Customers who take part in satisfaction surveys, to a vast extent, typically report that they were highly satisfied with a product or service. This phenomenon has gained a lot of attention in the field of product reviews as well and is known by the j-shape distribution of star ratings. The distributional characteristics of such satisfaction metrics are challenging when the task is to model the relationship with other potentially influencing variables. As marketers are keen in understanding why customers give a certain rating, choosing the correct model is of vital importance to avoid biased and hence unreliable estimates leading to wrong managerial implications. In her presentation for the AG Marketing at TU Clausthal Nadine Schröder proposes a count model based on a Tweedie distribution and shows how this can deal with the distributional characteristics of the star rating.

 The project is joint work with Giampiero Marra (University College London, United Kingdom), Rosalba Radice (Bayes Business School – City, University of London, United Kingdom), and Thomas Reutterer.