The selection of manipulated and controlled variables as well as the structure and parameters of their interconnecting controller is an important synthesis problem that has received a lot of attention from the process control community. This paper presents a mathematical programming methodology for the solution of this multifaceted synthesis problem. The method is applicable to linear or linearised nonlinear systems. In addition, a quantitative index is proposed for the estimation of the advantages in using unstructured controllers, such as model predictive controller (MPC, when compared to classical structured controllers such as multivariable proportional-integral-derivative (PID) controllers. Two case studies are presented where the potential of the method as a rigorous technique for solving the regulatory control structure selection problem for linear systems is established.