Product design selection is heavily constrained by its customer preference data acquisition process. Traditionally, the customer preference data is collected through survey-based methods such as conjoint; sometimes product prototypes are generated and evaluated by focused groups of customers. In this way, the data acquisition process can become costly and require a significant amount of time.

  The goal of this dissertation is to overcome the limitation of the traditional customer preference data acquisition process by making use of a new type of customer data - online customer reviews. Because online customer reviews are, to a large extent, freely available on the Internet copiously, using them for product design can significantly reduce the cost as well as the time. Of course, the data obtained from online reviews have some disadvantages too. For example, online reviews are freely expressed and can contain a lot of noise. 

  In this dissertation, a new methodology is developed to extract useful data from online customer reviews from a single website, construct customer preference models and select a product design that provides a maximum expected profit. However, online customer reviews from a single website may not represent the market well. Furthermore, different websites may have their own procedures and formats to acquire customer reviews. A new approach is developed to systematically elicit customer data from multiple websites, construct customer preference models by considering website heterogeneity, and select a product design. The model from multiple websites is also extended to account for customer preference heterogeneity. The models obtained from the online customer reviews for single and multiple websites are compared and validated using a set of out-of-sample data. To demonstrate the applicability of the proposed models, a smartphone case study is used throughout the dissertation.