In this paper, a theoretical analysis of the latent variable model predictive control (LV-MPC) algorithm, originally proposed by Flores-Cerrillo and MacGregor,(1) is presented. The properties of the algorithm in terms of stability, robustness, and control performance are analyzed. An off-line design methodology, using Internal Model Control framework (IMC), to determine the adequate range of the LV-MPC parameters is proposed. The excellent control performance and robustness of the off line design methodology Is illustrated for the temperature tracking of an emulsion polymerization process and an exothermic chemical reaction. system.