Fisheries production is subject to a significant variability caused no only by the stochastic nature of fisheries due to uncontrolled environmental and biological conditions, but also by factors related to production activity. It is necessary to take into consideration all these factors to avoid biases on production model estimates. This paper aims to go through this variability in Stochastic Frontier Analysis to account for observed and unobserved heterogeneity together with technical efficiency and randomness. This paper exemplifies the application of a Latent Class Stochastic Frontier model to the anchovy fishery of Southeastern Black Sea to account for production frontier heterogeneity. Results show a mean level of technical efficiency of 55%, which is higher than those produced by the standard stochastic frontier model. Moreover, results allow identifying two latent classes in the fleet. They also provide sound scientific advice for de management of the fishery.