Feature importance in predicting customer offer acceptance in online retail

Abstract

Accurately predicting whether customers will accept offers is essential for improving marketing strategies and increasing engagement in the highly competitive online retail sector. This study applies predictive modeling and key feature analysis to a dataset of 2,198 customers, each characterized by 20 features, including demographics, spending patterns, and engagement metrics. Techniques such as Random Forest, Support Vector Machine, and Logistic Regression were employed to predict offer acceptance, with performance evaluated through five metrics including confusion matrices, accuracy, precision, recall, and F1-score. These approaches achieved high accuracies of 99.32%, 98.86%, and 98.41%, respectively, indicating strong predictive reliability. Among the various features, customer income emerged as the most significant predictor of offer acceptance across all models, emphasizing the importance of financial metrics in understanding customer behavior and preferences. These findings underscore the value of financial insights, providing data-driven guidance for marketing strategies to enhance engagement, optimize conversion rates, and boost sales in online retail.