + 0.56 * P(D = 144, E = 8.91, 2) + 0.06 * P(D = 96, E = 10.89, 2) + 0.14 * P(D = 96, E = 8.91, 2) = 0.24 * 3,320,000 + 0.56 * 3,320,000 + 0.06 * 1,880,000 + 0.14 * 1,880,000 = €3,032,000 The present value of the expected profit in Period 2 discounted to Period 1 is given by PVEP1D = 120, E = 9.90, 12 = EP1D = 120, E = 9.90, 12>11 + k2 = 3,032,000/1.1 = €2,756,364 Next we evaluate the profits at the onshore plant at the node D ϭ 120, E ϭ 9.90 from its operations in Period 1, in which the onshore plant produces 120,000 panels at a variable cost of €40 and obtains revenue of €70 per panel. Revenues and costs are evaluated as follows: Revenue from manufacture and sale of 120,000 panels = 120,000 * 70 = € 8,400,000 Fixed + variable cost of onshore plant = 1,000,000 + 120,000 * 40 = €5,800,000 The expected profit for D-Solar at the node D ϭ 120, E ϭ 9.90 is obtained by adding the operational profits at this node in Period 1 and the discounted expected profits from the four nodes that may result in Period 2. The expected profit at this node in Period 1 is given by P1D = 120, E = 9.90, 12 = 8,400,000 - 5,800,000 + PVEP(D = 120, E = 9.90, 1) = 2,600,000 + 2,756,364 = €5,356,364 The expected profits for all nodes in Period 1 are calculated similarly and shown in Table 6-15.

166 Chapter 6 • Designing Global Supply Chain Networks Table 6-15 Period 1 Profits for Onshore Option Production Revenue Cost Expected Profit (euro) D E Sales Cost Quantity (euro) (euro) 5,356,364 120 9.90 120,000 120,000 8,400,000 5,800,000 5,356,364 120 8.10 120,000 120,000 8,400,000 5,800,000 2,934,545 5,600,000 4,200,000 2,934,545 80 9.90 80,000 80,000 5,600,000 4,200,000 80 8.10 80,000 80,000 PERIOD 0 EVALUATION In Period 0, the demand and exchange rate are given by D ϭ 100, E ϭ 9. In addition to the revenue and cost at this node, we also need to consider the discounted expected profit from the four nodes in Period 1. The expected profit is given by EP(D = 100, E = 9.00, 0) = 0.24 * P(D = 120, E = 9.90, 1) + 0.56 * P(D = 120, E = 8.10, 1) + 0.06 * P(D = 80, E = 9.90, 1) + 0.14 * P(D = 80, E = 8.10, 1) = 0.24 * 5,356,364 + 0.56 * 5,5356,364 + 0.06 * 2,934,545 + 0.14 * 2,934,545 = € 4,872,000 The present value of the expected profit in Period 1 discounted to Period 0 is given by PVEP1D = 100, E = 9.00, 02 = EP1D = 100, E = 9.00, 02>11 + k2 = 4,872,000/1.1 = €4,429,091 Next we evaluate the profits from the onshore plant’s operations in Period 0 from the manufacture and sale of 100,000 panels. Revenue from manufacture and sale of 100,000 panels = 100,000 * 70 = € 7,000,000 Fixed + variable cost of onshore plant = 1,000,000 + 100,000 * 40 = €5,000,000 The expected profit for D-Solar at the node D ϭ 100, E ϭ 9.00 in Period 0 is given by P1D = 100, E = 9.00, 02 = 7,000,000 - 5,000,000 + PVEP(D = 100, E = 9.00,0) = 2,000,000 + 4,429,091 = € 6,429,091 Thus, building the onshore plant has an expected payoff of €6,429,091 over the evaluation period. This number accounts for uncertainties in demand and exchange rates and the ability of the onshore facility to react to these fluctuations. Evaluating the Offshore Option As with the onshore option, we start by evaluating profits at each node in Period 2 and then back our evaluation to Periods 1 and 0. Each node is represented by the corresponding values of D and E. Recall that the offshore option is not fully flexible and can change production levels (and thus variable costs) only between 100,000 and 130,000 panels. Thus, if demand falls below 100,000 panels, D-Solar still incurs the variable production cost of 100,000 panels. If demand increases above 130,000 panels, the offshore facility can meet demand only up to 130,000 panels. At each node, given the demand, we calculate the expected profits accounting for the exchange rate that influences offshore costs evaluated in euro.

Chapter 6 • Designing Global Supply Chain Networks 167 PERIOD 2 EVALUATION The detailed analysis for the node D ϭ 144 (solar panel demand of 144,000), E ϭ 10.89 (exchange rate of 10.89 yuan per euro) is as follows. Even though demand is for 144,000 panels, given its lack of volume flexibility, the offshore facility is able to produce only 130,000 panels at a variable cost of 340 yuan each and sell each panel for revenue of €70. Revenues and costs are evaluated as follows: Revenue from manufacture and sale of 130,000 panels = 130,000 * 70 = € 9,100,000 Fixed + variable cost of offshore plant = 8,000,000 + 130,000 * 340 = 52,200,000 yuan The total profit for D-Solar at the node D ϭ 144, E ϭ 10.89 for the offshore option (evaluated in euro), is thus given by P1D = 144, E = 10.89, 22 = 9,100,000 - 152,200,00>10.892 = € 4,306,612 Using the same approach, we can evaluate the profit in each of the nine states (represented by the corresponding values of D and E) in Period 2 as shown in Table 6-16. Observe that the lack of flexibility at the offshore facility hurts D-Solar whenever demand is above 130,000 (lost margin) or below 100,000 (higher costs). For example, when the demand drops to 64,000 panels, the offshore facility continues to incur variable production costs for 100,000 panels. Profits are also hurt when the yuan is stronger than expected. PERIOD 1 EVALUATION In Period 1, there are four outcome nodes to be analyzed. As with the onshore option, a detailed analysis for one of the nodes D ϭ 120, E ϭ 9.90 is presented here. In addition to the revenue and cost from operations at this node, we also need to consider the present value of the expected profit in Period 2 from the four nodes that may result. The transi- tion probability into each of the four nodes is as shown in Figure 6-3. The expected profit in Period 2 from the node D ϭ 120, E ϭ 9.90 is thus given by EP(D = 120, E = 9.90,1) = 0.24 * P(D = 144, E = 10.89, 2) + 0.56 * P(D = 144, E = 8.91, 2) + 0.06 * P(D = 96, E = 10.89, 2) + 0.14 * P(D = 96, E = 8.91, 2) = 0.24 * 4,306,612 + 0.56 * 3,241,414 + 0.06 * 2,863,251 + 0.14 * 2,006,195 = € 3,301,441 Table 6-16 Period 2 Profits for Offshore Option Production Cost (yuan) Profit (euro) D E Sales Cost Quantity Revenue (euro) 52,200,000 4,306,612 144 10.89 130,000 130,000 9,100,000 52,200,000 3,241,414 144 8.91 130,000 130,000 9,100,000 42,000,000 2,863,251 100,000 6,720,000 42,000,000 2,006,195 96 10.89 96,000 100,000 6,720,000 52,200,000 1,939,506 96 8.91 96,000 130,000 9,100,000 42,000,000 144 7.29 130,000 100,000 6,720,000 42,000,000 958,683 96 7.29 96,000 100,000 4,480,000 42,000,000 623,251 64 10.89 64,000 100,000 4,480,000 –233,805 64 8.91 64,000 100,000 4,480,000 3,560,000 –1,281,317 64 7.29 64,000

