In this contribution we describe a new approach to detecting seasonal and fashion trends, by statistically modeling how clients’ reaction to style units change with time. In our framework, client reactions are required to take the form of binary outcome variables (e.g., buy vs. do not buy, click vs. do not click). Client behavior can then be studied with generalized linear models and mixed-effect models that include temporal features. We discuss how the coefficients of such models inform which styles are going in or out of season or fashion and demonstrate these methods using simulated data.