Abstract
Purpose: The aim of this study is to assess the effectiveness of loyalty programme promotions on customer acquisition by developing a machine learning model using an ensemble approach. This research expands upon Vroom’s Expectancy Theory (VET) to incorporate social and normative factors that influence consumer decision-making, providing a forward-looking framework for unpacking how promotions drive customer acquisition.
Design/Methodology/Approach: This study adopts a quantitative, dual-phase design, distinguishing between exploratory data analysis and predictive modelling. The research integrates VET with the Theory of Planned behaviour (TPB) and the Diffusion of Innovation Theory (DIT) to establish a robust conceptual framework. Ensemble classifiers, specifically Gradient Boosting and Random Forest, were deployed to evaluate 3,010 records from the 2018 flight activities of a Canadian airline.
Findings: The results illustrate that predictive analytics can accurately forecast consumer behaviour by identifying key drivers such as marital status (60% acquisition rate among married couples), tenure, education level and income level. While the Random Forest model indicates that all proposed attributes of the Expanded Expectancy Theory significantly influence acquisition, the Gradient Boosting model offers higher precision for identifying specific high-risks segments. The study concludes that model selection is use-case dependent: Gradient Boosting is recommended when acquisition costs are high or budgets are low, while Random Forest is preferred when the primary objective is maximising the identification of all potential acquired customers (recall).
Originality/Value: This research addresses a critical gap in the literature by conceptualising an Expanded Expectancy Theory, through integrating complementary theories that addresses the significant influence of socio-economic variables and attitudinal loyalty measures into a predictive machine learning architecture. It provides practitioners with a validated template for navigating loyalty programmes and optimising promotional spend through theoretically grounded ensemble methods.