In this paper, we focus on finding defuzzification methods which are appropriate for a consumer-agent based diffusion model to forecast product diffusion in a competitive automobile market. The consumer agent model defines product choice process by imitating word-of-mouth effect between consumers in a real market. We assume that information on products delivered by word-of-mouth and heterogeneous characteristics of consumers are described in linguistic terms, which are modeled with fuzzy numbers. Therefore, defuzzification method should be used for transforming the fuzzy numbers into crisp values for final product choice. Thus, effective defuzzification methods are necessary for the simulation of the agent model. We apply seven defuzzification methods to transform fuzzy operation for the product choice process of consumer agent into crisp value, and examine the appropriateness of each method by comparing its model data with real data.