Predicting Annual Customer Spending in E-commerce: A Comparative Analysis of Mobile App and Website Interaction Metrics

In the competitive world of e-commerce, companies must continuously optimize their digital platforms toincrease customer engagement and drive revenue. This project focuses on predicting annual customer incomefor an e-commerce company based in London, which operates both online clothing sales and in-store personalstylist consultations. The company aims to determine whether to focus on improving its mobile app orwebsite to maximize customer spending. By analyzing user behavior—such as time spent on the mobile app,time spent on the website, average in-store session length, and membership duration—this study identifiesthe factors that most significantly impact customer spending.To achieve this, three machine learning approaches were employed: Recurrent Neural Networks (RNN),Regression Analysis, and Decision Tree Regression (DT Regression). RNNs were used to capturesequential patterns in user behavior, while Regression Analysis provided a baseline for understanding linearrelationships between the features and the target variable. However, Decision Tree Regressionoutperformed both models by effectively capturing non-linear relationships and offering high interpretability,making it the most accurate method for predicting annual income.The results show that time spent on the mobile app is the strongest predictor of customer spending, followedby membership length. Time spent on the website, while still significant, had a weaker influence on annualincome. Additionally, in-store session length had a relatively smaller impact, highlighting the greaterimportance of digital interactions in influencing customer purchases. These insights lead to a clearrecommendation: the company should prioritize improving its mobile app experience to drive highercustomer engagement and revenue.Overall, this study provides valuable, data-driven guidance for the company's strategic decision-making,helping them allocate resources effectively between their mobile app, website, and in-store services.