Using Data-Driven Marketing to Improve Customer Retention for U.S. Businesses

This study investigates how data-driven marketing might help businesses in the United States retain more customers. Through the utilization of predictive models, we aimed to discover successful solutions. An Ordinal Neural Network (ONN) and an ensemble in which Random Forest (RF) and LightGBM were combined were the two primary models that we evaluated and contrasted. A score of 0.9063 on the Cohen's Kappa test indicated that the ONN was superior to the ensemble model in terms of its alignment with actual consumer behavior. This contrasts with the score of 0.5780 that the ensemble model received. Even though the ensemble model had a Mean Absolute Error (MAE) of 0.3236, which was slightly lower than the ONN's value of 0.3258, the ONN's higher Kappa score demonstrated that it was superior in terms of capturing the ranking connections seen in the data. In addition, our findings brought to light a few problems, such as protecting the confidentiality of data and comprehending the intricacies of model predictions, which can make implementing these methods difficult. This study suggests using data-driven personalization to retain customers and target high-risk consumers. We recommend adding real-time data and scaling models to handle larger datasets to improve projections. US firms must pay special attention to these facts to maximize marketing efforts and build consumer loyalty.