I Know What You Want Before You Know It: A Model for Predicting Time-Dependent Customer Behavior

Abstract

Today, customers are intensively exposed to (digital) information, such as advertisements. However, many advertisements appear irrelevant for customers. But what would happen if advertisments could match the precise customer interests better, even over time? Using psychological reactance theory, we show that there is a high need for targeting customers along their current and future interests as precisely as possible to prevent consumer reactance such as advertising avoidance. A specific branch of research dealing with this topic is behavioral targeting and more specifically, predictive behavioral targeting (PBT) models. PBT models allow a more accurate prediction of interests to personalize advertising and generate clicks that push sales. However, many models assume temporally stable consumer interests without considering that interests vary over time. Hence, we develop a dynamic PBT model based on decision trees and neural networks that allows a temporal differentiation of of interests into short-, medium-, and long-term. Thereby, we introduce a transformation function based on a time series of interest rates that permits the model to switch between multiple behavioral targeting (BT) and PBT models depending on the type of interest. By underlining the effects of our model with the psychological reactance theory, we offer insights for research and practice.