We measure the technical efficiency of Peruvian and Chilean port terminals by taking into account the existence of unobserved heterogeneity within the sample. We follow a Latent Class Stochastic Frontier Model (LCSFM), in order to account for possible technological differences among terminals. We will also identify whether a successful reform process has contributed to attracting investment that has rapidly expanded the infrastructure and new technology. The results of this paper show that LCSFM, with two classes, fits the unobserved heterogeneity better than the four Standard Stochastic Frontier Models (SSFMs) that we used as alternatives.