| Our major objective is to study the relationship between some variables and new product growth patterns by efficiently utilizing several sources of information, which include: (1) sales histories of previously introduced product categories; (2) basic product characteristics such as the type of the product (consumer goods, industrial goods, or both) and whether there is the need for developing complementary products; (3) market conditions, which include the introduction time, whether there are alternative products, and status of the market pioneer (established or startup), and the time span between invention of the technology and market introduction. We propose a hierarchical Bayesian mixture model as the basic framework, which is built upon the hierarchical Bayesian diffusion model of Lenk and Rao (1990) and product segmentation of Bayus (1993). This model allows us to segment previously introduced product categories based on their diffusion patterns in a way that, before its market introduction, it is possible to assign segment membership probabilities to a new product category with only some basic information on the product and its market. A general prediction of market growth pattern of the new product can be made, drawing from diffusion patterns of the respective product clusters. In the model, cluster membership probabilities are functions of product characteristics and the market conditions under which they are introduced. The proposed model also significantly improves upon the meta-analysis in the diffusion literature, without many limitations of the latter, such as the widely different standards in data requirement and estimation approaches. |
