The application of the response surface methodology in the optimization of industrial processes has had a great boom in recent decades, however, with a significant limitation, the null inclusion of qualitative factors in the noise variables. Since the methodology assumes the behavior of the noise factors as a continuous behavioral variable that follows a normal distribution. But what happens if this is not the case? How to treat a qualitative noise factor? What probability distribution would best fit the qualitative noise factor? What would be the correct inclusion of this type of noise factor in the methodology? This article summarizes the four-year research work from the mathematical approach to the new equations, case simulations using mathematical software and 2 real cases in maquiladora plants that manufacture plastic parts.