Service personalization has become a cornerstone of relationship marketing strategy, yet its aggregate effect on customer loyalty and the mechanisms through which it operates remain empirically underspecified. This study synthesizes empirical evidence across human and AI service contexts through a dual-method meta-analytic framework, addressing three gaps: the absence of path-specific aggregate effect size estimates, the lack of formal aggregate-level mediation testing, and the uninvestigated boundary condition of service delivery mode. A systematic meta-analysis (k = 15, N = 5,398) combined with Meta-Structural Equation Modeling (Meta-SEM) was conducted. Random-effects models using the DerSimonian-Laird estimator were applied to three path-specific pooled effect sizes. Moderator analyses examined service delivery mode (human-driven vs. AI-driven personalization) and cultural context through between-group Q-tests. Publication bias was assessed using Egger’s regression (b₀=−0.215, p>0.05) and the Duval and Tweedie (2000) trim-and-fill procedure (L₀=0). Personalization significantly predicts satisfaction (r = 0.391, 95% CI [0.360, 0.421]) and directly predicts loyalty (r = 0.344, 95% CI [0.301, 0.386]). Meta-SEM confirms satisfaction as a significant partial mediator (indirect β = 0.157; 45.6% of total effect). Human-driven personalization yields significantly stronger satisfaction effects than AI-driven personalization (Q_between = 4.822, p=0.028), establishing service delivery mode as a meaningful boundary condition. The individualistic-culture sample remains limited (k = 2), restricting formal cross-cultural moderation testing. The majority of included studies employed single-source cross-sectional surveys, raising common method variance concerns. Service managers should not assume AI-driven personalization fully substitutes for human service encounters in generating loyalty. The 13% effect size gap between human and AI personalization (r = 0.431 vs. r = 0.375) signals meaningful loyalty returns from investing in human-mediated personalization, particularly in high-stakes service contexts. This study is among the first to integrate Meta-SEM with moderation analysis to disentangle human-driven from AI-driven personalization effects on customer loyalty, providing a theoretically grounded synthesis spanning hospitality, banking, e-commerce, and AI service contexts.

Personalization and customer loyalty: a dual-method examination of relational and cognitive pathways

期刊论文
发表日期：

2026年6月3日

查看全文

Jangam Suneel Deexith

摘要

Service personalization has become a cornerstone of relationship marketing strategy, yet its aggregate effect on customer loyalty and the mechanisms through which it operates remain empirically underspecified. This study synthesizes empirical evidence across human and AI service contexts through a dual-method meta-analytic framework, addressing three gaps: the absence of path-specific aggregate effect size estimates, the lack of formal aggregate-level mediation testing, and the uninvestigated boundary condition of service delivery mode. A systematic meta-analysis (k = 15, N = 5,398) combined with Meta-Structural Equation Modeling (Meta-SEM) was conducted. Random-effects models using the DerSimonian-Laird estimator were applied to three path-specific pooled effect sizes. Moderator analyses examined service delivery mode (human-driven vs. AI-driven personalization) and cultural context through between-group Q-tests. Publication bias was assessed using Egger’s regression (b₀=−0.215, p>0.05) and the Duval and Tweedie (2000) trim-and-fill procedure (L₀=0). Personalization significantly predicts satisfaction (r = 0.391, 95% CI [0.360, 0.421]) and directly predicts loyalty (r = 0.344, 95% CI [0.301, 0.386]). Meta-SEM confirms satisfaction as a significant partial mediator (indirect β = 0.157; 45.6% of total effect). Human-driven personalization yields significantly stronger satisfaction effects than AI-driven personalization (Q_between = 4.822, p=0.028), establishing service delivery mode as a meaningful boundary condition. The individualistic-culture sample remains limited (k = 2), restricting formal cross-cultural moderation testing. The majority of included studies employed single-source cross-sectional surveys, raising common method variance concerns. Service managers should not assume AI-driven personalization fully substitutes for human service encounters in generating loyalty. The 13% effect size gap between human and AI personalization (r = 0.431 vs. r = 0.375) signals meaningful loyalty returns from investing in human-mediated personalization, particularly in high-stakes service contexts. This study is among the first to integrate Meta-SEM with moderation analysis to disentangle human-driven from AI-driven personalization effects on customer loyalty, providing a theoretically grounded synthesis spanning hospitality, banking, e-commerce, and AI service contexts.