PDF File: 

ASC6_wp_shapiro.pdf

SLIM Technologies, LLC

Enterprise resource optimization systems help managers explore their decision options for strategic, tactical, and operational planning.

Scholars studying organizational behavior have long recognized the need to balance exploration of new possibilities and exploitation of old certainties in managing and evolving the firm. "Exploration includes things captured by terms such as search, variation, risk taking, experimentation, play, flexibility, discovery, innovation," said J. G. March in A Primer on Decision Making: How Decisions Happen."Exploitation includes things captured by terms such as refinement, choice, production, efficiency, selection, implementation, execution."Excessive exploration will be detrimental to the firm if managers fail to achieve closure on numerous and expensive innovations. Excessive exploitation will be detrimental to the firm if managers fail to devote time and energy learning about and adapting to changing business conditions.

Realizing benefits from information technology involves a similar balancing of opportunities for exploring and exploiting data. Widespread implementation during the past decade of enterprise resource planning (ERP) systems and other transactional data systems offers the promise of homogeneous, transactional databases that can promote efficient and effective management. However, competitive advantage is not gained simply through the acquisition and use of systems providing faster and cheaper communication of data. And, as many managers have come to realize, ready access to transactional data does not automatically lead to better decision making.

Data-driven decision making requires the application of descriptive and prescriptive (optimization) models to transactional data that allow managers to forecast and explore new possibilities. The extent of such exploration depends heavily on the scope of the decision problem, specifically:

• Strategic planning entails robust exploration of plans for expanding, contracting, and configuring the firm's resources over the coming years;
• Tactical planning entails bounded exploration of plans for refining and allocating the firm's resources over the coming months; and
• Operational planning entails limited exploration of plans for using the firm's resources in executing the firm's mission over the coming hours and days.
• Our goal is to compare and contrast these three areas of exploration that support and improve supply chain decision-making.

Exploring Transactional Data

A high level view of the transformation of transactional data into descriptive and prescriptive knowledge that supports managerial decision making is shown in Figure 1. First, descriptive models, such as those for demand forecasting, data mining, or activity-based costing, are applied to ERP and other transactional data in creating a decision database. This database contains descriptions of the firm's future needed to analyze a class of planning problems (e.g., supply chain network optimization, corporate financial strategy, or project management).

Once the decision database has been assembled, it provides inputs to optimization models that allow managers to explore the space of planning options. Managerial intuition may be expanded and refined by running multiple data sets describing these planning problems. Stated another way, the purpose of descriptive and optimization models is to enhance human knowledge of decision problems, not replace it.

ERP and other transactional data systems are exploitative IT intended to efficiently and comprehensively amass raw data describing the firm's current state. They also assist managers and workers in executing short-term plans. Descriptive modeling systems are exploratory IT that aggregate and project these raw data in assisting analysts to identify accurate and meaningful descriptions of the firm's future in the decision database.

These data descriptions are called scenarios. They are passed to the optimization modeling system, which is also exploratory IT, that translates them into mathematical representations of the space of future decision options. An optimal plan for each scenario is computed using algorithms appropriate to the type of optimization model being solved. Depending on the scope of the application, analysts explore the space of possible plans by modifying the scenarios to reflect uncertainties as well as noneconomic policy constraints on the firm's future.

These comments are applicable to all types of decision making, not merely those concerned with supply chain management. In the discussion that follows, we focus on supply chain management, an area of decision making where managers are at the forefront in promoting and applying data-driven models. This interest stems from their recognition that analysis using holistic models of their firms' supply chains is the only means for unraveling the complex interactions and ripple effects that make supply chain planning important and difficult. Moreover, supply chain resources are largely physical resources, such as plant, machines, distribution centers, or trucks, whose costs and other characteristics are relatively easy to measure.

Robust Exploration

Robust exploration of strategic supply chain planning options involves both judgmental and data-driven analysis. An effective starting point is scenario planning, which is a formalism for assisting senior managers in defining scenarios of their firm's future that are consistent, plausible and comprehensive.¹ The judgments of senior managers and industry experts are combined in constructing these scenarios, which initially may be qualitative but quickly require data about projected demand for existing and new products in existing and new markets, and manufacturing costs and capacities including those for new technologies, transportation costs, and a host of other factors. Typical strategic planning supply chain studies include:

• Post-merger consolidation of two food companies with operational and cultural differences;
• Distribution network expansion to support business expansion in North America by a large retailing company;
• Worldwide sourcing of industrial chemical products to changing markets; and
• Capital investment planning to expand production capacity in a petrochemical company.

