Taguchi method and RSM (response surface method) are two of the most well known DOE (design of experiment) techniques. The levels of parameters are recommended to be taken far apart in the Taguchi method in order to cover a wide region to increase the chance of capturing nonlinearity of the relationship between the control and control factors. On the contrary, as long as the optimum is located within the region, RSM needs it to be as small as possible to identify the exact optimum. In this study, the Taguchi method is used to determine the rough region first, followed by RSM technique to determine the exact optimum value during turning on a CNC lathe. A new region reducing algorithm is introduced to narrow down the region of the Taguchi method for RSM. To achieve the goal, the result from the Taguchi method is fed to train the artificial neural network (ANN), whose optimum value is used to drive the region reducing algorithm. The proposed algorithm is tested under different cutting condition with different insert and work material. Data located in the literature is also used to inspect the adequacy of the region reducing algorithm. Both results show that the introduced algorithm has a good region reducing capability. In a separated experiment, it is shown that the obtained cutting condition from RSM gives a better result than that from the Taguchi method.