Dynamic Modeling of Brand Loyalty in Retail: A Semi-Supervised Approach Incorporating Temporal Effects and Purchase Behavior Sequences

In this paper, we introduce a comprehensive analytical framework meticulously designed to evaluate the extent of customer loyalty towards retail brands.In the current context, brand loyalty has emerged as one of the few enduring sources of competitive advantage for businesses. While numerous studies have focused on brand loyalty, most of them primarily discuss heuristic methods to measure it. In contrast, our approach emphasizes the use of machine learning techniques to predict brand loyalty, highlighting the need for a more comprehensive and systematic methodology.This study presents a data-driven, semi-supervised approach that incorporates temporal effects and purchase behavior sequences to evaluate brand loyalty of fast-moving consumer goods. We utilize Walmart's consumer panel data for fast-moving consumer goods as the training dataset and extracts behavioral features. We devise a novel method for calculating features reflecting recency, frequency, monetary, and persistence aspects of consumers' purchase behavior. Additionally, we develop a formula to create a brand loyalty label for model training. Subsequently, by applying machine learning methods, we assess whether a customer belongs to any of "Brand Loyalist", "Brand Switcher", or "Non-Loyalist" segments. Finally, the prediction model is refined and validated through experimentation with multiple product categories and time frames.The research paper concludes by providing clear directions for marketers regarding which segment of the population they should focus their finite advertising budget on to optimize their brand's success in both the immediate and extended future.