How many manufacturing plants and distribution center facilities should we have? Where should each of these facilities be located? What should the service territory of each facility be? Which products should be made where? How should products flow throughout our supply chain? These are some of the key questions companies typically ask when engaging in strategic network design. In order to confront these questions, companies have to sort through an abundance of supply chain data in order to understand the complexities of their supply chain and answer the question above. The number of possible combinations of solutions is extremely large, and firms must also balance various trade-offs when making the decisions. Therefore, an Excel spreadsheet will not be sufficient, and a mathematical tool, network optimization software, is a better option to help answer these questions. We sometimes refer to this as a network “model” because the tool is used to create a replica of a company’s supply chain. This model solves a specific objective function (usually minimizing cost) and utilizes constraints (a company’s business rules).
Although this model is extremely powerful, one cannot just push the “run” button and get the correct answers to a firm’s supply chain strategy questions. There is a great amount of work that goes into collecting, analyzing, and aggregating data in order to feed it into the network model. Additionally, just because a model “solves” that, it does not mean it is correct or the best solution for a company. For example, assume a model is run and Indianapolis is selected as the lowest cost single distribution center from a list of 30 potential cities across the U.S. One cannot assume that they have arrived at the optimal answer for their company. If the objective is least cost, the model will output the facility with the lowest cost subject to all the constraints, even if the second best solution is only $1 more per year! What would be the second best city? Let’s say that city is Cincinnati, where the firm currently has a distribution center, and the increase in cost is $20k. It would be hard to justify the movement of a DC from the firm’s current facility in Cincinnati to Indianapolis for little to no savings in annual supply chain costs. The business disruption and move costs would not justify such a change.
A second example would be a supply chain network where multiple DCs are being considered (assume we are comparing a 3DC vs. 4DC vs. 5DC network). Sometimes the difference between adding or removing a DC does not significantly change the costs of the network (some costs may go up and others go down, but the overall supply chain costs are not significantly different). In these cases, where there is a range of possible solutions, other factors not included in the model must be considered. Many of these factors are qualitative rather than quantitative or even difficult to quantify, but still must be considered. For example, how does the solution prevent lost sales, gain additional sales, or meet customer service expectations? What are the risks, complexity, and disruption costs associated with the recommended solution? How flexible is the solution with changes from growth assumptions? How transparent is the solution to customers? Additionally, there are other costs to consider that are not typically included in the model – taxes and tax incentives, detailed site selection, and labor availability to name a few.
The model is a very powerful mathematical tool used to evaluate numerous possibilities of complex, multi-echelon supply chain problems. However, this is just a tool, and many other factors must be considered before moving forward in the implementation of an “optimal” solution for a firm that is aligned with its own strategy.
— Brad Barry, St. Onge Company