168 Chapter 6 • Designing Global Supply Chain Networks The present value of the expected profit in Period 2 discounted to Period 1 is given by PVEP1D = 120, E = 9.90, 12 = EP1D = 120, E = 9.90, 12>11 + k2 = 3,301,441/1.1 = €3,001,310 Next we evaluate the profits at the offshore plant at the node D ϭ 120, E ϭ 9.90 from its operations in Period 1. The offshore plant produces 120,000 panels at a variable cost of 340 yuan and obtains revenue of €70 per panel. Revenues and costs are evaluated as follows: Revenue from manufacture and sale of 120,000 panels = 120,000 * 70 = € 8,400,000 Fixed + variable cost of onshore plant = 8,000,000 + 120,000 * 340 = 48,800,000 yuan The expected profit for D-Solar at the node D ϭ 120, E ϭ 9.90 in Period 1 is given by P1D = 120, E = 9.90, 12 = 8,400,000 - 148,800,00>9.902 + PVEP(D = 120, E = 9.90, 1) = 3,470,707 + 3,001,310 = €6,472,017 For the offshore option, the expected profits for all nodes in Period 1 are shown in Table 6-17. Observe that for the node D ϭ 80, E ϭ 8.10, D-Solar has a lower expected profit from the offshore option (Table 6-17) relative to the onshore option (see Table 6-15) because the offshore plant incurs high variable cost given its lack of flexibility (cost is incurred for 100,000 units even though only 80,000 are sold), and all offshore costs become expensive given the strong yuan. PERIOD 0 EVALUATION In Period 0, the demand and exchange rate are given by D ϭ 100, E ϭ 9. In addition to the revenue and cost at this node, we also need to consider the present value of expected profit from the four nodes in Period 1. The expected profit for the offshore option is given by EP(D = 100, E = 9.00, 0) = 0.24 * P(D = 120, E = 9.90, 1) + 0.56 * P(D = 120, E = 8.10, 1) + 0.06 * P(D = 80, E = 9.90, 1) + 0.14 * P(D = 80, E = 8.10, 1) = 0.24 * 6,472,017 + 0.56 * 4,301,354 + 0.06 * 3,007,859 + 0.14 * 1,164,757 = € 4,305,580 The present value of the expected profit in Period 1 discounted to Period 0 is given by PVEP1D = 100, E = 9.00, 02 = EP1D = 100, E = 9.00, 02>11 + k2 = 4,305,580/1.1 = €3,914,164 Table 6-17 Period 1 Profits for Offshore Option Production Expected Profit D E Sales Cost Quantity Revenue (euro) Cost (yuan) (euro) 120 9.90 120,000 120,000 8,400,000 48,800,000 6,472,017 120 8.10 120,000 120,000 8,400,000 48,800,000 4,301,354 80 9.90 80,000 100,000 5,600,000 42,000,000 3,007,859 80 8.10 80,000 100,000 5,600,000 42,000,000 1,164,757

Chapter 6 • Designing Global Supply Chain Networks 169 Next we evaluate the profits from the offshore plant’s operations in Period 0 from the manufacture and sale of 100,000 panels. Revenue from manufacture and sale of 100,000 panels = 100,000 * 70 = € 7,000,000 Fixed + variable cost of offshore plant = 8,000,000 + 100,000 * 340 = 42,000,000 yuan The expected profit for D-Solar at the node D ϭ 100, E ϭ 9.00 in Period 0 is given by P1D = 100, E = 9.00, 02 = 7,000,000 - (42,000,00>9.002 + PVEP(D = 100, E = 9.00, 0) = 2,333,333 + 3,914,164 = € 6,247,497 Thus, building the offshore plant has an expected payoff of €6,247,497 over the evaluation period. Observe that the use of a decision tree that accounts for both demand and exchange rate fluctuation shows that the onshore option and its flexibility is in fact more valuable (worth €6,429,091) than the offshore facility (worth €6,247,497), which is less flexible. This is in direct contrast to the decision that would have resulted if we had simply used the expected change in demand and exchange rate from one year to the next. When using the expected change in demand, the onshore option provided expected profits of €6,429,091, while the offshore option provided expected profits of €6,822,302. The offshore option is overvalued in this case because the potential fluctuations in demand and exchange rates are wider than the expected fluctuations. Using the expected fluctuation thus does not fully account for the lack of flexibility in the offshore facility and the big increase in costs that may result if the yuan strengthens more than the expected value. De Traville and Trigeorgis (2010) discuss the importance of evaluating all global supply chain design decisions using the decision tree or real options methodology. They give the example of Flexcell, a Swiss company that offered lightweight solar panels. In 2006, the company was looking to expand its operations by building a new plant. The three locations under discussion were China, eastern Germany, and near the company headquarters in Switzerland. Even though the Chinese and eastern German plants were cheaper than the Swiss plant, Flexcell management justified building the high-cost Swiss plant because of its higher flexibility and ability to react to changing market conditions. If only the expected values of future scenarios had been used, the more expensive Swiss plant could not be justified. This decision paid off for the company because the Swiss plant was flexible enough to handle the considerable variability in demand that resulted during the downturn in 2008. When underlying decision trees are complex and explicit solutions for the underlying decision tree are difficult to obtain, firms should use simulation for evaluating decisions (see Chapter 13). In a complex decision tree, thousands or even millions of possible paths may arise from the first period to the last. Transition probabilities are used to generate probability- weighted random paths within the decision tree. For each path, the stage-by-stage decision and the present value of the payoff are evaluated. The paths are generated in such a way that the probability of a path being generated during the simulation is the same as the probability of the path in the decision tree. After generating many paths and evaluating the payoffs in each case, the payoffs obtained during the simulation are used as a representation of the payoffs that would result from the decision tree. The expected payoff is then found by averaging the payoffs obtained in the simulation. Simulation methods are very good at evaluating a decision when the path itself is not decision dependent—in other words, when transition probabilities from one period to the next are not dependent on the decision taken during a period. They can also take into account real-world constraints as well as complex decision rules. In addition, they can easily handle different forms of uncertainty even when uncertainty between different factors is correlated.

170 Chapter 6 • Designing Global Supply Chain Networks Simulation models require a higher setup cost to start and operate compared to decision tree tools. However, their main advantage is that they can provide high-quality evaluations of complex situations. 6.7 MAKING GLOBAL SUPPLY CHAIN DESIGN DECISIONS UNDER UNCERTAINTY IN PRACTICE Managers should consider the following ideas to help them make better network design decisions under uncertainty. 1. Combine strategic planning and financial planning during global network design. In most organizations, financial planning and strategic planning are performed independently. Strategic planning tries to prepare for future uncertainties but often without rigorous quantitative analysis, whereas financial planning performs quantitative analysis but assumes a predictable or well-defined future. This chapter presents methodologies that allow integration of financial and strategic planning. Decision makers should design global supply chain networks considering a portfolio of strategic options—the option to wait, build excess capacity, build flexible capacity, sign long-term contracts, purchase from the spot market, and so forth. The various options should be evaluated in the context of future uncertainty. 2. Use multiple metrics to evaluate global supply chain networks. As one metric can give only part of the picture, it is beneficial to examine network design decisions using multiple metrics such as firm profits, supply chain profits, customer service levels, and response times. Good decisions perform well along most relevant metrics. 3. Use financial analysis as an input to decision making, not as the decision-making process. Financial analysis is a great tool in the decision-making process, as it often produces an answer and an abundance of quantitative data to back up that answer. However, financial methodologies alone do not provide a complete picture of the alternatives, and other nonquantifiable inputs should also be considered. 4. Use estimates along with sensitivity analysis. Many of the inputs into financial analysis are difficult, if not impossible, to obtain accurately. This can cause financial analysis to be a long and drawn-out process. One of the best ways to speed the process along and arrive at a good decision is to use estimates of inputs when it appears that finding an accurate input would take an inordinate amount of time. As we discuss in some of the other practice-oriented sections, using estimates is fine when the estimates are backed up by sensitivity analysis. It is almost always easier to come up with a range for an input than it is to come up with a single point. By performing sensitivity analysis on the input’s range, managers can often show that no matter where the true input lies within the range, the decision remains the same. When this is not the case, they have highlighted a key variable to making the decision and it likely deserves more attention to arrive at a more accurate answer. In summary, to make supply chain design decisions effectively, managers need to make estimates of inputs and then test all recommendations with sensitivity analysis. 6.8 SUMMARY OF LEARNING OBJECTIVES 1. Identify factors that need to be included in total cost when making global sourcing decisions. Besides unit cost, total cost should include the impact of global sourcing on freight, inventories, lead time, quality, on-time delivery, minimum order quantity, working capital, and stockouts. Other factors to be considered include the impact on supply chain visibility, order communication, invoicing errors, and the need for currency hedging. 2. Define uncertainties that are particularly relevant when designing global supply chains. The performance of a global supply chain is impacted by uncertainty in a number of input factors such as demand, price, exchange rates, and other economic factors. These uncertainties

