Abstract Many processes and techniques have been proposed to improve the heavy oil recovery from fractured reservoirs. Such complex processes require careful design to achieve optimal efficiency with minimal costs and environmental impacts. Steam injection is one of the options for heavy-oil recovery from fractured reservoirs but significant steam requirement for effective matrix heating due to heterogeneous structure poses important challenges in terms of cost, water availability, and environment impacts due to water processing and steam generation. Al-Bahlani and Babadagli, 2008 , Al-Bahlani and Babadagli, 2009a proposed a new process called Steam-Over-Solvent in Fractured Reservoirs (SOS-FR) by adding solvent component to minimize the heat needed. The SOS-FR technique consists of a heating phase using steam injection, subsequent solvent injection, and low temperature steam injection for solvent retrieval and additional oil recovery. Optimization of this process is a critical step to determine optimal injection (and soaking) schedules as the heterogeneous structure of this kind of reservoirs may easily yield an inefficient process due to high cost and excessive use of steam and solvent. In this study, we adopted a global optimization scheme, where genetic algorithm is integrated with orthogonal arrays and response surface proxies for better convergence behavior and higher computational efficiency, to optimize the SOS-FR process for cyclic injection option. The results show that one may be able to double the profit obtained with the benchmark model using the optimal injection scheme suggested by our optimization procedure. The results are valid for models with low fracture density; further work should be performed using models with high fracture density and that the fracture orientation is not in the direction of the injector to the producer.