Forecasting modellen voor nieuwe producten

Abstract

When forecasting the demand of a new product, historical data are unavailable for managers to predict future sales or adoption of the product. A broad range of models has been developed with the purpose of tackling this problem. This dissertation discusses a sub-branch of quantitative forecasting techniques, namely diffusion models. The basic premise of the diffusion models is that adoption of products follows a sigmoidal trend. In the beginning, few people adopt the product. This initial period is followed by a spike in the adoption rate. Eventually the market is saturated, and growth slows down. The three most significant models that express the sigmoidal growth curve are the Bass, logistic and Gompertz model. Despite the popularity of diffusion models in the research on adoption of new products, they also have received a great deal of criticism. The lack of distinctiveness between the models, the inseparable characteristic of the need for data and the lack of practical business cases are major concerns impacting the general applicability of diffusion models. For the purpose of testing the ability of the three basic models to describe the diffusion pattern of an innovation, the functions are fitted for mobile subscription data of five different European countries. If the actual data showed little trend fluctuations, the models provided a good fit. However, the data also demonstrated various fluctuations the diffusion models couldn’t account for.