Chapter 6 • Designing Global Supply Chain Networks 171 and any flexibility in the supply chain network must be taken into account when evaluating alternative designs of a supply chain. 3. Explain different strategies that may be used to mitigate risk in global supply chains. Operational strategies that help mitigate risk in global supply chains include carrying excess capacity and inventory, flexible capacity, redundant suppliers, improved responsiveness, and aggregation of demand. Hedging fuel costs and currencies are financial strategies that can help mitigate risk. It is important to keep in mind that no risk mitigation strategy will always pay off. These mitigation strategies are designed to guard against certain extreme states of the world that may arise in an uncertain global environment. 4. Understand decision tree methodologies used to evaluate supply chain design decisions under uncertainty. When valuing the streams of cash flows resulting from the performance of a supply chain, decision trees are a basic approach to valuing alternatives under uncertainty. Uncertainty along different dimensions over the evaluation period is represented as a tree with each node corresponding to a possible scenario. Starting at the last period of the evaluation interval, the decision tree analysis works back to Period 0, identifying the optimal decision and the expected cash flows at each step. Discussion Questions 6. What are the major financial uncertainties faced by an electronic components manufacturer deciding whether to 1. Why is it important to consider uncertainty when evaluating build a plant in Thailand or the United States? supply chain design decisions? 7. What are some major nonfinancial uncertainties that a company 2. What are the major sources of uncertainty that can affect the should consider when making decisions on where to source value of supply chain decisions? product? 3. Describe the basic principle of DCFs and how they can be used to compare different streams of cash flows. 4. Summarize the basic steps in the decision tree analysis methodology. 5. Discuss why using expected trends for the future can lead to different supply chain decisions relative to decision tree analysis that accounts for uncertainty. Exercises 20 percent in the second year and a 50 percent chance of remaining where they are. 1. Moon Micro is a small manufacturer of servers that currently builds its entire product in Santa Clara, California. As the Use a decision tree to determine whether Moon should market for servers has grown dramatically, the Santa Clara add capacity to its Santa Clara plant or if it should outsource plant has reached capacity of 10,000 servers per year. Moon is to Molectron. What are some other factors that would affect considering two options to increase its capacity. The first option this decision that we have not discussed? is to add 10,000 units of capacity to the Santa Clara plant at an 2. Unipart, a manufacturer of auto parts, is considering two B2B annualized fixed cost of $10,000,000 plus $500 labor per marketplaces to purchase its MRO supplies. Both marketplaces server. The second option is to have Molectron, an independent offer a full line of supplies at very similar prices for products assembler, manufacture servers for Moon at a cost of $2,000 for and shipping. Both provide similar service levels and lead times. each server (excluding raw materials cost). Raw materials cost $8,000 per server, and Moon sells each server for $15,000. However, their fee structures are quite different. The Moon must make this decision for a two-year time first marketplace, Parts4u.com, sells all of its products with a horizon. During each year, demand for Moon servers has an 5 percent commission tacked on top of the price of the product 80 percent chance of increasing 50 percent from the year (not including shipping). AllMRO.com’s pricing is based on a before and a 20 percent chance of remaining the same as the subscription fee of $10 million that must be paid up front for a year before. Molectron’s prices may change as well. They are two-year period and a commission of 1 percent on each fixed for the first year but have a 50 percent chance of increasing transaction’s product price.

172 Chapter 6 • Designing Global Supply Chain Networks 4. Bell Computer is reaching a crossroads. This PC manufacturer has been growing at a rapid rate, causing problems for its Unipart spends about $150 million on MRO supplies each operations as it tries to keep up with the surging demand. Bell year, although this varies with their utilization. Next year will executives can plainly see that within the next half year, the likely be a strong year, in which high utilization will keep MRO systems used to coordinate its supply chain are going to fall spending at $150 million. However, there is a 25 percent chance apart because they will not be able to handle the volume of that spending will drop by 10 percent. The second year, there is Bell projects they will have. a 50 percent chance that the spending level will stay where it To solve this problem, Bell has brought in two supply was in the first year and a 50 percent chance that it will drop by chain software companies that have made proposals on another 10 percent. Unipart uses a discount rate of 20 percent. systems that could cover the volume and the complexity of Assume all costs are incurred at the beginning of each year tasks Bell needs to have handled. These two software (so Year 1 costs are incurred now and Year 2 costs are incurred companies are offering different types of products, however. in a year). The first company, SCSoftware, proposes a system for which Bell will purchase a license. This will allow Bell to use Which B2B marketplace should Unipart buy its parts the software as long as it wants. However, Bell will be from? responsible for maintaining this software, which will require 3. Alphacap, a manufacturer of electronic components, is trying to significant resources. select a single supplier for the raw materials that go into its main The second company, SC–ASP, proposes that Bell pay a product, the doublecap. This is a new capacitor that is used by subscription fee on a monthly basis for SC–ASP to host Bell’s cellular phone manufacturers to protect microprocessors from supply chain applications on SC–ASP’s machines. Bell power spikes. Two companies can provide the necessary employees will access information and analysis via a Web materials—MultiChem and Mixemat. browser. Information will be fed automatically from the ASP servers to the Bell servers whenever necessary. Bell will MultiChem has a solid reputation for its products and continue to pay the monthly fee for the software, but all charges a higher price on account of its reliability of supply and maintenance will be performed by SC–ASP. delivery. MultiChem dedicates plant capacity to each customer, How should Bell go about making a decision regarding and therefore supply is assured. This allows MultiChem to which software company to choose? What are the specific charge $1.20 for the raw materials used in each doublecap. pieces of information that Bell needs to know (both about the software and about the future conditions Bell will experience) Mixemat is a small raw materials supplier that has limited in order to make a decision? What are some of the qualitative capacity but charges only $0.90 for a unit’s worth of raw issues Bell must think about when making this decision? materials. Its reliability of supply, however, is in question. Mixemat does not have enough capacity to supply all its 5. Reliable is a cell phone manufacturer serving the Asian and customers all the time. This means that orders to Mixemat are North American markets. Current annual demand of its product not guaranteed. In a year of high demand for raw materials, in Asia is 2 million, whereas the demand in North America is Mixemat will have 90,000 units available for Alphacap. 4, million. Over the next two years, demand in Asia is expected In low-demand years, all product will be delivered. to go up either by 50 percent with a probability of 0.7 by 20 percent with a probability of 0.3. Over the same period, If Alphacap does not get raw materials from suppliers, it demand in North America is expected to go up by 10 percent needs to buy them on the spot market to supply its customers. with a probability of 0.5 or go down 10 percent with a Alphacap relies on one major cell phone manufacturer for the probability of 0.5. Reliable currently has a production facility in majority of its business. Failing to deliver could lead to losing Asia with a capacity of 2.4 million units per year and a facility this contract, essentially putting the firm at risk. Therefore, in North America with a capacity of 4.2 million per year. Alphacap will buy raw material on the spot market to make up The variable production cost per phone in Asia is $15, and the for any shortfall. Spot prices for single-lot purchases (such as variable cost per phone in North America is $17. It costs $3 to Alphacap would need) are $2.00 when raw materials demand ship a phone between the two markets. Each phone sells for $40 is low and $4.00 when demand is high. in both markets. Reliable is debating whether to add 2 million units or Demand in the raw materials market has a 75 percent 1.5 million units of capacity to the Asia plant. The larger plant chance of being high each of the next two years. Alphacap increase will cost $18 million, whereas the smaller addition sold 100,000 doublecaps last year and expects to sell 110,000 will cost $15 million. Assume that Reliable uses a discount this year. However, there is a 25 percent chance it will sell only factor of 10 percent. What do you recommend? 100,000. Next year, the demand has a 75 percent chance of rising 20 percent over this year and a 25 percent chance of 6. A European apparel manufacturer has production facilities in falling 10 percent. Alphacap uses a discount rate of 20 percent. Italy and China to serve its European market, where annual Assume all costs are incurred at the beginning of each year demand is for 1.9 million units. Demand is expected to stay at (Year 1 costs are incurred now and Year 2 costs are incurred in the same level over the foreseeable future. Each facility has a a year) and that Alphacap must make a decision with a capacity of 1 million units per year. With the current exchange two-year horizon. Only one supplier can be chosen, as these two suppliers refuse to supply someone who works with their competitor. Which supplier should Alphacap choose? What other information would you like to have to make this decision?

