A video-based automated recommender (VAR) system for garments

Change log
Xiao, L 
Ding, M 

In this paper, we propose an automated and scalable garment recommender system using real-time in-store videos that can improve the experiences of garment shoppers and increase product sales. The video-based automated recommender (VAR) system is based on observations that garment shoppers tend to try on garments and evaluate themselves in front of store mirrors. Combining state-of-the-art computer vision techniques with marketing models of consumer preferences, the system automatically identifies shoppers’ preferences based on their reactions and uses that information to make meaningful personalized recommendations. First, the system uses a camera to capture a shopper’s behavior in front of the mirror to make inferences about her preferences based on her facial expressions and the part of the garment she is examining at each time point. Second, the system identifies shoppers with preferences similar to the focal customer from a database of shoppers whose preferences, purchasing, and/or consideration decisions are known. Finally, recommendations are made to the focal customer based on the preferences, purchasing, and/or consideration decisions of these like-minded shoppers. Each of the three steps can be implemented with several variations, and a retailing chain can choose the specific configuration that best serves its purpose. In this paper, we present an empirical test that compares one specific type of VAR system implementation against two alternative, nonautomated personal recommender systems: self-explicated conjoint (SEC) and self-evaluation after try-on (SET). The results show that VAR consistently outperforms SEC and SET. A second empirical study demonstrates the feasibility of VAR in real-time applications. Participants in the second study enjoyed the VAR experience, and almost all of them tried on the recommended garments. VAR should prove to be a valuable tool for both garment retailers and shoppers.

Data, as supplemental material, are available at http://dx.doi.org/10.1287/mksc.2016.0984.

retailing, video analysis, collaborative filtering
Journal Title
Marketing Science
Conference Name
Journal ISSN
Volume Title
Institute for Operations Research and the Management Sciences (INFORMS)
The authors thank the participants in presentations given by the authors in College of Business at City Univeristy of HongKong and Cambridge Judge Business School for their feedback, as well as the Editor, the Area Editor, and two anonymous Marketing Science reviewers for their insightful comments. This research was supported by two National Natural Science Foundation of China Fund (Grants 71232008 & 71502039), and the Institute for Sustainable Innovation and Growth (iSIG) at School of Management, Fudan University.