When to Message: Investigating User Response Prediction with Machine Learning for Advertisement Emails

Abstract

Direct marketing message campaigns are a common way for businesses to deliver updates, recommendations, or coupons to their user base to spark interest in brands and products. In this paper, we explore the possibility of using machine learning to predict the response behavior of users to regular newsletter emails from a real-world e-commerce business. In doing so, we train and evaluate classification models, such as random forests and artificial neural networks, to predict the probability of a user interacting with an email based on past behavior. Further investigation is conducted into the potential of using the sending time of a message to influence responses, based on the assumption that a user’s likeliness to be engaged depends on the time of day a message is received. We identify two user groups that have a preference regarding morning and evening messages and can show that this preference holds for a subsequent message campaign. Thus, our results demonstrate a clear potential for time-aware response modeling approaches for marketing campaigns.