Analyzing Customer Churn Prediction in the E-commerce Industry Using Advanced Statistical Models

Thesis Abstract

Customer retention remains a critical challenge in the highly competitive e-commerce industry, where understanding and predicting customer churn can substantially influence business strategies and profitability. This study aims to develop and evaluate advanced statistical models to accurately predict customer churn within an e-commerce platform, thereby enabling proactive retention measures. The specific objectives include identifying key customer behavior variables associated with churn, comparing the effectiveness of various predictive techniques, and proposing a robust, interpretable model for operational deployment. A quantitative research design underpins this investigation, employing a retrospective analysis of customer transaction and engagement data. The population comprises active customers of an e-commerce company with a total customer base of approximately 200,000 individuals over a three-year period. A stratified random sampling technique is used to select a sample of 5,000 customers, ensuring proportional representation across customer segments such as demographic groups, purchase frequency, and geographic regions. Data collection involves extracting transactional records, website interaction logs, customer service interactions, and feedback surveys stored within the company's enterprise resource planning (ERP) and customer relationship management (CRM) systems. To ensure the reliability and validity of data, preprocessing procedures include handling missing values through multiple imputation, normalizing variables to facilitate comparison, and conducting exploratory data analysis to detect outliers and multicollinearity. The study employs multiple analytical techniques, with logistic regression as a baseline, complemented by advanced machine learning models such as Random Forest, Gradient Boosting Machines (GBM), and Support Vector Machines (SVM). Model performance evaluation utilizes metrics including accuracy, precision, recall, F1-score, and the Area Under the Receiver Operating Characteristic curve (AUC-ROC). Feature importance analysis identifies the most influential variables in churn prediction, aiding interpretability and practical application. The theoretical framework is grounded in the Theory of Planned Behavior and the Customer Loyalty Model, which suggest that customer intentions and satisfaction influence loyalty and churn. The study proposes a conceptual model illustrating the relationship between customer engagement metrics, satisfaction indicators, and churn probability. Data analysis involves multicollinearity diagnostics, feature selection via recursive feature elimination, and hyperparameter tuning through grid search. Ethical considerations include ensuring customer data confidentiality and securing necessary permissions in accordance with data protection regulations. It is anticipated that the results will demonstrate that machine learning models, particularly GBM, outperform traditional logistic regression in predictive accuracy, with key predictors including purchase frequency, customer service satisfaction scores, and website engagement levels. The findings are expected to contribute to the existing literature by providing a comparative analysis of statistical models within the context of e-commerce churn prediction and identifying critical customer behavior indicators. Furthermore, the study aims to develop an interpretable and scalable model that the company can integrate into its customer relationship management system for real-time churn alerts. The main conclusion underscores the significant role of advanced statistical and machine learning models in enhancing customer retention strategies. The study recommends the adoption of the superior predictive model for targeted retention interventions and emphasizes the importance of continuous data monitoring and model updating as customer behaviors evolve. Additionally, further research should explore the integration of sentiment analysis from customer feedback statements and social media interactions to enrich predictive accuracy. Overall, this research addresses a vital area for e-commerce sustainability, offering practical insights and methodological advancements in customer churn analytics.