In service industries, productivity growth requires matching the level of demand of the consumer and the level of service of the provider. This matching requires the service provider to have knowledge of consumer-related factors, such as the satisfaction level or the concept of value of the consumer. An intelligent model of the consumer is needed in order to estimate such factors because these factors cannot be observed directly by the service provider. However, obtaining knowledge of such factors in real services using conventional consumer behavior theory is difficult because the models are not designed for practical application, but rather are intended to provide a comprehensive and elaborative understanding of consumer behaviors. In addition, most conventional models are qualitative, and so cannot provide quantitative information for decision making by the providers. The present paper describes a method for computational modeling of the consumer by understanding the behavior based on large-scale datasets observed in real services. It is difficult to model the behavior or decision making process of the customer, which involves nonlinear or non-Gaussian variables, using conventional statistical modeling techniques, which assume linear or Gaussian models. We use a Bayesian network method, which can handle nonlinear and non-Gaussian variables as conditional probabilities. The models are constructed based on large-scale datasets observed in real services and present some practical applications of the models to retail and content providing services. The proposed method is efficient for many other services that use a variety of large-scale datasets.