Abstract
Customer loyalty programs (LPs) are widely adopted in retail to incentivise repeat purchases and foster long-term customer relationships. In South Africa, LP participation is high, yet many programs struggle to achieve sustained retention due to reliance on generic transactional incentives. This study investigates how data mining techniques can enhance customer retention in retail loyalty programs by addressing key theoretical and practical gaps.
Adopting a quantitative research design and leveraging secondary data analysis, the study applies a multi-method framework comprising predictive modelling, clustering, and association rule mining to the publicly available Starbucks Loyalty Program dataset (17,000 customers, 300,000+ events). Despite the international origin of the dataset, the behavioural patterns observed such as offer responsiveness and recency, mirror South African consumer dynamics, offering transferable insights for local retail contexts.
Predictive models using Logistic Regression and Support Vector Machines achieved strong performance, with AUC scores exceeding 0.96 and classification accuracy above 89%, identifying recency and frequency as dominant churn predictors. Clustering revealed distinct customer segments, including disengaged high-churn groups and low-activity but stable members. Association rule mining uncovered behavioural patterns, such as incomplete offer engagement being strongly linked to churn.
The study is theoretically grounded in Relationship Marketing Theory, Brand Loyalty Theory, and Customer Lifecycle Management, bridging CRM theory with practice by operationalising constructs like trust, commitment, and behavioural loyalty. It addresses theoretical gaps in behavioural loyalty modelling, practical gaps in CRM personalisation, and methodological gaps in feature engineering for churn prediction.
Actionable recommendations include implementing recency-based reactivation triggers and segment-specific retention strategies. Overall, the research contributes a replicable, interpretable framework for enhancing loyalty program effectiveness in emerging markets like South Africa.