Bayesian optimization algorithms, i.e., algorithms using Gaussian Processes, are often resorted to when the number of calls to the objective function is strongly limited. In the last decade, these algorithms have been extended to multi-objective optimization and parallelized versions have appeared. In this talk, we show how a faster multi-objective optimization is possible by complementing Bayesian approaches with a preference point. The gain comes in part from a first phase of the search, where only one point of the Pareto front, deducted from the preference point and the Gaussian processes, is targeted. Once convergence to this point is detected, the portion of the Pareto front that can be attained within the remaining budget is estimated and becomes the final goal. The method involves a new analytically tractable Bayesian criterion, the mEI. A default and updates of the preference point are proposed. The method relies on Pareto front simulations. We will also present how, at the two stages of the algorithm, new search points can be produced in batches, making the method parallel while keeping the choice of the points optimal in a Bayesian sense.