Abstract:While computing Customer Satisfaction Model, structural equation model is always used to solve the influ- ence of complicated multi-dimensional. Partial Least Squares (PLS), used in computing of structural equation model, works effectively with the problem that a large amount of latent variables cannot be sampling directly. In the process of model analysis, the PLS path model builds up the relationship between latent variables and manifest variables, and enhances the intuitive and generality of model. In order to improve the efficiency of PLS, we parallelize the computational process to get more efficiency. This article mainly talks about the PLS path model based on customer satisfaction model and the distributed algorithm based on Java multi-thread. It is proved on multi-core computer that PLS path model implementing on multi-thread can obvious improve speed of computing.