rates, the production and distribution cost from Italy is 10 euros Chapter 6 • Designing Global Supply Chain Networks 173 per unit, whereas the production and distribution cost from China is 7 euros. Over each of the next three years, the Chinese 4 million units of capacity in North America and building currency could rise relative to the euro by 15 percent with a 2 million units of capacity in each of the two locations. probability of 0.5 or drop by 5 percent with a probability of 0.5. Building two plants will incur an additional one-time cost of An option being considered is to shut down 0.5 million units of $2 million. The variable cost of production in North America capacity in Italy and move it to China at a one-time cost of (for either a large or a small plant) is currently $10/kg, while 2 million euros. Assume a discount factor of 10 percent over the the cost in Europe is 9 euros/kg. The current exchange rate is three years. Do you recommend this option? 1 euro for US$1.33. Over each of the next three years, the 7. A chemical manufacturer is setting up capacity in Europe and dollar is expected to strengthen by 10 percent with a North America for the next three years. Annual demand in probability of 0.5 or weaken by 5 percent with a probability of each market is 2 million kilograms (kg) and is likely to stay at 0.5. Assume a discount factor of 10 percent. What should the that level. The two choices under consideration are building chemical manufacturer do? At what initial cost differential from building the two plants will the chemical manufacturer be indifferent between the two options? Bibliography Amram, Martha, and Nalin Kulatilaka. Real Options. Cambridge, Johnson, Norman L., Samuel Kotz, and N. Balakrishnan. MA: Harvard Business School Press, 1999. Continuous Univariate Distributions. New York: John Wiley & Sons, 1994. Bovet, David. “The Supply Chain Manager as Global Economist.” Supply Chain Management Review (September 2008): 17–24. Jordan, W. C., and S. C. Graves. “Principles on the Benefits of Manufacturing Process Flexibility.” Management Science 41 Brealey, Richard A., and Stewart C. Myers. Principles of (April 1995): 577–594. Corporate Finance. New York: McGraw-Hill, 1996. Lim, Michael, Achal Bassamboo, Sunil Chopra, and Mark Chang, Shih-Chia, Neng-Pai Lin, and Chwen Sheu. “Aligning S. Daskin. “Flexibility and Fragility: Supply Chain Network Manufacturing Flexibility with Environmental Uncertainty: Design with Disruption Risks” (working paper, Northwestern Evidence from High-Technology Component Manufacturers University, Evanston, IL, 2008). in Taiwan.” International Journal of Production Research 40, no. 18 (2002): 4765–4780. Luehrman, Timothy A. “Capital Projects as Real Options: An Introduction.” Harvard Business School Case 9–295–074, 1995. Chopra, Sunil, and ManMohan S. Sodhi. “Managing Risk to Avoid Supply Chain Breakdown.” Sloan Management Review 46, Luehrman, Timothy A. “Investment Opportunities as Real no. 1 (2004): 53–61. Options: Getting Started on the Numbers.” Harvard Business Review (July–August 1998): 51–67. De Treville, Suzanne, and Lenos Trigeorgis. “It May Be Cheaper to Manufacture at Home.” Harvard Business Review (October Luehrman, Timothy A. “Strategy as a Portfolio of Real Options.” 2010): 84–87. Harvard Business Review (September–October 1998): 89–99. Farrell, Diana. “Beyond Offshoring: Assess Your Company’s Global Potential.” Harvard Business Review (December 2004): Ross, Sheldon M. Introduction to Stochastic Dynamic 82–90. Programming. New York: Academic Press, 1983. Farreira, John, and Len Prokopets. “Does Offshoring Still Make Sheffi, Y. The Resilient Enterprise. Cambridge: The MIT Press, Sense?” Supply Chain Management Review (January/February 2005. 2009): 20–27. Smith, Adam. An Inquiry into the Nature and Causes of the Favre, Donavon, and John McCreery. “Coming to Grips with Wealth of Nations, 5th ed. London: Methuen & Co., Ltd., Rising Supplier Risk.” Supply Chain Management Review 1904. (September 2008): 26–32. Stokey, Nancy L., Robert E. Lucas Jr., and Edward C. Prescott. Garber, Randy, and Suman Sarkar. “Want a More Flexible Supply Recursive Methods in Economic Dynamics. Cambridge, MA: Chain?” Supply Chain Management Review (January/February Harvard University Press, 1989. 2007): 28–34. Swaminathan, Jayashankar M., and Brian Tomlin. “How to Avoid Goel, Ajay, Nazgol Moussavi, and Vats N. Srivatsan. “Time to the Six Risk Management Pitfalls.” Supply Chain Management Rethink Offshoring?” McKinsey on Business Technology 14 Review (July/August 2007): 34–42. (Winter 2008): 32–35. Tanowitz, Marc, and David Rutchik. “Squeezing Opportunity Out Harding, Mary Lu. “Gauging Total Cost, Supplier by Supplier.” of Higher Fuel Costs.” Supply Chain Management Review CSCMP’s Supply Chain Quarterly (Q4 2007): 64–68. (October 2008): 34–40. Horngren, Charles T., George Foster, and Srikant M. Datar. Cost Trigeorgis, Lenos. Real Options. Cambridge, MA: The MIT Accounting. Upper Saddle River, NJ: Prentice Hall, 1997. Press, 1996.

174 Chapter 6 • Designing Global Supply Chain Networks CASE STUDY For 2009, sales of each product by region and the production and capacity at each plant are shown in BioPharma, Inc.1 Table 6-18. The plant capacity, measured in millions of kilograms of production, can be assigned to either In 2009, Phillip (Phil) Landgraf faced several glaring chemical as long as the plant is capable of producing problems in the financial performance of his company, both. BioPharma has forecast that its sales for the two BioPharma, Inc. The firm had experienced a steep chemicals are likely to be stable for all parts of the decline in profits and very high costs at its plants in world, except for Asia without Japan, where sales are Germany and Japan. Landgraf, the company’s president expected to grow by 10 percent annually for each of the for worldwide operations, knew that demand for the next five years before stabilizing. company’s products was stable across the globe. As a result, the surplus capacity in his global production The Japanese plant is a technology leader within network looked like a luxury he could no longer afford. the BioPharma network in terms of its ability to handle regulatory and environmental issues. Some develop- Any improvement in financial performance was ments in the Japanese plant had been transferred to other dependent on having the most efficient network in place, plants in the network. The German plant is a leader in because revenues were unlikely to grow. Cutting costs terms of its production ability. The plant has routinely was thus a top priority for the coming year. To help had the highest yields within the global network. The design a more cost-effective network, Landgraf assigned Brazilian, Indian, and Mexican plants have somewhat a task force to recommend a course of action. outdated technology and are in need of an update. Background Current Plant Costs at BioPharma BioPharma, Inc., is a global manufacturer of bulk After considerable debate, the task force identified the cost chemicals used in the pharmaceutical industry. The structure at each plant in 2009 as shown in Table 6-19. company holds patents on two chemicals that are called Each plant incurs an annual fixed cost that is independent Highcal and Relax internally. These bulk chemicals are of the level of production in the plant. The fixed cost used internally by the company’s pharmaceutical includes depreciation, utilities, and the salaries and fringe division and are also sold to other drug manufacturers. benefits of employees involved in general management, There are distinctions in the precise chemical specifica- tions to be met in different parts of the world. All plants, however, are currently set up to be able to produce both chemicals for any part of the world. . Table 6-18 Sales by Region and Production/Capacity by Plant of Highcal and Relax in Million Kilograms Highcal Relax Region Plant Capacity 2009 2009 2009 2009 Latin America Brazil 18.0 Sales Production Sales Production Europe Germany 45.0 Asia w/o Japan India 18.0 7.0 11.0 7.0 7.0 Japan Japan 10.0 15.0 15.0 12.0 0.0 Mexico Mexico 30.0 10.0 8.0 U.S. U.S. 22.0 5.0 3.0 0.0 7.0 2.0 8.0 18.0 3.0 12.0 3.0 17.0 18.0 17.0 5.0 1 This case was inspired by Applichem (A), Harvard Business School Case 9–685–051, 1985.

