Previous literature uses first‐order meta‐analyses to overcome the divergent theories about customer relational benefit, which produced mixed results in terms of the associations found. Originally conceptualised as unidimensional, customer relational benefit is an integrative multidimensional concept that includes trust, commitment, loyalty, satisfaction, word of mouth, and other constructs involved in long‐term relational exchanges. The authors conduct a second‐order meta‐analysis on 41 first‐order meta‐analyses on customer relational benefits, generating 3.8 million observations and 3462 independent samples, resulting in an overall generalisation of the relationships between these constructs. We test three theoretical models and propose an alternative rival framework. Main results indicate that the product‐ versus consumer‐related model generates strong variance in loyalty. Second, the cognitive–affective–conative–behavioural model is not superior in terms of explaining and predicting loyalty, but our three‐stage model performs better. Third, the service‐profit chain model is an alternative in terms of generating a sequence of effects (serial perspective). Fourth, our new alternative framework (a three‐stage model based on the temporal sequence of loyalty) works better for explaining some constructs.