Unravelling The Dynamics Of Online Ratings

Online product ratings are an immensely important source of information for consumers and accordingly a strong driver of commerce. Nonetheless, interpreting a particular rating in context can be very challenging. Ratings show significant variation over time, so understanding the reasons behind that variation is important for consumers, platform designers, and product creators. In this paper we contribute a set of tools and results that help shed light on the complexity of ratings dynamics. We consider multiple item types across multiple ratings platforms, and use a interpretable model to decompose ratings in a manner that facilitates comprehensibility. We show that the various kinds of dynamics observed in online ratings are largely understandable as a product of the nature of the ratings platform, the characteristics of the user population, known trends in ratings behavior, and the influence of recommendation systems. Taken together, these results provide a framework for both quantifying and interpreting the factors that drive the dynamics of online ratings.