Application of customer satisfaction study to derive customer knowledge

Abstract

Customer knowledge has become one of the most important economic assets to a firm. Knowing what customers value most in a company's products or services helps companies optimize resource utilization in areas for continuous improvement based on their needs and wants. This study extends previous research in customer satisfaction by showing how a customer satisfaction study can help answer questions, such as why customers are satisfied or dissatisfied, whether the company's complaint handling is effective and how to improve customers' satisfaction. We surveyed 532 customers served by a commercial banking organization in Taiwan. In addition, researchers demonstrated that there are differences in satisfaction levels among different genders. To explore further the reasons for these differences, we separated our sample to compare the specific causal links between the two gender samples and we draw broader inference from the results. Partial Least Squares (PLS) was used to test the theoretical model. PLS not only bridges theoretical and empirical knowledge but also facilitates the construction of a strategic management map and provides latent variable scores that could be utilized to report CSI scores. Our findings provide support for the application of CSI study to derive customer knowledge, which also reveals some interesting suggestions for future research.

Keywords:

• Customer satisfaction index
• PLS

Notes

1. PLS requires a minimum sample size that equals 10 times the greater of (1) the indicators on the most complex formative construct or (2) the largest number of antecedent constructs leading on endogenous construct.

2. The average R ² values of overall CSI and loyalty are 0.75 and 0.36 in ACSI study (see Fornell et al., 1996).

3. The formula of construct score is:

where h is the number of manifests variable within the latent variables and w_i are the unstandardized weights.

4. The ACSI score for banking industry can be obtained from: http://www.theacsi.org/fourth_quarter.htm#ban.

5. The calculation procedure for comparing corresponding paths across structural models is:

where S _pooled = pooled estimator for the standard deviation, t = t-statistic with N ₁ + N ₂-2 degree of freedom, N_i  = size of sample i, SE _i  = standard error of path in structural model of sample i, and PC _i  = path coefficient in structural model of sample i.