Title: Personalized Recommendation System Design for an Online B2B Platform
Authors: Vishal Gaur, Cornell SC Johnson College of Business, and Xiaoyan Liu, Leavey School of Business, Santa Clara University
We formulate the problem of designing a personalized recommendation system for an online business-to-business (B2B) marketplace, develop a method to solve it, and evaluate results using a field experiment. Our research is conducted in collaboration with IndiaMart, the dominant online B2B platform in India serving approximately 60 million buyer firms and 5 million seller firms in more than 50 million products and services. The problem faced by the platform is to accurately match heterogeneous requests for quotation (RFQs) real-time with suitable sellers with the highest likelihood of acceptance by addressing two major challenges: (1) high-dimensional and sparse data regarding product category and spatial engagement, and (2) class imbalance such that the volume of `accepted' records in historical data is significantly larger than that of `declined' records. We propose new variables motivated by the choice estimation literature to address high-dimensionality, and evaluate alternative approaches including Synthetic Minority Over-sampling Technique (SMOTE) and a new resampling approach, which we call Panel Data Augmentation Technique (PDATE), to counter class imbalance. A controlled field experiment conducted at IndiaMart shows that our method provides a consistent and significant improvement in the quality of recommendations sustained over time. Since the writing of this paper, our method is being scaled by the company to its entire platform.