The study presents the first empirical analysis of customer churn prediction within the scholarly publishing industry. Utilizing a 6.5-year dataset provided by a major academic publisher, the researchers developed a predictive model to infer the likelihood of customer defection based on resampled usage patterns, specifically the volume and frequency of content downloads. This approach emphasizes the utility of minimal data inputs, focusing on temporal patterns rather than granular behavioral attributes.
Benefits & outcomes
The predictive models, particularly ensemble methods, achieved substantial accuracy in forecasting customer churn nearly a year in advance. Notably, the models maintained high performance even when specific behavioral attributes were omitted, suggesting that simplified models based on resampled usage data can effectively predict churn in subscription-based markets.
Notes & quotes
“We show that this approach can be both accurate as well as uniquely useful in the business-to-business context.”
“Modelling churn on the basis of resampling customers’ use of resources over subscription time is a better (simplified) approach than when considering the high granularity that can often characterise consumption behaviour.”
Publisher / company: Undisclosed Major Academic Publisher
Date of case study: 18/11/2022
Source: https://arxiv.org/pdf/2211.09970