In this work, we consider the Flexible Job Shop problem (FJSP), which consists of ordering several jobs involving operations, in its multiobjective form. We focus on the integration of the DM's preferences in the optimization process when considering three criteria: makespan, total machine processing time and balanced machine utilization. We assume that the DM is able to identify reference performance levels, but that the heterogeneity of the criteria scales makes it difficult to use compensatory models. We use the Ranking using Multiple Profiles (RMP) model embedded in a hybrid Tabu Search method. Lastly, we compare the a priori and a posteriori strategies.