Pricing Subscription Services with Text Data using Hedonic Pricing and Machine Learning

Abstract

Subscription e-commerce market has grown by more than 100 percent a year over the past five years and the importance of study on subscription services is self-evident nowadays. However, prices in subscription services usually have unclear structures even though offering a reasonable price is one of the most important elements of customer relationship management in service area. Thus, the aim of the study is to consider attributes based on customer preferences obtained by user-generated data and predict price for subscription services more objectively in a way that customers accept it reasonable. Target data are 10,000 web scraped reviews from representative subscription services of video streaming service brands. Using machine learning techniques, topic modeling, vector space model, dimensionality shrinkage and deriving the value of attributes were done, and hedonic pricing model was defined. In this process, the result shows that a highly predictive value can be obtained by considering the covariance between each value through Partial Least Squares regression that enables supervised Latent Dirichlet Allocation when pricing services with text-based data. Moreover, regarding the result of Netflix, Amazon Prime Video, and HBO Max being overpriced, Disney+, and Hulu being underpriced, this paper presented insights and implications for each service through considering the relationships between price and attributes that customers value and other situations that the services face.