Abstract
The swift advancement of digital technologies has reshaped how retailers engage with customers in South Africa, positioning omni-channel strategies at the heart of loyalty programme effectiveness. This study investigates the effect of omni-channel engagement on loyalty outcomes and examines the potential of machine learning (ML) in predicting the most effective engagement channels. Grounded in the Theory of Planned Behaviour (TPB), Social Exchange Theory (SET), and Service-Dominant Logic (SDL), the research integrates behavioural theory with predictive analytics to model how multi-channel interactions influence offer redemption.
Using the Starbucks Offer Completion dataset (Kaggle 2025), the study analysed over 139,000 transactions through descriptive, correlational, and inferential techniques, including logistic regression and machine learning classification. The findings reveal that mobile and web engagement are the strongest predictors of offer completion, far exceeding the influence of demographic factors such as age, income, or gender. Offer design characteristics particularly duration and difficulty also significantly influenced redemption behaviour, whereas reward value alone did not.
The research addresses a critical empirical gap in South African retail, where limited studies have quantified the relative effectiveness of individual marketing channels or applied ML to predict loyalty outcomes. The novel contribution lies in combining behavioural theory with ML modelling to operationalise psychological constructs in a predictive framework. Practically, the study recommends that South African retailers adopt mobile-first, data-driven loyalty strategies to enhance customer engagement while ensuring omni-channel coherence. Although the study was limited by the use of secondary international data, the study provides a transferable methodological foundation for loyalty optimisation in emerging markets.