Because these and other studies are concerned with long-term decisions involving the firm's entire supply chain network comprised of vendors, facilities, and customers, prescriptive models for analyzing strategy are called supply chain network optimization models.

A well-constructed supply chain network optimization model allows senior management to simultaneously explore: a portfolio of investment, divestment, and reconfiguration resource options including new and existing facilities, technologies, equipment, products; and, for each portfolio, an aggregate resource allocation plan describing how these resources may best be employed in managing the firm's supply chain.

To accomplish this breadth and depth of exploration, the model must incorporate zero-one decision variables for each portfolio option where a value of one corresponds to accepting the option, and a value of zero corresponds to rejecting the option. The mathematical form of the model is called mixed integer programming because it involves these zero-one decision variables as well as resource allocation decision variables and constraints. Such models are optimized by an algorithm that implicitly considers all possible portfolios in seeking one that optimally balances the resource portfolio selection and allocation decisions. For any trial portfolio being considered by the algorithm, the selected options provide the supply chain with resources that are optimally allocated to meet the supply chain's demand and other requirements.

For any scenario specifying the firm's future, the model is used to determine the portfolio of resource options and an associated resource allocation plan that is optimal, as measured by an objective such as maximization of discounted net revenues over the strategic planning horizon, or maximization of return on investment. Multiple scenarios are constructed and optimized to explore the risks associated with uncertainties regarding future product demand, currency exchange rates, the cost of raw materials, and other important parameters.² Scenarios may also reflect political realities, i.e., regardless of economic analysis, the manufacturing plant located in the city where the firm has its corporate headquarters cannot be shut down.

Despite an expanding number and variety of strategic supply chain modeling projects, exploration of strategic supply chain plans is less robust in many studies than it could be. Too often, strategic decision making for purchasing, manufacturing, and distribution planning is weakly integrated. Many companies would realize great benefits by truly integrated supply chain analysis and planning.

Moreover, strategic planning should address integration of supply chain management with demand management.³ The application of a supply chain network optimization model to identify strategic plans for meeting fixed and given product demand provides the marginal cost, and therefore the marginal profit, for delivering each product in each time period to each market. At a minimum, this knowledge should be incorporated in strategic marketing plans. A more powerful approach would be to construct and apply models that explore coordinated supply chain and marketing decision options that lead to the maximization of net revenue, which equals gross revenue from sales minus total supply chain and marketing costs. Finally, models that maximize net revenue by integrating supply chain and marketing decisions should be extended to explore corporate financial decisions. On the one hand, net revenue maximization figures over the coming years largely determine the firm's earnings before interest and taxes, which are key financial performance measures. Conversely, strategic expansion of the firm's supply chain often involves large capital outlays that the firm's financial planners amass from loans, retained earnings, and the issuance of corporate bonds. The portfolio of investment options for improving supply chain performance should be holistically analyzed.

Bounded Exploration

Tactical supply chain planning differs from strategic supply chain planning in two important ways. First, the major investment and divestment options open to strategic planning are considered fixed and given over the shorter tactical planning horizon. A supply chain network optimization model used for strategic planning is easily adapted for tactical planning by fixing investment and divestment options to their values determined by the strategic plan. The focus is to use the adapted model to holistically explore plans for managing the company's supply chain over the coming months. Typical tactical plans to be explored include:

• Inventory buildups of products with seasonal demand;
• Assignment of realized and projected product demand to manufacturing facilities;
• Timing and sourcing of acquisitions from vendors; and
• Shift scheduling in manufacturing plants and distribution centers.

Second, unlike strategic planning studies that can be performed independently of ongoing management of the company's supply chain, tactical planning exercises must be carried out as a key element of such management. Thus, exploration of the firm's tactical plans requires the creation of new business processes that involve a team of planners and analysts responsible for:

• Creating, validating, and maintaining the tactical supply chain decision database;
• Performing monthly scenario analyses (or some other regular cycle) using the supply chain network optimization model;
• Disseminating trial tactical plans to the firm's supply chain managers; and
• Negotiating acceptable final tactical plans among the firm's supply chain managers.

The tactical version of the supply chain network optimization model performs bounded, not robust, exploration of the firm's tactical plans. Its application differs in two important ways from the version used for strategic planning. First, scenarios examined during the monthly tactical planning cycle are more limited and focused than those examined during a strategic planning study. Second, the decision space of tactical plans examined by the optimization model is more restricted to reflect shorter-term realities of the firm's supply chain operations. Still, exploratory options, such as major shifts in sourcing, production, or distribution plans, can and should be examined during the tactical planning process.

