This study examines the performance of a penalized neural network and the replication of a customer engagement survey scale with text information in the hotel industry. Although the empirical analysis shows highly accurate model performance only in the training sample, the results also clarify the issues of the engagement scale.

This study reviews the SERVQUAL model theoretically and statistically in relation to the nonlinearity of the perceived service quality. Perceived service quality measurement based on the SERVQUAL model assumes that consumers evaluate service quality by comparing their perceptions with their expectations. However, previous studies adopt linear factor analysis to discuss the SERVQUAL model. The present study assumes that consumers evaluate the service quality with a standard to admit the difference between their expectations and perceptions, and that their perceived service quality follows a nonlinear response. A nonlinear SERVQUAL model based on a nonlinear factor analysis model is proposed to understand the characteristics. The proposed model employs a threshold specification that represents the space in which consumers admit their discrepancy. The study extends a nonlinear factor analysis model to a nonparametric model in order to examine the functional aspects that from the perceived service quality. A nonparametric SERVQUAL model is adopted without any assumptions of the functional form. The empirical studies on the retail sector shows that the nonlinear model performs better than the linear model, and that the nonparametric model estimates the nonlinear function for perceived service quality. The results from the proposed model in our study provide insights in a consumer’s perceived service quality recognized as nonlinear and asymmetric. We conclude that the functional form for perceived service quality should be considered when specifying the measurement model for SERVQUAL. In addition, we discuss future work for a nonlinear measurement model and a nonparametric factor analysis.