Variable demand and cost uncertainty are the key issues that have been taken into consideration while planning its supply chain network. A new approach has been developed to determine the production capacities, distribution locations and materials flow for a new product launch. The model provides solution for all potential scenarios of demand growth and also addresses the financial viability of the company during the introduction and growth phase of the product. But there are so many other potential factors that have been ignored which might have played a significant role in planning the new supply chain network. This paper discusses the roles of those key drivers that have been overlooked in this case study and how the supply chain network could have been planned more efficiently had those factors been taken into account.Keywords: supply chain design, data management, robust supply chain model, New product launch Introduction: The fortune 200 company has developed a method to design the supply chain for the new and spin off product focusing on two main ideas. First, the company needs to have sufficient cash reserve to support new products longer until they become viable. Second, to find a solution that is flexible and robust so that it works well under many different potential growth scenarios.Supply Chain Design: The supply chain network has been divided into three phases: Data interpretation, modeling and solution.

The data interpretation phase involves identifying the alternatives for manufacturing and distribution facilities and equipment which includes existing and new facilities as well as different size, technology and cost options. It also involves the identification of important uncertain model parameters among cost, capacities, demand, labor , raw material and transportation. These parameters are then combined to form reasonable scenarios.The modelling phase consists of developing mixed integer programming model and selection of robustness criterion (robustness criteria determines how to define a good solution to all of our defined scenarios). In this case, deviation from the optimal profit was chosen as the robust criterion. The robustness criterion can be adjusted depending on the risk averseness of the company. A less risk averse approach is to use a weighted average of the deviation from the optimal profit for all scenarios. One approach to find out probabilities for the likelihood of each scenario is to assign scenarios equal probabilities and another approach involves getting expert opinions and then infer the overall probabilities.

The final phase involves fine tuning and solving the model by performing sensitivity analysis to see how the solution would change if the probabilities and robust parameter change. Based on the model output, the decision maker decides which supply chain investments to pursue during launching of the product.Robust Supply Chain Model: The company developed a scenario based multi period production/distribution supply chain design model which addresses strategic planning for production and distribution investments simultaneously with the concept of viability. The model outputs a single set of decisions for production machines and distribution centers to be used for all scenarios and also speciﬁes the production and shipment quantities for each scenario.Case Study: In response to the shift in demand away from company’s core products due to advancement in technology, the Fortune 200 company developed new products for manufacture on their existing equipment. In the planning of the supply chain network, some segments of the supply chain of the company’s current product were used. Additionally, the new product is sold as an input material to manufacturers, while the existing product is sold in the retail market. The supply chain network, the problem data and the model results of this case study have been described below.Supply Chain: The supply chain network for the new product launch consists of one U.S. manufacturing facility, seven possible distribution center locations, and four customers located in Asia – namely China, Japan, Korea, and Taiwan. The new product line consists of a “high-grade” and a “low-grade” product for the initial launch and after the initial entry into the market, the company hopes to add more features to its products to gain an advantage over competitors. To transport the product from the manufacturing facility in the U.S. to Asia, it must be shipped by truck to a nearby airport or seaport from where it can then be transported by airplane or ship to Asia.

Air shipment is signiﬁcantly costlier than sea, but the lead-time is much shorter. Because the model does not include inventory cost explicitly, the inventory charges for the transit time have been added to the shipping cost to obtain the total cost for each shipment.Problem Data: Uncertainty in supply chain parameters like demand and transportation costs have been incorporated in the model. 13 scenarios have been defined based on the different levels of demand and transportation cost parameters. The scenarios deal with two levels of demand, “low” being 10% of the total market forecast and “high” being 25% of the market. Special attention is paid to Japan and Korea in the scenarios, because they represent 80% of the market, and building these relationships is important for the success of the product launch. In the model, the demand combinations are paired with two levels of transportation costs to form the ﬁrst eight robust case scenarios, R1-R8. Two levels of transportation costs are included, the “current” costs and the “high” transportation costs corresponding to a 30% hike over current costs.

