Predicting the Incremental Benefits of Online Information Search for Heterogeneous Consumers

Consumer information search (CIS), the process by which a consumer browses and inspects a shopping environment for appropriate information to select a product or service from available options, has been a research focus in the context of online business. One of the key questions related to CIS is how to determine how much information to search (i.e., when to stop searching). Extensive literature on behavioral science has revealed that consumers often search either ‘too little’ or ‘too much,’ even with the help of existing consumer decision support systems (CDSSs). To address this issue, this article introduces a new method of CDSSs that provides effective estimation of incremental search benefits. The method, called the personalized distribution‐based prediction method (PDM), is developed from the perspective of machine learning and utilizes consumer preference information generated by collaborative filtering (CF) algorithms. In contrast to related methods that assume that all consumers follow the same distribution function in terms of product preference, the PDM method is designed to consider the diversified search behaviors of consumers through the incorporation of heterogeneous preference distribution functions. Experiments based on data provided by Netflix illustrate that the proposed method is effective and advantageous over existing applicable techniques. Theoretical analyses are also provided to explain the advantageous performance of PDM.