Exploring the Mediating Role of Affective and Cognitive Satisfaction on the Effect of Service Quality on Loyalty

This research aims to test the mediating role of both affective and cognitive satisfaction on the effect of service quality on loyalty. The affective satisfaction is represented by Kansei Engineering-based measures and the cognitive satisfaction is represented by overall customer satisfaction. The study is based on survey through personal interviewing and face-to-face questionnaire. There were 102 respondents from 24 hotels ranging from three-star to five-star hotels in Surabaya, Indonesia. There are four latent variables, namely, service quality, overall customer satisfaction, Kansei, and loyalty. The relationships among the constructs are modeled via structural equation modeling. The mediation analysis is carried out using bias-corrected bootstrap sampling. We found that both overall customer satisfaction and Kansei partially mediate the relationship between service quality and loyalty (approximately 52% mediation effects). In particular, the two mediators, namely, Kansei and overall customer satisfaction accounts for 24% and 28% of the effect of service quality on loyalty, respectively. Additionally, we failed to find a statistically significant relationship between Kansei and overall customer satisfaction. This might be due to relatively low sample size. The generalization of the results from the study is limited because of the relatively small sample size. Furthermore, the study was carried out in a single service setting, that is, hotel services. Affective satisfaction is equally important as cognitive satisfaction in service encounter. Both are shown to have significant mediating effects on the relationship between service quality and loyalty. To create higher loyalty, service provider should therefore not overlook the importance of both affective and cognitive satisfaction. This research complements the previous research by taking into account both cognitive and affective satisfaction as mediators at the same time. It is shown that the two-mediator model fits the data better than that of using one mediator or no mediator