Chapter 8. Strategic and Tactical Supply Chain Planning

Part III:

Applications

Part III includes four chapters in which we explore supply chain modeling applications and supporting concepts:

Chapter 8. Strategic and Tactical Supply Chain Planning

Chapter 9. Operational Supply Chain Planning

Chapter 10. Inventory Planning

Chapter 11. Supply Chain Decision Making Under Uncertainty

Optimization models for strategic and tactical planning are similar in their holistic analysis of a company’s supply chain. Both types of models are intended to optimize integrated decision making across purchasing, manufacturing, distribution, and transportation activities. Thus, it is appropriate and convenient to discuss them together in Chapter 8. The principal difference is that optimization models for strategic planning include redesign options, such as the location of new distribution centers (DCs) or the acquisition of a manufacturing firm, whereas such options are treated as fixed and given when considering the company’s tactical plans.

Optimization models for operational supply chain planning are usually much more myopic in their scope and more detailed in their descriptions of the decisions to be analyzed. In Chapter 9, we consider a variety of operational modeling applications. We also address the design of modeling systems to support operational planning and their integration with other systems used by the company to acquire and communicate data.

Inventory planning, which is important at all levels of decision making, is discussed as a stand-alone topic in Chapter 10. The topic merits special attention because inventory models and solution methods are much different than optimization models and methods. Artistry is required to integrate them. Chapter 10 also contains descriptions of successful modeling applications in which inventory decisions were given priority.

In Chapter 11, we examine modeling approaches that explicitly treat uncertainties faced by the decision maker. The goal of these stochastic models is to explicitly identify contingency plans and hedging strategies for dealing with uncertainties. Although the models have been successfully applied to a few supply chain planning problems, they are too sophisticated for today’s typical planning situations. Still, we believe stochastic modeling approaches are becoming more realistic as well as more important and therefore well worth our attention.

Chapter Eleven:

Supply Chain Decision Making

Under Uncertainty

Uncertainties underlying supply chain decision making are a reality that managers may accommodate in a number of ways depending on the scope of the decisions to be made. The extent and possible impact of uncertainties are large for strategic planning, lower for tactical planning, and still lower for operational planning. Uncertainties about strategic planning may be addressed by analyzing scenarios describing a variety of likely and not-so-likely long-term futures. Uncertainties about tactical planning may be addressed by models and processes that regularly update details of scenarios of the medium-term future and revise decisions to be implemented. Uncertainties about operational planning may be addressed by short-term contingency plans that use stand-by resources.

For strategic and tactical supply chain planning, data-driven network optimization models such as those discussed in Chapter 8 provide important insights about the impact of uncertainties by optimizing decisions associated with scenarios of the future. For a given scenario, modeling analysis is said to be deterministic when it treats the scenario data as certain to occur. Deterministic optimization of multiple scenarios is the current state of the art for evaluating supply chain uncertainties. The decision maker uses results from several optimization runs in selecting her implementation plans.

Stochastic programming models extend deterministic optimization models. They provide deeper insights because they optimize decisions over multiple scenarios linked together in a single model, each with an associated probability of occurrence. The optimization criterion for a stochastic programming model may be minimization of expected total costs or maximization of expected net revenues across all the scenarios; that is, costs and revenues associated with each scenario are weighted by their probability of occurrence and then summed. By simultaneously determining optimal contingency plans for each scenario and short-term strategies that optimally hedge against these contingencies, a stochastic programming model can provide more effective decisions than a deterministic model. A stochastic programming model also allows the imposition of constraints on decisions that limit risks faced by the company—for example, bounds on net revenue losses allowed under any scenario.

We begin our discussion with a review of the discipline of scenario planning in Section 11.1. Scenario planning is an important methodology independent of its role in supporting modeling analysis because it helps senior managers define scenarios that are consistent, plausible, and comprehensive. In Section 11.2, we examine contingency planning and discuss an application of data-driven models to contingency planning for the natural gas network that supplies electric power generators in the northeastern United States. In Section 11.3, we examine decision trees that are formalisms for describing and analyzing states of uncertainty that evolve over time. One approach to stochastic programming model construction is to combine a decision tree describing the evolution of major uncertainties with linear and mixed-integer programming submodels describing resource acquisition and allocation decisions occurring at each state or node in the decision tree. This construction is illustrated in Section 11.4 along with a more traditional construction based on combining multiple versions of a deterministic model.

Two applications of stochastic programming are reviewed in Section 11.5. One application involves a data-driven model to support quick-response decisions based on early sales of retail products. The other application involves a two-stage planning approach for the manufacture of hybrid corn seed products. In Section 11.6, we explore modeling approaches to risk management of supply chains including the mitigation of risk using real assets as well as financial instruments. Section 11.7 contains our final thoughts about the prospects for stochastic programming applications to supply chain management.