For case scenarios R9-R12, the low and high levels of demand are for the high-grade product only. In scenario R13, the demand for the product gradually falls to zero after seven periods. The model is solved for 16 six-month periods for a total of eight years where years 1-4 are growth years and years 5-8 are in the mature stage. Solving the model: Solving the model involves two main tasks to ﬁne-tune the appropriate scenario probabilities and the risk tolerance by adjusting the robustness parameter. First, determining the impact of the scenario probabilities was focused on while ignoring the robustness parameter. For this analysis, the robustness parameter was set to zero allowing the model to act as a simple expected value model. Next, the probabilities are kept constant to explore the effect of different robustness parameters on the decisions.The probability for scenario R13 (i.e product failure) is varied from 10% to 90%. The remaining probability is allocated to the other 12 scenarios (R1-R12) in the following fashion. The likelihood of scenarios R1-R8 is same as scenarios R9-R12. If the product is successful there is approximately 20% chance that the demand will shift to high-grade only (R9-R12); therefore, 20% of the probability of success (probabilities R1-R12) is allocated equally to scenarios R9-R12. This analysis produces three distinct combinations of machines and distribution centers: Solution 1 for product success below 20%, Solution 2 for success in the range of 20-70%, and Solution 3 for the success rate above 70%.When the probability of success is below 20%, Machines 1 and 2 are open in Period 1. When the success rate is between 20% and 70%, Machine 3 is added in Period 5. Machine 4 is added in Period 4 when the probability of success is above 70%. Distribution Centers 1-4 are open in Period 1 in all the cases. When the probability of success is less than 70% there is not enough production capacity to satisfy all the demand. The DC in China is closed in Period 8 when the demand of the other customers becomes large enough to gain economies of scale by focusing on fewer, but larger customers.Although varying the probability of success produced only three facility solutions, for each probability level, a different net proﬁt was produced. The minimum net cash positions for all 13 scenarios for Solutions 1 and 2 are -$0.8 million and for Solution 3 is -$9.4 million, which occurs in scenario R13 (product failure).

The net proﬁt for Solution 1 ranges from $3.2 million to $122.7 million, Solution 2 ranges from $2.8 million to $133.7 million and Solution 3 has the highest maximum net proﬁt of $144.7 million, but also the lowest net proﬁt of -$9.4 million when the product fails. As the probability of success increases, the maximum scenario net proﬁt increases, but the minimum scenario net proﬁt decreases. This indicates that when the probability of success is high, the model suggests accepting slightly more risk.The expected value model shows how the scenario probabilities, particularly the probability assigned to product failure, change the solution. The manufacturing capacity increased as the probability of success increased from 10% to 90%. Varying the probability of product success in the expected value model provides alternative solutions for consideration. The pessimistic scenario can be used to determine how the network should develop to deal with uncertainty. It establishes the minimum investment required to enter into the market. The optimistic scenarios provide an alternative solution that indicates when future investment decisions should be reconsidered.Now holding the scenario probabilities constant, the second task to validate the model involves determining the correct robustness parameter λ, i.e., the weight or value of lost proﬁt in dollars from not having accurate information. As a point of contrast, we demonstrate the effect of the robustness parameter with two cases: an optimistic case with a 90% probability of success and a pessimistic case with a 10% probability of success was demonstrated. The value of λ represents the value in dollars of being 100% away from the optimal solution’s net proﬁt. However, the value of the net proﬁt depends on scenario, so in order to have one λ for all scenarios, the weighted average of the scenario net proﬁts may be appropriate. The weighted average net proﬁt is $112.8 million for the optimistic case and $16.7 million for the pessimistic case.

