This paper develops an efficient multistage parametric approach for eigenstructure assignment in linear multivariable systems by output-feedback control. The sets of closed-loop eigenvalues and associated eigenvectors are suitably divided into subsets and the entire eigenstructure is constructed by parts in two (or more) consecutive stages. This allows the subsets of right and left eigenvectors to be assigned in separate stages and thus the computational algorithm is relaxed from the orthogonality conditions. The number of effective free parameters beyond eigenvalue assignment is determined, and the notion of redistributing these parameters among the assignable right and left eigenvectors is introduced. Numerical examples are included to demonstrate the design procedure.