The selection of a statistical/mathematical software (SMS) system is complicated. The satisfaction derived from a system depends on many variables. This study analyses the influence of SMS predictor variables on satisfaction as determined by multiple regression. The null hypothesis, that multiple correlation coefficient is zero, is rejected. The alternative hypothesis is accepted. This study confirms the theories that suggest that SMS ease of operation, computer reliability, and ease of programming are the major determinants of computer user satisfaction. The factors identified were used in the design of an SMS Expert System based on Artificial Intelligence (ESAI).