The D-optimal design criteria use a maximum determinant of matrix information in determining the optimal design. The D-optimal design is most often used because it is quite simple and reduces uncertainty in estimating model parameters. However, this criterion very depends on the assumption of the model. The improper model assumption can reduce the accuracy of optimal point estimation. The Bayesian method can handle the dependence of D-optimal criteria on model assumptions. Design matrix in the Bayesian D-optimal is divided into primary and potential terms. The primary term is a certain parameter then the potential term is a parameter that possible in the model. Each primary and potential term has a prior distribution which gives us information in determining the posterior distribution. The Bayesian D-optimal algorithm will be applied to a cake mixture that consists of three components with certain constraint functions. The limited development of research in the Bayesian approach to D-optimal design, this study is limited by assuming linear model for numerical responses. The result, from nineteen candidates, twelve were chosen as optimal points. It found that the optimal point to be consistent in the selection of . The high number of  is indication that the potential term is feasible to the actual model. The optimal point obtained also represents the overall points in the design area.