Sales forecasting using combination of diffusion model and forecast market – an adaption of prediction/preference markets

Forecasting sales accurately for a new product is difficult and complex due to non-availability ofpast data. However, such forecast information is crucial for successful introduction of new products which, inturn, determines the survival of companies, in many cases. Decisions relating to new products depend criticallyon reliable period-by-period sales forecasts (otherwise called forecast time series) as early as possible in the newproduct development cycle. This information is crucial in assessing cash flow and NPV relating to the newproduct. There have been many attempts to use growth curves (otherwise called diffusion models), such as theBass model. These models made use of past data about analogous products to do this task. However, thismethod, although considered the best method, available, has many problems, such as identifying analogousproducts which can reliably mimic the new product in its sales characteristics. These difficulties explain why theaccuracy of forecasts reported by such methods is, at best, 50%. Here we propose an innovative conceptualframework to obtain time series data required for forming the growth curve for the new product bybootstrapping the growth curve models with a novel ‘Forecast market’ mechanism. The effectiveness of the‘Forecast market’ in obtaining accurate estimates of the time series data itself is likely to be enhanced by lettingthe ‘Forecast market’ participants use product information ranging from simple pictures of the product to highendvirtual reality systems which enable them to visualise and appreciate the features of the new product