Chapter 6 • Designing Global Supply Chain Networks 175 Table 6-19 Fixed and Variable Production Costs at Each BioPharma Plant in 2009 (US$) Highcal Relax Plant Plant Highcal Relax Raw Production Raw Production Brazil Fixed Cost Fixed Cost Fixed Cost Material cost Material cost Germany (million $) (million $) (million $) ($/kg) ($/kg) India ($/kg) 5.1 ($/kg) 6.6 Japan 20.0 5.0 5.0 3.6 7.0 4.6 8.5 Mexico 45.0 13.0 14.0 3.9 4.5 5.0 6.0 U.S. 18.0 3.6 7.5 4.5 9.0 17.0 4.0 4.0 3.9 5.0 5.1 6.5 30.0 6.0 6.0 3.6 5.0 4.6 6.5 21.0 6.0 6.0 3.6 4.5 5.0 5.0 scheduling, expediting, accounting, maintenance, and so BioPharma transports the chemicals in specialized forth. Each plant that is capable of producing either containers by sea and in specialized trucks on land. Highcal or Relax also incurs a product-related fixed cost The transportation costs between plants and markets are as that is independent of the quantity of each chemical shown in Table 6-20. Historical exchange rates are shown produced. The product-related fixed cost includes depreci- in Table 6-21 and the regional import duties in Table 6-22. ation of equipment specific to a chemical and other fixed Given regional trade alliances, import duties in reality vary costs that are specific to a chemical. If a plant maintains based on the origin of the chemical. For simplicity’s sake, the capability to produce a particular chemical, it incurs however, the task force has assumed that the duties are the corresponding product-related fixed cost even if the driven only by the destination. Local production within chemical is not produced at the plant. each region is assumed to result in no import duty. Thus, production from Brazil, Germany, and India can be sent to The variable production cost of each chemical Latin America, Europe, and the rest of Asia excluding consists of two components: raw materials and production Japan, respectively, without incurring any import duties. costs. The variable production cost is incurred in propor- Duties apply only to the raw material, production, and tion to the quantity of chemical produced and includes transportation cost component and not to the fixed cost direct labor and scrap. The plants themselves can handle component. Thus, a product entering Latin America with a varying levels of production. In fact, they can also be idled raw material, production, and transportation cost of $10 for the year, in which case they incur only the fixed cost, incurs import duties of $3. none of the variable cost. Table 6-20 Transportation Costs from Plants to Markets (US$/kg) From/To Latin America Europe Asia w/o Japan Japan Mexico U.S. Brazil 0.20 0.45 0.50 0.50 0.40 0.45 Germany 0.45 0.20 0.35 0.40 0.30 0.30 India 0.50 0.35 0.20 0.30 0.50 0.45 Japan 0.50 0.40 0.30 0.10 0.45 0.45 Mexico 0.40 0.30 0.50 0.45 0.20 0.25 U.S. 0.45 0.30 0.45 0.45 0.25 0.20 (continued)

176 Chapter 6 • Designing Global Supply Chain Networks (continued) Table 6-21 History of Exchange Rates in Currency/US$1 (at the Beginning of Each Year) Brazilian Real Euro Indian Rupee Japanese Yen Mexican Peso U.S. Dollar 2009 2.70 0.74 43.47 103.11 11.21 1.00 2008 2.90 0.80 45.60 107.00 11.22 1.00 2007 3.50 0.96 48.00 119.25 10.38 1.00 2006 2.30 1.11 48.27 131.76 1.00 2005 1.95 1.06 46.75 114.73 9.12 1.00 2004 1.81 0.99 43.55 102.33 9.72 1.00 9.48 Table 6-22 Import Tariffs (Percentage of Value of Product Imported, Including Transportation) Latin America Europe Asia w/o Japan Japan Mexico U.S. 30% 3% 27% 6% 35% 4% Network Options Under Consideration Questions The task force is considering a variety of options for 1. How should BioPharma have used its production network its analysis. One option is to keep the global network in 2009? Should any of the plants have been idled? What is with its current structure and capabilities. Other the annual cost of your proposal, including import duties? options include shutting down some plants or limiting the capability of some plants to producing only one 2. How should Landgraf structure his global production chemical. Closing down a plant eliminates all variable network? Assume that the past is a reasonable indicator of costs and saves 80 percent of the annual fixed costs the future in terms of exchange rates. (the remaining 20 percent accounts for costs that are incurred related to the plant shutdown). Similarly, if a 3. Is there any plant for which it may be worth adding a plant is limited to producing only one chemical, the million kilograms of additional capacity at a fixed cost of plant saves 80 percent of the fixed cost associated with $3 million per year? the chemical that is no longer produced. The two options being seriously considered are shutting the 4. How are your recommendations affected by the reduction Japanese plant and limiting the German plant to a of duties? single chemical. 5. The analysis has assumed that each plant has a 100 percent yield (percent output of acceptable quality). How would you modify your analysis to account for yield differences across plants? 6. What other factors should be accounted for when making your recommendations? CASE STUDY The Sourcing Decision at Forever Young Forever Young is a retailer of trendy and low-cost The company has historically outsourced production to apparel in the United States. The company divides the China given the lower costs. Sourcing from the Chinese year into four sales seasons of about three months each supplier costs 55 yuan/unit (inclusive of all delivery and brings in new merchandise for each season. costs), which at the current exchange rate of 6.5 yuan/

Chapter 6 • Designing Global Supply Chain Networks 177 $ gives a variable cost of under $8.50/unit. The Chinese The short lead time of the local supplier allows supplier, however, has a long lead time, forcing Forever Forever Young to keep bringing product in a little bit at a Young to pick an order size well before the start of the time based on actual sales. Thus, if the local supplier is season. This does not leave the company any flexibility used, the company is able to meet all demand in each if actual demand differs from the order size. period without having any unsold inventory or lost sales. In other words, the final order from the local supplier A local supplier has come to management with a will exactly equal the demand observed by Forever proposal to supply product at a cost of $10/unit but do so Young. quickly enough that Forever Young will be able to make supply in the season exactly match demand. A Potential Hybrid Strategy Management is concerned about the higher variable cost but finds the flexibility of the onshore supplier very The local supplier has also offered another proposal that attractive. The challenge is to value the responsiveness would allow Forever Young to use both suppliers, each provided by the local supplier. playing a different role. The Chinese supplier would produce a base quantity for the season and the local Uncertainties Faced by Forever Young supplier would cover any shortfalls that result. The short lead time of the local supplier would ensure that no sales To better compare the two suppliers, management are lost. In other words, if Forever Young committed to a identifies demand and exchange rates as the two major base load of 900 units with the Chinese supplier in a uncertainties faced by the company. Over each of the given period and demand was 900 units or less, nothing next two periods (assume them to be a year each), would be ordered from the local supplier. If demand, demand may go up by 10 percent with a probability of however, was larger than 900 units (say 1,100), the 0.5 or down by 10 percent with a probability of 0.5. shortfall of 200 units would be supplied by the local Demand in the current period was 1,000 units. Similarly, supplier. Under a hybrid strategy, the local supplier over each of the next two periods, the yuan may would end up supplying only a small fraction of the strengthen by 5 percent with a probability of 0.5 or season’s demand. For this extra flexibility and reduced weaken by 5 percent with a probability of 0.5. The volumes, however, the local supplier proposes to charge exchange rate in the current period was 6.5 yuan/$. $11/unit if she is used as part of a hybrid strategy. Ordering Policies with the Two Suppliers Questions Given the long lead time of the offshore supplier, 1. Draw a decision tree reflecting the uncertainty over the Forever Young commits to an order before observing next two periods. Identify each node in terms of demand any demand signal. Given the demand uncertainty over and exchange rate and the transition probabilities. the next two periods and the fact that the margin from each unit (margin of about $11.50) is higher than the 2. If management at Forever Young is to pick only one of the loss if the unit remains unsold at the end of the season two suppliers, which one would you recommend? What is (loss of about $8.50), management decides to commit to the NPV of expected profit over the next two periods for an order that is somewhat higher than expected demand. each of the two choices? Assume a discount factor of Given that expected demand is 1,000 units over each of k ϭ 0.1 per period. the next two periods, management decides to order 1,040 units from the Chinese supplier for each of the 3. What do you think about the hybrid approach? Is it next two periods. If demand in a period turns out to be worth paying the local supplier extra to use her as part higher than 1,040 units, Forever Young will sell 1,040 of a hybrid strategy? For the hybrid approach, assume units. However, if demand turns out to be lower than that management will order a base load of 900 units 1,040, the company will have left over product for which from the Chinese supplier for each of the two periods, it will not be able to recover any revenue. making up any shortfall in each period at the local supplier. Evaluate the NPV of expected profits for the hybrid option assuming a discount factor of k ϭ 0.1 per period.

