Assessment of user price sensitivity

Abstract

Systems and methods for assessment of user price sensitivity using a predictive model are disclosed. An example method may be performed by one or more processors of a retention system and include retrieving traversal sequences including lists of pages users accessed prior to price notifications, labeling the traversal sequences based on whether users terminated sessions upon the notifications, transforming the traversal sequences into graphs based on identifiers assigned to the pages and instances in which users successively accessed pages, defining predictive features suggesting an extent to which identified structural attributes of given graphs affect user price sensitivity, generating model training data based on the labeled traversal sequences and predictive features, determining optimal weights for the predictive features, and generating a model incorporating the optimal weights and trained to predict a likelihood that a user will terminate a session when notified of a price to continue given the user's traversal sequence.