Additionally, robustness parameter λ can be a measure for the risk tolerance of the company. The λ is multiplied by the percentage difference between the optimal and the robust net proﬁt. Thus, as λ increases, the lower performing scenarios are more emphasized, resulting in more conservative expansion plans.The λ values indicated in Exhibit 1 were varied from $0 to the highest net proﬁt of all scenarios, $144.7 million. For the optimistic case, increasing λ produces two robust solutions (combinations of machines and distribution centers) both of which have identical decisions for Periods 1–3. For the solution with higher values of λ (λ>$16.8 million), Machine 4 is not opened in Period 4 and DC 1 is closed in Period 8. The higher values of λ coincide with the more conservative solution. Utilizing Machine 4 in the less conservative solution increases the production capacity from 6,500 units per year to 10,400 units, which corresponds to meeting an additional 31% of the maximum expected demand. Likewise, the λ values were varied for the pessimistic case with a 10% probability of success. For this example, increasing λ results in six different robust solutions (as per Exhibit 1). The decisions differ only in facility closings. As λ increases more facilities close. Before closing machines, these solutions could meet 41% of the maximum anticipated demand.Exhibit 1. Robust Facility Decisions[image: ]The weighted net proﬁt for both the optimistic and pessimistic cases are plotted against the λ values (in $ million) in Exhibit 2. The weighted average net proﬁt is $112.8 million for the optimistic case and $16.7 million for the pessimistic case. At these points, the net proﬁt curves are relatively ﬂat, indicating little change in net proﬁt with changes in λ near the weighted average. (The weighted averages are marked on the curves in Exhibit 2.) In the optimistic case, the net proﬁt decreases by 3% as λ increases from $4.7 million to $16.8 million. The net proﬁt remains unchanged for λ>$16.8 million. In the pessimistic case, the net proﬁt continues to decrease as λ increases, with a sharper decline as λ>$109.0 million. The net proﬁt declines by 12% from λ=$0 to λ=$109.0 million and then an additional 39% for λ=$109.0 to λ=$144.7 million.Exhibit 2: Weighted Net Proﬁt versus λ Values[image: ]In conclusion, the robust model analysis shows how with the increase in the penalty parameter λ the weight on scenarios with lower net proﬁt increases and thus provides a more conservative expansion plan; however, the solutions for λ values near the weighted average of the net proﬁt do not change signiﬁcantly with λ. Further, the solutions for all λ values,for both the optimistic and pessimistic cases have the same decisions for Period 1. Therefore, in Period 3 the model could be solved again, before additional investment is made. This helps in taking future decisions with the advantage of more current information about the uncertain demand and costs. The robust model solutions for the case study demonstrate that a conservative expansion plan is warranted initially, with temporary resolving before additional investment is considered. This approach, especially in high-risk scenarios, is reassuring to the practicing engineering manager.Evaluation: Though the company has come up with an excellent supply chain design for the launch of its new product, there are gaps in the analysis. Every product is different and thus no one supply chain planning strategy is fit for all the products. In this case study there is no mention of the kind of product that the company is planning to launch. So it will be very difficult to say if the strategy taken by them will bring success in the long run. Moreover, there are challenges other than uncertainty in demand and transportation costs that needs to be addressed in order to make the supply chain design more effective.First, products have shorter life cycles due to rapidly changing market demands. As trends will not last for a long time there is always a pressure to keep up with the latest trends and innovate by introducing new products, while keeping the total manufacturing cost low. This demands a flexible supply chain that can be utilized for manufacturing other products and for future projects.Second, customers demand high-quality products. If the products are of inferior quality or are damaged, the company might lose demand. Thus, enterprises are under increasing pressure to create high-quality products and to create them consistently. This can be done by addressing quality at every level of the supply chain, such as raw materials procurement, manufacturing, packaging, logistics, and product handling.Third, product quality often goes hand-in-hand with compliance.

The data interpretation phase involves identifying the alternatives for manufacturing and distribution facilities and equipment which includes existing and new facilities as well as different size, technology and cost options. It also involves the identification of important uncertain model parameters among cost, capacities, demand, labor , raw material and transportation. These parameters are then combined to form reasonable scenarios.The modelling phase consists of developing mixed integer programming model and selection of robustness criterion (robustness criteria determines how to define a good solution to all of our defined scenarios). In this case, deviation from the optimal profit was chosen as the robust criterion. The robustness criterion can be adjusted depending on the risk averseness of the company. A less risk averse approach is to use a weighted average of the deviation from the optimal profit for all scenarios. One approach to find out probabilities for the likelihood of each scenario is to assign scenarios equal probabilities and another approach involves getting expert opinions and then infer the overall probabilities.

The final phase involves fine tuning and solving the model by performing sensitivity analysis to see how the solution would change if the probabilities and robust parameter change. Based on the model output, the decision maker decides which supply chain investments to pursue during launching of the product.Robust Supply Chain Model: The company developed a scenario based multi period production/distribution supply chain design model which addresses strategic planning for production and distribution investments simultaneously with the concept of viability. The model outputs a single set of decisions for production machines and distribution centers to be used for all scenarios and also speciﬁes the production and shipment quantities for each scenario.Case Study: In response to the shift in demand away from company’s core products due to advancement in technology, the Fortune 200 company developed new products for manufacture on their existing equipment. In the planning of the supply chain network, some segments of the supply chain of the company’s current product were used. Additionally, the new product is sold as an input material to manufacturers, while the existing product is sold in the retail market. The supply chain network, the problem data and the model results of this case study have been described below.Supply Chain: The supply chain network for the new product launch consists of one U.S. manufacturing facility, seven possible distribution center locations, and four customers located in Asia – namely China, Japan, Korea, and Taiwan. The new product line consists of a “high-grade” and a “low-grade” product for the initial launch and after the initial entry into the market, the company hopes to add more features to its products to gain an advantage over competitors. To transport the product from the manufacturing facility in the U.S. to Asia, it must be shipped by truck to a nearby airport or seaport from where it can then be transported by airplane or ship to Asia.