7 {{{ Demand Forecasting in a Supply Chain LEARNING OBJECTIVES After reading this chapter, you will be able to 1. Understand the role of forecasting for both an enterprise and a supply chain. 2. Identify the components of a demand forecast. 3. Forecast demand in a supply chain given historical demand data using time-series methodologies. 4. Analyze demand forecasts to estimate forecast error. All supply chain decisions made before demand has materialized are made to a forecast. In this chapter, we explain how historical demand information can be used to forecast future demand and how these forecasts affect the supply chain. We describe several methods to forecast demand and estimate a forecast’s accuracy. We then discuss how these methods can be implemented using Microsoft Excel. 7.1 THE ROLE OF FORECASTING IN A SUPPLY CHAIN Demand forecasts form the basis of all supply chain planning. Consider the push/pull view of the supply chain discussed in Chapter 1. All push processes in the supply chain are performed in anticipation of customer demand, whereas all pull processes are performed in response to customer demand. For push processes, a manager must plan the level of activity, be it production, transportation, or any other planned activity. For pull processes, a manager must plan the level of available capacity and inventory but not the actual amount to be executed. In both instances, the first step a manager must take is to forecast what customer demand will be. A Home Depot store selling paint orders the base paint and dyes in anticipation of customer orders, whereas it performs final mixing of the paint in response to customer orders. Home Depot uses a forecast of future demand to determine the quantity of paint and dye to have on hand (a push process). Farther up the supply chain, the paint factory that produces the base also needs forecasts to determine its own production and inventory levels. The paint factory’s suppliers also need forecasts for the same reason. When each stage in the supply chain makes its own separate forecast, these forecasts are often very different. The result is a mismatch between supply and demand. When all stages of a supply chain work together to produce a collaborative forecast, it tends to be much more accurate. The resulting forecast accuracy enables supply chains to be both more responsive and more efficient in serving their customers. Leaders in many supply chains, from PC manufacturers to packaged-goods retailers, have improved their ability to match supply and demand by moving toward collaborative forecasting. Consider the value of collaborative forecasting for Coca-Cola and its bottlers. Coca-Cola decides on the timing of various promotions based on the demand forecast over the coming quarter. Promotion decisions are then incorporated into an updated demand forecast. The updated forecast is essential for the bottlers to plan their capacity and production decisions. A bottler operating without an updated forecast based on the promotion is unlikely to have sufficient supply available for Coca-Cola, thus hurting supply chain profits. 178

Chapter 7 • Demand Forecasting in a Supply Chain 179 Mature products with stable demand, such as milk or paper towels, are usually easiest to forecast. Forecasting and the accompanying managerial decisions are extremely difficult when either the supply of raw materials or the demand for the finished product is highly unpredictable. Fashion goods and many high-tech products are examples of items that are difficult to forecast. In both instances, an estimate of forecast error is essential when designing the supply chain and planning its response. Before we begin an in-depth discussion of the components of forecasts and forecasting methods in the supply chain, we briefly list characteristics of forecasts that a manager must understand to design and manage his or her supply chain effectively. 7.2 CHARACTERISTICS OF FORECASTS Companies and supply chain managers should be aware of the following characteristics of forecasts. 1. Forecasts are always inaccurate and should thus include both the expected value of the forecast and a measure of forecast error. To understand the importance of forecast error, consider two car dealers. One of them expects sales to range between 100 and 1,900 units, whereas the other expects sales to range between 900 and 1,100 units. Even though both dealers anticipate average sales of 1,000, the sourcing policies for each dealer should be very different given the difference in forecast accuracy. Thus, the forecast error (or demand uncertainty) must be a key input into most supply chain decisions. Unfortunately, most firms do not maintain any estimate of forecast error. 2. Long-term forecasts are usually less accurate than short-term forecasts; that is, long-term forecasts have a larger standard deviation of error relative to the mean than short-term forecasts. Seven-Eleven Japan has exploited this key property to improve its performance. The company has instituted a replenishment process that enables it to respond to an order within hours. For example, if a store manager places an order by 10 A.M., the order is delivered by 7 P.M. the same day. Therefore, the manager only has to forecast what will sell that night less than 12 hours before the actual sale. The short lead time allows a manager to take into account current information that could affect product sales, such as the weather. This forecast is likely to be more accurate than if the store manager had to forecast demand one week in advance. 3. Aggregate forecasts are usually more accurate than disaggregate forecasts, as they tend to have a smaller standard deviation of error relative to the mean. For example, it is easy to forecast the gross domestic product (GDP) of the United States for a given year with less than a 2 percent error. However, it is much more difficult to forecast yearly revenue for a company with less than a 2 percent error, and it is even harder to forecast revenue for a given product with the same degree of accuracy. The key difference among the three forecasts is the degree of aggregation. The GDP is an aggregation across many companies, and the earnings of a company are an aggregation across several product lines. The greater the aggregation, the more accurate is the forecast. 4. In general, the farther up the supply chain a company is (or the farther it is from the consumer), the greater is the distortion of information it receives. One classic example of this is the bullwhip effect (see Chapter 10), in which order variation is amplified as orders move farther from the end customer. As a result, the farther up the supply chain an enterprise is, the larger is the forecast error. Collaborative forecasting based on sales to the end customer helps upstream enterprises reduce forecast error. In the next section, we discuss the basic components of a forecast, explain the four classi- fications into which forecasting methods fall, and introduce the notion of forecast error.

180 Chapter 7 • Demand Forecasting in a Supply Chain 7.3 COMPONENTS OF A FORECAST AND FORECASTING METHODS Yogi Berra, the former New York Yankees catcher who is famous for his malapropisms, once said, “Predictions are usually difficult, especially about the future.” One may be tempted to treat demand forecasting as magic or art and leave everything to chance. What a firm knows about its customers’ past behavior, however, sheds light on their future behavior. Demand does not arise in a vacuum. Rather, customer demand is influenced by a variety of factors and can be predicted, at least with some probability, if a company can determine the relationship between these factors and future demand. To forecast demand, companies must first identify the factors that influence future demand and then ascertain the relationship between these factors and future demand. Companies must balance objective and subjective factors when forecasting demand. Although we focus on quantitative forecasting methods in this chapter, companies must include human input when they make their final forecast. Seven-Eleven Japan illustrates this point. Seven-Eleven Japan provides its store managers with a state-of-the-art decision support system that makes a demand forecast and provides a recommended order. The store manager, however, is responsible for making the final decision and placing the order, because he or she may have access to information about market conditions that are not available in historical demand data. This knowledge of market conditions is likely to improve the forecast. For example, if the store manager knows that the weather is likely to be rainy and cold the next day, he or she can reduce the size of an ice cream order to be placed with an upstream supplier, even if demand was high during the previous few days when the weather was hot. In this instance, a change in market conditions (the weather) would not have been predicted using historical demand data. A supply chain can experience substantial payoffs from improving its demand forecasting through qualitative human inputs. A company must be knowledgeable about numerous factors that are related to the demand forecast, including the following: • Past demand • Lead time of product replenishment • Planned advertising or marketing efforts • Planned price discounts • State of the economy • Actions that competitors have taken A company must understand such factors before it can select an appropriate forecasting methodology. For example, historically a firm may have experienced low demand for chicken noodle soup in July and high demand in December and January. If the firm decides to discount the product in July, the situation is likely to change, with some of the future demand shifting to the month of July. The firm should make its forecast taking this factor into consideration. Forecasting methods are classified according to the following four types: 1. Qualitative: Qualitative forecasting methods are primarily subjective and rely on human judgment. They are most appropriate when little historical data are available or when experts have market intelligence that may affect the forecast. Such methods may also be necessary to forecast demand several years into the future in a new industry. 2. Time series: Time-series forecasting methods use historical demand to make a forecast. They are based on the assumption that past demand history is a good indicator of future demand. These methods are most appropriate when the basic demand pattern does not vary significantly from one year to the next. These are the simplest methods to implement and can serve as a good starting point for a demand forecast. 3. Causal: Causal forecasting methods assume that the demand forecast is highly corre- lated with certain factors in the environment (the state of the economy, interest rates, etc.). Causal forecasting methods find this correlation between demand and environmental factors and use estimates of what environmental factors will be to forecast future demand. For example,

