This year’s largest marketing analytics conference organised by the INFORMS Society for Marketing Science was held in Los Angeles at the University of Southern California. Oliver chaired a session on forecasting new products with diffusion models and presented a talk entitled “Pre-launch diffusion model parameter estimation using online search traffic data”. Obtaining valuable pre-launch forecasts is one of the biggest challenges for marketers and demand planners. Current methods are often based heavily on expert judgment or information from similar products. The approach, in development by Oliver and his supervisors Dr Nikos Kourentzes and Professor Robert Fildes, uses search traffic data from Google Trends to augment the estimation of market potential.

Preliminary results, from forecasting physical video game sales, shows encouraging results where models with search traffic augmented market potential estimators outperform purely analogy based models. The findings are consistent across parametric models such as the Bass model but also for a non-parametric model which tends to perform better.

Furthermore, search traffic which is considered as a proxy for advertising expenditures and consumer interest typically found to provide useful information up to two months prior to launch.

The entire presentation can be found below and anyone from the industry wishing to collaborate in the research area of new product forecasting is very welcome to contact Oliver directly or Georgie Watson.

Downloads:

• O. Schaer, Pre-launch diffusion model parameter estimation using online search traffic data‌