Diffusion Forecasts Using Social Interactions Data



We propose an approach for incorporating data on social interactions into the estimation of diffusion models. We nest major extant diffusion models with models that jointly capture the process of adoption conditional on the number of social interactions and the process that leads to social interactions. Data on social interactions (e.g., number of recommendations received and/or given by consumers) may be used to calibrate the models, in addition to the data used to calibrate comparable extant models. Incorporating such additional data gives rise to tighter diffusion forecasts and allows calibrating diffusion models at an earlier stage of the diffusion process, when forecasts are more critical. We illustrate the proposed approach in the context of a field study conducted in collaboration with a viral marketing company.
Contact information:
Dr. S. Puntoni