Managing Uncertainty
Tuesday, April 15

9:10am-10:00am
Continuous Improvement for Demand Planning
Jim T. Williams, MS, Manager of Operations Research, Land O’Lakes, Inc.

Companies that implement an ERP system call it “a pretty good day” when the system is up and running at all, much less to optimal capabilities. In particular, the system’s demand planning logic is a key driver of inventory investment and the success (failure) of Sales and Operations Planning processes. Demand planners may be left to configure the system on their own, starting with “one-size-fits-all” algorithm settings. With a daunting array of options, how can you proceed along a continuous improvement path? This talk will focus on the experience at Land O’Lakes:
• Establishing useful performance metrics,
• Partitioning the forecast hierarchy into clusters that prefer similar algorithm settings,
• Reducing manual steps in the forecast cycle,
• Enabling higher-payoff planning initiatives,
• Finding the balance between quantitative rigor and ease of implementation.

10:30am-11:20am
Managing Bias in Demand Forecasting
Jay W. Hopman, MBA, Researcher & Strategic Analyst, Intel Corporation (IT Innovation & Research)

Internal studies of forecasting performance have revealed the presence and impacts of bias in the signals used to generate forecasts. Furthermore, all players in the forecasting process – from customers to sales personnel to managers to the forecasters themselves – impart bias as forecasts are developed. Since 2003 Intel has been developing and testing new methods to measure and manage bias:
• A standardized intelligence collection process is providing objective data to guide forecasts of new product ramps.
• New forecasting tools based on prediction markets are providing more accurate and stable forecasts of aggregate sales volumes.
• A survey mechanism is providing a better understanding of key business variables by measuring beliefs and bias across job functions and areas of expertise.
Hopman will describe how Intel is piloting these methods and present sample results.

11:30am-12:20pm
Simulation Methods to Improve Alignment of Supply Chain with Consumer Demand for Novel Products
Michael A. Kubica, MS, MBA, President, Applied Quantitative Sciences, Inc.

Supply chain management for commercialized products is challenging in its own right, but the planning process for novel products is particularly so. Supply chain planning and management for novel products begins long before launch and must account for uncertainties both in demand and production capacity. Single-point estimation of such uncertainties can lead to significant over- or under-budgeting, leading to costly inventories or stock-outs. This is especially true when raw materials are common among multiple novel products being launched by diverse operating units. Recent advances in business modeling technologies make it possible to efficiently link probabilistic forecasts from across the enterprise to facilitate planning at explicit confidence levels for the supply chain. In this talk, Kubica will describe how libraries of probabilistic forecasts were developed for unique business units within a major medical device company, and how these forecasts were used as stochastic library components to establish budgeting and supply chain planning decisions.

Decision-makers, from individuals up to heads of national governments, face multiple risks all the time. Focusing too much on one tends to increase vulnerability to others. Real-life examples: A well-known insurance company trying to limit losses wound up driving away many of its best customers and nearly went bankrupt. A major high-tech company focused on avoiding investment losses gave away an immensely profitable line of business. Another major company, trying to innovate and hold market share above all else, backed a hugely unprofitable project far too long. Several medicines commonly prescribed for diabetes tend to cause heart attacks, some heart medications tend to cause depression, and some antidepressants tend to cause diabetes. If we have a major flu epidemic, our hospital waiting rooms are likely to be the major spread vector. The U. S. Government's post-9/11 antiterrorism initiatives greatly weakened its capability to respond to natural disasters. Using OR risk analysis, Samuelson will explore how to develop policies and practices that more appropriately balance competing risks, at all levels from individual health to global security.

4:10pm-5:00pm
Designing Distribution Networks Using Simulation and Optimization
William T. Cartwright, BA, Senior Manager, Logistics Strategy, Planning & Analytics, JCPenney

Contemplating radical process changes in 1999, the complexity of the redesign required that JCPenney develop a perspective of its distribution network that would be difficult without sophisticated network models. “Go-forward” strategies were established with the help of system enablers and a consulting firm who administrated network models on the retailer’s behalf. Years later and with the need for additional throughput capacities, consultants were engaged and the network reexamined. The result of the second project determined the location for further investment in cross-dock and consolidation capital. With a deepening grasp of modeling methods, a growing sense of the discipline’s potential uses as well as the evolution and accessibility of modeling applications, JCPenney subsequently set out to select software and develop internal competencies in the modeling and simulation disciplines. In this session, Cartwright will trace his experiences, transitioning to where network models are now developed and administrated internally, discuss sourcing decisions that lead to the selection of an application, and highlight the evolution of its increasing applications within JCPenney’s supply chain network.