The first speaker for the summer semester 2022 at our Research Seminar Series was Stephan Seiler from Imperial College London. Stephan presented his and his co-authors’ work on how to best design personalized pricing strategies, answering questions of Is purchase history or demographic information more important for personalizing price promotions? And What method is best for making predictions in that area?

His results show that customers' shopping habits are more informative than demographics for making profitable targeting pricing promotions. This can be valuable for practitioners, especially in the digital channels, where purchase history can be tracked. Comparing different prediction models with a blanket couponing strategy, Stephan also found that the performance of machine learning models can range from -21% loss to 17% gain, and that a Bayesian model performs best and achieves a 17.5% gain. From what we heard, the Bayesian model is also the least time-consuming option and does not require a lot of adjustments, which makes it easier to be widely applied.

Thank you for coming, Stephan!