Traditional process capability analysis takes assumption that there is only a variation source in production/process quality characteristics. This paper presents a method for estimating process capability indices, which roots in nested mode of experiment design, and uses variance components to estimate total variance as well as each variance component. In addition, the confidence interval of Cp is derived based on MVA. A case is studied, and it is shown that MVA is good at analyzing the process variation patterns, so the method can remark the process scientifically and rationally.[