Data-Driven vs. Theory-Driven Customer Retention Projections: A Comparison Based on Rolling-Window Cross-Validation

Customer retention is one of the critical metrics for businesses to evaluate and project the further growth and profitability. There are two main methods used to evaluate customer retention rate: theory-based and data-driven. The aim of the research is to estimate which of the methods help to evaluate customer retention projection more accurately. The theory-based method is more common and widely used among researchers, while data-driven methods might be underestimated. This study directly compares the performance of these two approaches using rolling-window cross-validation on. The study is motivated by the following question: Which of the methods: a data-driven approach or a theory-driven approach is the most relevant to measuring customer retention approach for companies nowadays? In this study the secondary data-analyses are used: 10 datasets of the customer churn rate. For each dataset, both data-driven and theory-driven models are developed and their projected results are compared with the actual customer retention rates to estimate the accuracy. The findings of this study indicate that the performance of data-driven projections, based on machine learning algorithms and statistical models, outperformed theory-driven projections derived from established marketing theories or expert opinions. Specifically, the data-driven models consistently achieved higher accuracy, precision, and predictive power compared to the theory-driven models. In addition data-driven methods are simpler in usage and can be implied without special programs, but rather simple packages (e. g. MS Excel). By leveraging predictive analytics and machine learning algorithms, organizations can gain a competitive advantage in retaining customers and optimizing their business strategies.