Air shipment is signiﬁcantly costlier than sea, but the lead-time is much shorter. Because the model does not include inventory cost explicitly, the inventory charges for the transit time have been added to the shipping cost to obtain the total cost for each shipment.Problem Data: Uncertainty in supply chain parameters like demand and transportation costs have been incorporated in the model. 13 scenarios have been defined based on the different levels of demand and transportation cost parameters. The scenarios deal with two levels of demand, “low” being 10% of the total market forecast and “high” being 25% of the market. Special attention is paid to Japan and Korea in the scenarios, because they represent 80% of the market, and building these relationships is important for the success of the product launch. In the model, the demand combinations are paired with two levels of transportation costs to form the ﬁrst eight robust case scenarios, R1-R8. Two levels of transportation costs are included, the “current” costs and the “high” transportation costs corresponding to a 30% hike over current costs.

For case scenarios R9-R12, the low and high levels of demand are for the high-grade product only. In scenario R13, the demand for the product gradually falls to zero after seven periods. The model is solved for 16 six-month periods for a total of eight years where years 1-4 are growth years and years 5-8 are in the mature stage. Solving the model: Solving the model involves two main tasks to ﬁne-tune the appropriate scenario probabilities and the risk tolerance by adjusting the robustness parameter. First, determining the impact of the scenario probabilities was focused on while ignoring the robustness parameter. For this analysis, the robustness parameter was set to zero allowing the model to act as a simple expected value model. Next, the probabilities are kept constant to explore the effect of different robustness parameters on the decisions.The probability for scenario R13 (i.e product failure) is varied from 10% to 90%. The remaining probability is allocated to the other 12 scenarios (R1-R12) in the following fashion. The likelihood of scenarios R1-R8 is same as scenarios R9-R12. If the product is successful there is approximately 20% chance that the demand will shift to high-grade only (R9-R12); therefore, 20% of the probability of success (probabilities R1-R12) is allocated equally to scenarios R9-R12. This analysis produces three distinct combinations of machines and distribution centers: Solution 1 for product success below 20%, Solution 2 for success in the range of 20-70%, and Solution 3 for the success rate above 70%.When the probability of success is below 20%, Machines 1 and 2 are open in Period 1. When the success rate is between 20% and 70%, Machine 3 is added in Period 5. Machine 4 is added in Period 4 when the probability of success is above 70%. Distribution Centers 1-4 are open in Period 1 in all the cases. When the probability of success is less than 70% there is not enough production capacity to satisfy all the demand. The DC in China is closed in Period 8 when the demand of the other customers becomes large enough to gain economies of scale by focusing on fewer, but larger customers.Although varying the probability of success produced only three facility solutions, for each probability level, a different net proﬁt was produced. The minimum net cash positions for all 13 scenarios for Solutions 1 and 2 are -$0.8 million and for Solution 3 is -$9.4 million, which occurs in scenario R13 (product failure).

The net proﬁt for Solution 1 ranges from $3.2 million to $122.7 million, Solution 2 ranges from $2.8 million to $133.7 million and Solution 3 has the highest maximum net proﬁt of $144.7 million, but also the lowest net proﬁt of -$9.4 million when the product fails. As the probability of success increases, the maximum scenario net proﬁt increases, but the minimum scenario net proﬁt decreases. This indicates that when the probability of success is high, the model suggests accepting slightly more risk.The expected value model shows how the scenario probabilities, particularly the probability assigned to product failure, change the solution. The manufacturing capacity increased as the probability of success increased from 10% to 90%. Varying the probability of product success in the expected value model provides alternative solutions for consideration. The pessimistic scenario can be used to determine how the network should develop to deal with uncertainty. It establishes the minimum investment required to enter into the market. The optimistic scenarios provide an alternative solution that indicates when future investment decisions should be reconsidered.Now holding the scenario probabilities constant, the second task to validate the model involves determining the correct robustness parameter λ, i.e., the weight or value of lost proﬁt in dollars from not having accurate information. As a point of contrast, we demonstrate the effect of the robustness parameter with two cases: an optimistic case with a 90% probability of success and a pessimistic case with a 10% probability of success was demonstrated. The value of λ represents the value in dollars of being 100% away from the optimal solution’s net proﬁt. However, the value of the net proﬁt depends on scenario, so in order to have one λ for all scenarios, the weighted average of the scenario net proﬁts may be appropriate. The weighted average net proﬁt is $112.8 million for the optimistic case and $16.7 million for the pessimistic case.

