The optimization of a mixing unit by means of CFD (computational fluid dynamics) and RSM (response surface methodology) is presented in this work. The starting geometry is the one studied by Fang and co-workers [14] consisting in a T-type microchannel with the mixing unit inserted in the straight main channel. The mixing takes place at a very low Reynolds number and is promoted by means of two bars at 69 each with respect to the mixing unit walls. These angles are the input parameters of the optimization process while the output ones are the mixing efficiency, the needed pumping power to run the channel and the mixing energy cost. Fang and co-workers found numerically an efficiency of around 22% when one mixing unit was employed, while an almost perfect mixing could be reached by using 28 of them. However, they did not carry out a grid convergence study and their results were got for just one mesh. Due to this, before the optimization tasks and thanks to the Grid Convergence Index, it is shown that Fang and co-workers' mixing efficiency has an uncertainty of around 40%. This is due to the fact that, with the optimal grid, the mixing efficiency is around 12.5% which is quite far from what Fang and co-workers said. Additionally, with the RSM and by using the optimal mesh, it is found that the optimal angles alpha(1) and alpha(2) are: similar to 76 degrees and similar to 62 degrees to get the highest efficiency which is around 14%; and similar to 72 degrees and similar to 74 degrees to get both the lowest pumping power to run the channel and the lowest mixing energy cost. The performing of the micromixer as a microreactor, where a basic reaction must take place, is finally assessed.