The need to harmonize exploration and exploitation in creating and employing tactical supply chain planning processes is particularly difficult and important. As shown in Figure 2, individuals who create modeling systems have an exploratory cultural orientation. By contrast, IT administrators who manage the integration of modeling systems and ensure their reliable use by supply chain managers and analysts have an exploitative cultural orientation. Their responsibility is to integrate modeling systems used for supply chain planning with data management and reporting systems and to ensure that use of these systems is reliable and efficient.

The potential clash of cultures and the need to harmonize them occurs not only in the initial development phase of a modeling system. It also occurs periodically when exploration by the modeling system needs to be reviewed and revised to fit new realities of the firm's supply chain tactical planning environment. Because the form of a model is often viewed as a black box by IT administrators, the need to fix something that appears to be unbroken can lead them to resist necessary and desirable updates.

Limited Exploration

Unlike strategic and tactical planning, exploration of operational supply chain plans is myopic, not holistic. Separate modeling systems and processes are used to route vehicles, schedule production, and maintain inventories. Myopic analysis and planning is indicated because operational plans cannot realistically be evaluated holistically due to the large quantities of data that must be analyzed and the tight timing required by efficient decision making.

An important area of operational supply chain planning in the retailing industry is the management of inventories in each of the company's stores, which is the logic underlying a state- of-the-art modeling system for conducting retailing inventory management (see Figure 3). Each store has a point-of-sale (POS) system that generates daily information about sales of SKUs often numbering in the tens of thousands. A challenging exercise in descriptive modeling is to mine these data by simultaneously aggregating them into clusters with similar demand characteristics and estimating the parameters for the demand distribution of each cluster.

The result is a generalized ABC classification of items into those with fast- (type A), medium- (type B), and slow-moving (type C) distributions. The derived distributions are used to forecast demand for the items. In addition, these forecasts coupled with inventory theory models are used to order items in optimal quantities from the firm's suppliers at the optimal re-order levels. The information regarding forecasts and re-orders are fed back into the POS system so that salespeople are aware of planned changes in inventories over the coming days.

The descriptive and prescriptive modeling systems depicted in Figure 3 are merely one example of the types of sophisticated systems required for operational supply chain management. A balancing of exploration and exploitation as depicted in Figure 2 is clearly needed for the successful development and use of such systems. Exploration is critically needed to identify the knowledge inherent in large data sets. Exploitation is critically needed to ensure efficient and reliable identification of operational plans and their communication to workers responsible for executing the firm's business.

The optimization models underlying the selection, timing, and sizing of inventory re-orders of individual items have been thoroughly studied over the past 50 years. However, re-ordering may not strictly follow the results of these models due to the company's desire to send efficient, full-load shipments of many items to individual stores. Thus, an unresolved tension exists between exploitation of old inventory models and exploration of new, more flexible plans for restocking the stores. In general, similar tensions exist in the design and use of most operational modeling systems. More applied research is needed to understand how best to harmonize these tensions.

Conclusions

The previous decade has seen extensive developments in ERP, POS, and other exploitative IT for collecting, organizing, and communicating transactional data regarding the firm's supply chain. We have argued that the time is ripe for wider application of exploratory IT, which is comprised of descriptive and prescriptive models, that translates these transactional data to knowledge enabling supply chain managers to make better decisions. The extent of exploration varies according to the scope of the planning problems: robust for strategic supply chain management; bounded for tactical supply chain management; and, limited for operational supply chain management. Success to date with exploratory modeling systems indicates that substantial improvements in supply chain performance can be realized by creating, validating, and employing these systems. Firms that are innovators in applying these tools will achieve significant and sustainable competitive advantage.

Endnotes

¹ See pp. 116-121 in J. E. Russo and P. J. H. Schoemaker, Winning Decisions, Currency Books, Doubleday, New York, NY.

² Stochastic programming is an advanced modeling approach that allows multiple scenarios with associated probabilities of occurrence to be simultaneously optimized to identify optimal contingency plans for each scenario and an immediate plan for hedging against these contingencies. Recent developments suggest that stochastic programming will soon be a practical tool for strategic supply chain management.

³ Integration of strategic supply chain decision making with demand and corporate financial decision making is discussed in J. F. Shapiro, "Beyond Supply Chain Optimization to Enterprise Optimization," ASCET Volume 4, Montgomery Research Inc., San Francisco, 2001.