Scenario analysis of demand in a technology market using leading indicators

Abstract

This paper proposes an approach to analyzing demand scenarios in technology-driven markets where product demands are volatile, but follow a few identifiable lifecycle patterns. After examining a large amount of semiconductor data, we found that not only can products be clustered by lifecycle patterns, but in each cluster there exists a leading indicator product that provides advanced indication of changes in demand trends. Motivated by this finding, we propose a scenario analysis structure in the context of stochastic programming. Specifically, the demand model that results from this approach provides a mechanism for building a scenario tree for semiconductor demand. Using the Bass growth model and a Bayesian update structure, the approach streamlines scenario analysis by focusing on parametric changes of the demand growth model over time. The Bayesian structure allows expert judgment to be incorporated into scenario generation while the Bass growth model allows an efficient representation of time varying demands. Further, by adjusting a likelihood threshold, the method generates scenario trees of different sizes and accuracy. This structure provides a practical scenario analysis method for manufacturing demand in a technology market. We demonstrate the applicability of this method using real semiconductor data.