Predicting Donor Churn and Customer Sentiment from Reviews Using Logistic Regression and NLP: A Data-Driven Approach to Retention and Sentiment Analysis

This study employs a dual-analytical approach to explore donor churn prediction and customer sentiment analysis using logistic regression and natural language processing (NLP). Drawing on a dataset of 2,000 donors from a non-profit organization (2012–2017), we use logistic regression to identify key determinants of donor attrition, including direct marketing, TV and Facebook advertising, publicity, and demographic variables. Our best-performing churn model achieved an AUC of 0.8629, highlighting the value of personalized direct marketing and demographic segmentation in donor retention strategies. In parallel, we analyze 2,500 Amazon magazine subscription reviews using sentiment analysis and Latent Dirichlet Allocation (LDA) topic modeling. Despite accounting for negativity bias, most reviews reflected positive sentiment. Six key themes emerged from topic modeling, including lifestyle, technology, and delivery concerns, offering actionable insights for consumer engagement and product improvement. By integrating quantitative and textual data, this research provides a data-driven framework for improving donor retention and understanding customer sentiment. These findings offer strategic guidance for marketing, fundraising, and review-based customer analytics in both nonprofit and commercial sectors.