We are happy to announce that Olivier Schaer will give a guest talk on „The predictive power of online buzz“ on Wednesday 6th June.


Recently, in predictive analytics, there has been substantial interest in augmenting aggregate sales forecasts with individual consumer data collected from internet platforms, such as search traffic or social network shares. The aim of this work is to investigate the usefulness and limitations of this information for forecasting purposes in different applications. We develop appropriate predictive models to take advantage of this type of inputs to augment conventional forecasting methods. We investigate two different predictive tasks, one marketing and one operation related. We find that search traffic improves pre-launch life-cycle predictions compared to analogy based methods, providing new insights of how such information can be useful to marketers. On the other hand, we show that it is less helpful for short-term operational decision-making. Interestingly, this is in contrast to a large body of recent research, which reports positive findings. We identify a series of weakness in many of these studies, including lack of adequate benchmarks or rigorous evaluations.