Additionally, robustness parameter λ can be a measure for the risk tolerance of the company. The λ is multiplied by the percentage difference between the optimal and the robust net proﬁt. Thus, as λ increases, the lower performing scenarios are more emphasized, resulting in more conservative expansion plans.The λ values indicated in Exhibit 1 were varied from $0 to the highest net proﬁt of all scenarios, $144.7 million. For the optimistic case, increasing λ produces two robust solutions (combinations of machines and distribution centers) both of which have identical decisions for Periods 1–3. For the solution with higher values of λ (λ>$16.8 million), Machine 4 is not opened in Period 4 and DC 1 is closed in Period 8. The higher values of λ coincide with the more conservative solution. Utilizing Machine 4 in the less conservative solution increases the production capacity from 6,500 units per year to 10,400 units, which corresponds to meeting an additional 31% of the maximum expected demand. Likewise, the λ values were varied for the pessimistic case with a 10% probability of success. For this example, increasing λ results in six different robust solutions (as per Exhibit 1). The decisions differ only in facility closings. As λ increases more facilities close. Before closing machines, these solutions could meet 41% of the maximum anticipated demand.Exhibit 1. Robust Facility Decisions[image: ]The weighted net proﬁt for both the optimistic and pessimistic cases are plotted against the λ values (in $ million) in Exhibit 2. The weighted average net proﬁt is $112.8 million for the optimistic case and $16.7 million for the pessimistic case. At these points, the net proﬁt curves are relatively ﬂat, indicating little change in net proﬁt with changes in λ near the weighted average. (The weighted averages are marked on the curves in Exhibit 2.) In the optimistic case, the net proﬁt decreases by 3% as λ increases from $4.7 million to $16.8 million. The net proﬁt remains unchanged for λ>$16.8 million. In the pessimistic case, the net proﬁt continues to decrease as λ increases, with a sharper decline as λ>$109.0 million. The net proﬁt declines by 12% from λ=$0 to λ=$109.0 million and then an additional 39% for λ=$109.0 to λ=$144.7 million.Exhibit 2: Weighted Net Proﬁt versus λ Values[image: ]In conclusion, the robust model analysis shows how with the increase in the penalty parameter λ the weight on scenarios with lower net proﬁt increases and thus provides a more conservative expansion plan; however, the solutions for λ values near the weighted average of the net proﬁt do not change signiﬁcantly with λ. Further, the solutions for all λ values,for both the optimistic and pessimistic cases have the same decisions for Period 1. Therefore, in Period 3 the model could be solved again, before additional investment is made. This helps in taking future decisions with the advantage of more current information about the uncertain demand and costs. The robust model solutions for the case study demonstrate that a conservative expansion plan is warranted initially, with temporary resolving before additional investment is considered. This approach, especially in high-risk scenarios, is reassuring to the practicing engineering manager.Evaluation: Though the company has come up with an excellent supply chain design for the launch of its new product, there are gaps in the analysis. Every product is different and thus no one supply chain planning strategy is fit for all the products. In this case study there is no mention of the kind of product that the company is planning to launch. So it will be very difficult to say if the strategy taken by them will bring success in the long run. Moreover, there are challenges other than uncertainty in demand and transportation costs that needs to be addressed in order to make the supply chain design more effective.First, products have shorter life cycles due to rapidly changing market demands. As trends will not last for a long time there is always a pressure to keep up with the latest trends and innovate by introducing new products, while keeping the total manufacturing cost low. This demands a flexible supply chain that can be utilized for manufacturing other products and for future projects.Second, customers demand high-quality products. If the products are of inferior quality or are damaged, the company might lose demand. Thus, enterprises are under increasing pressure to create high-quality products and to create them consistently. This can be done by addressing quality at every level of the supply chain, such as raw materials procurement, manufacturing, packaging, logistics, and product handling.Third, product quality often goes hand-in-hand with compliance.