The simplex-centroid mixture design is used for the development of new food products. This design saves time and money by generating mathematical models that predict the centesimal composition of new formulations without the need to repeat experiments. Thus, it is an important innovation tool for the food industry. The aim of this work was to produce different formulations of whey fermented beverages with buttermilk and cagaita pulp (Eugenia dysenterica) using the simplex-centroid design. Therefore, 7 formulations with different concentrations of independent variables were prepared: whey/buttermilk mix (x₁), milk (x₂), and cagaita pulp (x₃) and its centesimal composition (dependent variables) was determined. Using design, it was possible to determine the best regression model and to generate the equations that predict the behavior of each dependent variable by changing the relation between the concentrations of the independent constituents. The following equations were obtained, where y = response estimate. For moisture: y = 78.12x₁ + 84.33x₂ + 83.41x₃ + 8.25x₁x₃ – 13.38x₂x₃; lipids: y = 0.50x₁ + 3.40x₂ + 2.73x₃ – 1.76x₁x₃ – 5.83x₂x₃; proteins: y = 0.50x₁ + 1.55x₂ + 2.58x₃ + 3.49x₁x₂ + 0.02x₁x₃ – 0.20x₂x₃; and total energy value: y = 88.50x₁ + 75.89x₂ + 79.49x₃ – 45.53x₁x₃ + 24.00x₂x₃. The interaction x₁x₃ shows synergism with moisture and antagonism with lipid and total energy value content. The interaction x₂x₃ shows antagonism with the moisture, lipids, and protein. A positive correlation was observed by interaction x₁x₂ only with the protein content of formulations. By using the models generated through the simplex-centroid design, it is possible to predict components like moisture (which influences the shelf life of the product), lipids and proteins (which influence the viability of microorganisms and the sensorial characteristics of the product), as well as total energy value and other constituents of the food. Therefore, the composition can be predicted and modified without having to do new experiments.