A Forecasting and Inventory Model for Short Lifecycle Products with Seasonal Demand Patterns

The lifecycles of many products are becoming shorter and shorter as innovation and timeto-market become key aspects of modern corporations' success in the marketplace. Few methods for inventory management exist that are capable of contending with product lifecycles measured in months rather than years. This situation is especially acute in the semiconductor industry, where a single generation of a product typically spends less than 6 months in the marketplace demand for these products is therefore highly nonstationary and stochastic. Supply chain management is further complicated by nonstationary, stochastic lead times and the necessity for complex demand forecasting procedures the use of time-series forecasting models is therefore generally infeasible for these products. We present a new method in which exogenous short and medium-term forecasts are combined with a new method of exponential smoothing to generate replenishment forecasts for finished goods inventories. This method is designed to contend with highly non-stationary, stochastic demand and lead time patterns, daily seasonal effects, unusual probability distributions, capacity constraints, and short product lifecycles across a variety of products. Historical simulations of the model for many of Intel Corporation's Central Processing Unit products reveal dramatic inventory reductions compared to the ad hoc inventory management policies presently in use. Thesis Supervisor: Donald Rosenfield Senior Lecturer, Sloan School of Management Director, Leaders for Manufacturing Fellows Program Thesis Supervisor: James Masters Executive Director, Masters in Logistics Program