In this paper is presented an evaluation of different control structures for a fluidized catalytic cracker. A systematic methodology for evaluating such structures was developed which is suitable for complex nonlinear processes in which the number of manipulated variables is smaller than the number of variables that are part of the specifications. It is shown that the choice of the control structure and the variables entering it is far more important than that of the multivariable algorithms to be used. Criteria for the choice of the measured and manipulated variables entering the dynamic structure are presented. The role of additional so-called slow variables in steady state control is discussed and demonstrated by examples. It is shown that linear algorithms are sufficient for dynamic control,but nonlinear considerations dominate the choice of the control structure. This paper is one of the first that thoroughly discusses the compromises and choices a designer faces in designing a control system for a complex nonlinear system. Areas in need of further research are outlined.