| Abstract: | There has been a lot of interest in diffusion models as a basis for prelaunch estimates of the sales of new products, and indeed there have been several models developed that have achieved fairly good acceptance by new product managers. One of the limitations of such models, however, has been the requirement that a sales history for the new product, even a short one from a test market, for example, be available to derive the parameters of the model. For some types of products—consumer durables, services, industrial products, for example—a sales history isn't available. In this article, Professor Robert Thomas suggests some steps toward the development of models that incorporate the attractive features of diffusion models. His approach is to use, in a systematic way, the sales histories of products that can be considered to have analogous features from a buyer's point of view. He illustrates the approach by forecasting the sales of a new service. |
