Oliver Schaer, University of Virginia

Oliver Schaer will present her current work that focuses on improving new product forecasting. The first part of the talk includes a presentation of a working paper that looks into estimating the market potential with Google Trends. In contrast to existing research, which relies on pre-release buzz information only for the launch phase, Olivia will explain the ability of pre-release buzz to predict life-cycle sales. Olivia uses analogies by augmenting information from previous generations with pre-release online search traffic. First, Oliva will propose a model of pre-release online search traffic and market potential, establishing the connection between the two. Then, she’ll validate this relationship with an empirical experiment on sequential video game sales. This will demonstrate that one can obtain improved forecasting accuracy using a model that only requires one previous product generation.

The second part of the talk then looks into ways to improve the estimation of growth parameters of diffusion models across multiple generations. Often it can be challenging to estimate such models once the number of generations increases. Olivia proposes a new method to estimate diffusion curves for multiple generations using independent bootstrapping for coefficient estimation. This not only facilitates the estimation when having large numbers of generations but also gives insights about the change in adoption parameters over time. Olivia will empirically evaluate the method on a set of new technology and product adoption cases. She’ll also demonstrate that the proposed method can improve forecasting predictions while providing a robust way to measure changes in parameters across generations and improve estimation through variable selection.

Both papers make use of our open-source R package ‘diffusion’ for forecasting with diffusion models.

You must register for this event to receive the Zoom link.