A choice-based approach to the diffusion of a service: Forecasting fax penetration by market segments

Motivated by poor performance of standard estimation methods in the present application, the problem of modeling the diffusion of a new service (or a product) is considered without the assumption of a homogeneous population. The model consists of a two-stage procedure where customers receive purchase occasions according to a conventional diffusion model and at each purchase occasion they buy according to a binary choice model. This approach permits explicit incorporation of individual customer demographics and product attributes, whence one can study changes in diffusion as a function of changes in product attributes, prices, advertising, and customer demographics. The model has provided excellent results in a number of applications including the one on fax penetration reported here. In the present applications this approach has been useful in market segmentation, studying the effect of marketing strategies, and in evaluating the effects of new service features. A limited validation is carried out to judge the forecasting performance of the model used in the fax example. A simulation study is carried out to compare the proposed solution with the result obtained by fitting the Bass model to a set of market segments separately. The simulation indicates that the proposed approach substantially improves over the aggregate fitting.