Perceived Personalization of Luxury Hotel Brands: An Empirical Examination Using Structural Equation Modeling and Machine Learning

This study explores the outcomes of perceived personalization in the context of luxury hotels through the theoretical lenses of Stimulus-Organism-Response (SOR) Theory, Personalization-Privacy Paradox Theory, and Privacy Calculus Theory. The study incorporates multiple latent constructs, such as brand engagement, brand experience, self-brand connection, consumer-brand identification, privacy concerns, behavioral loyalty intentions, and willingness to disclose information, to understand how perceived personalization influences customer perceptions and behaviors. Understanding both data (e.g., willingness to disclose information) and marketing (e.g., behavioral loyalty intentions) constructs is crucial in the era of data-driven marketing, as it helps address the complexities of personalization.A sample of 476 U.S. luxury hotel consumers was collected using a Qualtrics panel. Structural Equation Modeling (SEM) (i.e., explanation) and Machine Learning (ML) using SHapley Additive exPlanations (SHAP) values (i.e., prediction) were integrated for comprehensive analysis. The measurement model revealed the need to establish two second-order constructs, “brand engagement experience” (combining brand engagement and brand experience) and “brand identification connection” (combining self-brand connection and consumer-brand identification), to better represent the underlying relationships. The structural model revealed that perceived personalization positively influences brand engagement experience and brand identification connection. Brand engagement experience partially mediated the relationship between perceived personalization and behavioral loyalty intentions, while brand identification connection fully mediated the relationship between perceived personalization and willingness to disclose personal information. Privacy concerns did not significantly moderate the relationship between personalization and the brand-related constructs. Additionally, brand engagement experience did not significantly lead to willingness to disclose information, and brand identification connection did not significantly lead to behavioral loyalty intentions. The performance of six ML models was compared for each outcome variable. Deep neural network was the best method for predicting willingness to disclose information, and k-nearest neighbors was the best for predicting behavioral loyalty intentions. A comparative analysis between the SEM direct effects and the SHAP mean scores for the top-performing ML models demonstrated convergence for willingness to disclose information but divergence for behavioral loyalty intentions. These findings offer theoretical, practical, and methodological insights within the context of luxury hotel personalization.