Customer satisfaction has a major influence on customer loyalty and knowing its determinants enables the manager to improve his quality of service and the customer lifetime value. Nevertheless, customer satisfaction is the result of a service encounter whose characteristics are numerous and often correlated. This paper describes the principle, characteristics and performance of several regression techniques in order to determine the weight of criteria, with satisfaction data collected by questionnaire. An empirical comparison on convergent validity, stability and dispersion of weights draws to the conclusion of the superiority of PLS and sequential stepwise regression.