Chapter 7 • Demand Forecasting in a Supply Chain 181 product pricing is strongly correlated with demand. Companies can thus use causal methods to determine the impact of price promotions on demand. 4. Simulation: Simulation forecasting methods imitate the consumer choices that give rise to demand to arrive at a forecast. Using simulation, a firm can combine time-series and causal methods to answer such questions as: What will be the impact of a price promotion? What will be the impact of a competitor opening a store nearby? Airlines simulate customer buying behavior to forecast demand for higher fare seats when there are no seats available at the lower fares. A company may find it difficult to decide which method is most appropriate for forecasting. In fact, several studies have indicated that using multiple forecasting methods to create a combined forecast is more effective than using any one method alone. In this chapter, we deal primarily with time-series methods, which are most appropriate when future demand is related to historical demand, growth patterns, and any seasonal patterns. With any forecasting method, there is always a random element that cannot be explained by his- torical demand patterns. Therefore, any observed demand can be broken down into a systematic and a random component: Observed demand 1O2 = systematic component 1S2 + random component 1R2 The systematic component measures the expected value of demand and consists of what we will call level, the current deseasonalized demand; trend, the rate of growth or decline in demand for the next period; and seasonality, the predictable seasonal fluctuations in demand. The random component is that part of the forecast that deviates from the systematic part. A company cannot (and should not) forecast the direction of the random component. All a company can predict is the random component’s size and variability, which provides a measure of forecast error. The objective of forecasting is to filter out the random component (noise) and estimate the systematic component. The forecast error measures the difference between the forecast and actual demand. On average, a good forecasting method has an error whose size is comparable to the random component of demand. A manager should be skeptical of a forecasting method that claims to have no forecasting error on historical demand. In this case, the method has merged the historical random component with the systematic component. As a result, the forecasting method will likely perform poorly. 7.4 BASIC APPROACH TO DEMAND FORECASTING The following five points are important for an organization to forecast effectively: 1. Understand the objective of forecasting. 2. Integrate demand planning and forecasting throughout the supply chain. 3. Identify the major factors that influence the demand forecast. 4. Forecast at the appropriate level of aggregation. 5. Establish performance and error measures for the forecast. Understand the Objective of Forecasting Every forecast supports decisions that are based on it, so an important first step is to identify these decisions clearly. Examples of such decisions include how much of a particular product to make, how much to inventory, and how much to order. All parties affected by a supply chain decision should be aware of the link between the decision and the forecast. For example, Wal-Mart’s plans to discount detergent during the month of July must be shared with the manufacturer, the transporter, and others involved in filling demand, as they all must make decisions that are affected by the forecast of demand. All parties should come up with a common forecast for the promotion and a shared plan of action based on the forecast. Failure to make these decisions jointly may result in either too much or too little product in various stages of the supply chain.

182 Chapter 7 • Demand Forecasting in a Supply Chain Integrate Demand Planning and Forecasting Throughout the Supply Chain A company should link its forecast to all planning activities throughout the supply chain. These include capacity planning, production planning, promotion planning, and purchasing, among others. In one unfortunately common scenario, a retailer develops forecasts based on promotional activities, whereas a manufacturer, unaware of these promotions, develops a different forecast for its production planning based on historical orders. This leads to a mismatch between supply and demand, resulting in poor customer service. To accomplish integration, it is a good idea for a firm to have a cross-functional team, with members from each affected function responsible for forecasting demand—and an even better idea is to have members of different companies in the supply chain working together to create a forecast. Identify Major Factors That Influence the Demand Forecast Next, a firm must identify demand, supply, and product-related phenomena that influence the demand forecast. On the demand side, a company must ascertain whether demand is growing or declining or has a seasonal pattern. These estimates must be based on demand—not sales data. For example, a supermarket promoted a certain brand of cereal in July 2011. As a result, the demand for this cereal was high while the demand for other, comparable cereal brands was low in July. The supermarket should not use the sales data from 2011 to estimate that demand for this brand will be high in July 2012, because this will occur only if the same brand is promoted again in July 2012 and other brands respond as they did the previous year. When making the demand forecast, the supermarket must understand what the demand would have been in the absence of promotion activity and how demand is affected by promotions and competitor actions. A combination of these pieces of information will allow the supermarket to forecast demand for July 2012 given the promotion activity planned for that year. On the supply side, a company must consider the available supply sources to decide on the accuracy of the forecast desired. If alternate supply sources with short lead times are available, a highly accurate forecast may not be especially important. However, if only a single supplier with a long lead time is available, an accurate forecast will have great value. On the product side, a firm must know the number of variants of a product being sold and whether these variants substitute for or complement one another. If demand for a product influences or is influenced by demand for another product, the two forecasts are best made jointly. For example, when a firm introduces an improved version of an existing product, it is likely that the demand for the existing product will decline because customers will buy the improved version. Although the decline in demand for the original product is not indicated by historical data, the historical demand is still useful in that it allows the firm to estimate the combined total demand for the two versions. Clearly, demand for the two products should be forecast jointly. Forecast at the Appropriate Level of Aggregation Given that aggregate forecasts are more accurate than disaggregate forecasts, it is important to forecast at a level of aggregation that is appropriate given the supply chain decision that is driven by the forecast. Consider a buyer at a retail chain who is forecasting in order to pick an order size for shirts. One approach is to ask each store manager the precise number of shirts needed and add up all the requests to get an order size with the supplier. The advantage of this approach is that it uses local market intelligence that each store manager has. The problem with this approach is that it makes store managers forecast well before demand arises at a time when their disaggregate forecasts are unlikely to be accurate. A better approach may be to forecast demand at the aggregate level when ordering with the supplier and ask each store manager to forecast only when the shirts are to be allocated across the stores. In this case, the long lead time forecast (supplier order) is aggregate, thus lowering error. The disaggregate store-level forecast is made close to the sales season when local market intelligence is likely to be most effective.

Chapter 7 • Demand Forecasting in a Supply Chain 183 Establish Performance and Error Measures for the Forecast Companies should establish clear performance measures to evaluate the accuracy and timeliness of the forecast. These measures should be highly correlated with the objectives of the business decisions based on these forecasts. Consider a mail-order company that uses a forecast to place orders with its suppliers up the supply chain. Suppliers take two months to send in the orders. The mail-order company must ensure that the forecast is created at least two months before the start of the sales season because of the two-month lead time for replenishment. At the end of the sales season, the company must compare actual demand to forecasted demand to estimate the accuracy of the forecast. Then plans for decreasing future forecast errors or responding to the observed forecast errors can be put into place. In the next section, we discuss techniques for static and adaptive time-series forecasting. 7.5 TIME-SERIES FORECASTING METHODS The goal of any forecasting method is to predict the systematic component of demand and estimate the random component. In its most general form, the systematic component of demand data contains a level, a trend, and a seasonal factor. The equation for calculating the systematic component may take a variety of forms: • Multiplicative: Systematic component = level * trend * seasonal factor • Additive: Systematic component = level + trend + seasonal factor • Mixed: Systematic component = (level + trend) * seasonal factor The specific form of the systematic component applicable to a given forecast depends on the nature of demand. Companies may develop both static and adaptive forecasting methods for each form. We now describe these static and adaptive forecasting methods. Static Methods A static method assumes that the estimates of level, trend, and seasonality within the systematic component do not vary as new demand is observed. In this case, we estimate each of these parameters based on historical data and then use the same values for all future forecasts. In this section, we discuss a static forecasting method for use when demand has a trend as well as a seasonal component. We assume that the systematic component of demand is mixed, that is, Systematic component = (level + trend) * seasonal factor A similar approach can be applied for other forms as well. We begin with a few basic definitions: L ϭ estimate of level at t ϭ 0 (the deseasonalized demand estimate during Period t ϭ 0) T ϭ estimate of trend (increase or decrease in demand per period) St ϭ estimate of seasonal factor for Period t Dt ϭ actual demand observed in Period t Ft ϭ forecast of demand for Period t In a static forecasting method, the forecast in Period t for demand in Period t ϩ l is a prod- uct of the level in Period t ϩ l and the seasonal factor for Period t ϩ l. The level in Period t ϩ l is the sum of the level in Period 0 (L) and (t ϩ l) times the trend T. The forecast in Period t for demand in Period t ϩ l is thus given as Ft+l = [L + (t + l)T]St+l (7.1)

