Reliability and Accuracy of Bootstrap and Monte Carlo Methods for Demand Distribution Modeling

Estimation of demand is one of the most important tasks in new product development. How customers come to appreciate and decide to purchase a new product impacts demand and hence profit of the product. Unfortunately, when designers select a new product concept early in the product development process, the future demand of the new product is not known. Conjoint analysis is a statistical method that has been used to estimate a demand of a new product concept from customer survey data. Although conjoint analysis has been increasingly incorporated in design engineering as a method to estimate a demand of a new product design, it has not been fully employed to model demand uncertainty. This paper demonstrates and compares two approaches that use conjoint analysis data to model demand uncertainty: bootstrap of respondent choice data and Monte Carlo simulation of utility estimation errors. Reliability of demand distribution and accuracy of demand estimation are compared for the two approaches in an illustrative example.