184 Chapter 7 • Demand Forecasting in a Supply Chain Table 7-1 Quarterly Demand for Tahoe Salt Year Quarter Period, t Demand, Dt 1 2 1 8,000 13 2 13,000 14 3 23,000 21 4 34,000 22 5 10,000 23 6 18,000 24 7 23,000 31 8 38,000 32 9 12,000 3 3 10 13,000 3 4 11 32,000 4 1 12 41,000 We now describe one method for estimating the three parameters L, T, and S. As an example, consider the demand for rock salt used primarily to melt snow. This salt is produced by a firm called Tahoe Salt, which sells its salt through a variety of independent retailers around the Lake Tahoe area of the Sierra Nevada Mountains. In the past, Tahoe Salt has relied on estimates of demand from a sample of its retailers, but the company has noticed that these retailers always overestimate their purchases, leaving Tahoe (and even some retailers) stuck with excess inventory. After meeting with its retailers, Tahoe has decided to produce a collaborative forecast. Tahoe Salt wants to work with the retailers to create a more accurate forecast based on the actual retail sales of their salt. Quarterly retail demand data for the past three years are shown in Table 7-1 and charted in Figure 7-1. In Figure 7-1, observe that demand for salt is seasonal, increasing from the second quarter of a given year to the first quarter of the following year. The second quarter of each year has the lowest demand. Each cycle lasts four quarters, and the demand pattern repeats every year. There is also a growth trend in the demand, with sales growing over the past three years. The company estimates that growth will continue in the coming year at historical rates. We now describe how each of the three parameters—level, trend, and seasonal factors—may be estimated. The following two steps are necessary to making this estimation: 1. Deseasonalize demand and run linear regression to estimate level and trend. 2. Estimate seasonal factors. 50,000Demand 40,000 30,000 20,000 10,000 0 1, 2 1, 3 1, 4 2, 1 2, 2 2, 3 2, 4 3, 1 3, 2 3, 3 3, 4 4, 1 Period FIGURE 7-1 Quarterly Demand at Tahoe Salt

Chapter 7 • Demand Forecasting in a Supply Chain 185 ESTIMATING LEVEL AND TREND The objective of this step is to estimate the level at Period 0 and the trend. We start by deseasonalizing the demand data. Deseasonalized demand represents the demand that would have been observed in the absence of seasonal fluctuations. The periodicity (p) is the number of periods after which the seasonal cycle repeats. For Tahoe Salt’s demand, the pattern repeats every year. Given that we are measuring demand on a quarterly basis, the periodicity for the demand in Table 7-1 is p ϭ 4. To ensure that each season is given equal weight when deseasonalizing demand, we take the average of p consecutive periods of demand. The average of demand from Period l ϩ 1 to Period l ϩ p provides deseasonalized demand for Period l ϩ (p ϩ 1)/2. If p is odd, this method provides deseasonalized demand for an existing period. If p is even, this method provides deseasonalized demand at a point between Period l ϩ (p/2) and Period l ϩ 1 ϩ (p/2). By taking the average of deaseasonalized demand provided by Periods l ϩ 1 to l ϩ p and l ϩ 2 to l ϩ p ϩ 1, we obtain the deseasonalized demand for Period l ϩ 1 ϩ (p/2). This procedure for obtaining the deseasonalized demand, D-t, for Period t, is formulated as follows: t - 1 + (p/2) D-t c Dt - (p/2) + Dt+(p/2) + a 2Di d n (2p) for p even e = i = t + 1 - (p/2) (7.2) t + [(p - 1)/2] a Di /p for p odd i = t - [(p - 1)/2] In our example, p ϭ 4 is even. For t ϭ 3, we obtain the deseasonalized demand using Equation 7.2 as follows: D-3 = c Dt - 1p>22 + Dt + 1p>22 + t - 1 + 1p>22 4 a 2Di d n (2 p) = D1 + D5 + a 2Di n8 i = t + 1 - 1p>22 i=2 With this procedure, we can obtain deseasonalized demand between Periods 3 and 10 as shown in Figures 7-2 and 7-3. The following linear relationship exists between the deseasonalized demand, D-t, and time t, based on the change in demand over time. D-t = L + Tt (7.3) Cell Cell Formula Equation Copied to C4 =(B2+B6+2*SUM(B3:B5))/8 7.2 C5:C11 FIGURE 7-2 Excel Workbook with Deseasonalized Demand for Tahoe Salt

186 Chapter 7 • Demand Forecasting in a Supply Chain 50,000 Actual Demand Deseasonalized 40,000 Demand 30,000 Demand 20,000 10,000 01 2 3 4 5 6 7 8 9 10 11 12 Period FIGURE 7-3 Deseasonalized Demand for Tahoe Salt Note that in Equation 7.3, D-t represents deseasonalized demand and not the actual demand in Period t, L represents the level or deseasonalized demand at Period 0, and T represents the rate of growth of deseasonalized demand or trend. We can estimate the values of L and T for the deseasonalized demand using linear regression with deseasonalized demand (see Figure 7-2) as the dependent variable and time as the independent variable. Such a regression can be run using Microsoft Excel (Data | Data Analysis | Regression). This sequence of commands opens the Regression dialog box in Excel. For the Tahoe Salt workbook in Figure 7-2, in the resulting dialog box, we enter Input Y Range: C4:C11 Input X Range: A4:A11 and click the OK button. A new sheet containing the results of the regression opens up. This new sheet contains estimates for both the initial level L and the trend T. The initial level, L, is obtained as the intercept coefficient, and the trend, T, is obtained as the X variable coefficient (or the slope) from the sheet containing the regression results. For the Tahoe Salt example, we obtain L ϭ 18,439 and T ϭ 524. For this example, deseasonalized demand D-t for any Period t is thus given by D-t = 18,439 + 524t (7.4) Note that it is not appropriate to run a linear regression between the original demand data and time to estimate level and trend because the original demand data are not linear and the resulting linear regression will not be accurate. The demand must be deseasonalized before we run the linear regression. ESTIMATING SEASONAL FACTORS We can now obtain deseasonalized demand for each dpeesrieoadsounsainligzeEdqdueamtioannd7.4D-.tTahnedsiesagsoivneanl factor St for Period t is the ratio of actual demand Dt to as S-t = Di (7.5) D-t For the Tahoe Salt example, the deseasonalized demand estimated using Equation 7.4 and the seasonal factors estimated using Equation 7.5 are shown in Figure 7-4. Given the periodicity, p, we obtain the seasonal factor for a given period by averaging seasonal factors that correspond to similar periods. For example, if we have a periodicity of p ϭ 4, Periods 1, 5, and 9 have similar seasonal factors. The seasonal factor for these periods is obtained

Chapter 7 • Demand Forecasting in a Supply Chain 187 Cell Cell Formula Equation Copied to C3:C13 C2 =18439+A2*524 7.4 D3:D13 D2 =B2/C2 7.5 FIGURE 7-4 Deseasonalized Demand and Seasonal Factors for Tahoe Salt as the average of the three seasonal factors. Given r seasonal cycles in the data, for all periods of the form pt + i, 1 … i … p, we obtain the seasonal factor as r-1 a Sjp + i j=0 Si = (7.6) r For the Tahoe Salt example, a total of 12 periods and a periodicity of p ϭ 4 imply that there are r ϭ 3 seasonal cycles in the data. We obtain seasonal factors using Equation 7.6 as S1 = 1S1 + S5 + S92/3 = 10.42 + 0.47 + 0.522/3 = 0.47 S2 = 1S2 + S6 + S102/3 = 10.67 + 0.83 + 0.552/3 = 0.68 S3 = 1S3 + S7 + S112/3 = 11.15 + 1.04 + 1.322/3 = 1.17 S4 = 1S4 + S8 + S122/3 = 11.66 + 1.68 + 1.662/3 = 1.67 At this stage, we have estimated the level, trend, and all seasonal factors. We can now obtain the forecast for the next four quarters using Equation 7.1. In the example, the forecast for the next four periods using the static forecasting method is given by F13 = 1L + 13T2S13 = 118,439 + 13 * 52420.47 = 11,868 F14 = 1L + 14T2S14 = 118,439 + 14 * 52420.68 = 17,527 F15 = 1L + 15T2S15 = 118,439 + 15 * 52421.17 = 30,770 F16 = 1L + 16T2S16 = 118,439 + 16 * 52421.67 = 44,794 Tahoe Salt and its retailers now have a more accurate forecast of demand. Without the sharing of sell-through information between the retailers and the manufacturer, this supply chain would have a less accurate forecast and a variety of production and inventory inefficiencies would result.