When using "conventional" approaches for the optimal sizing of electrical machines, the application of optimizationalgorithms makes it possible to search for the values of their input characteristics (dimensions, power supply, materials, etc.)giving the desired performances (torque, efficiency, mass, etc.), provided that a certain number of constraints are satisfied.This (one-way) approach has several weaknesses, the main one being that it is not clearly apparent whether this sizingproblem has at least one solution. If sets of solutions exist, it becomes interesting to determine the characteristics of all thesevalid machines. On the other hand, if no machine can be defined by exactly meeting the specifications, it becomes more usefulto look for the constraints to be relaxed to reveal solutions, or to determine and then choose among all the quasi-optimalconfigurations those that are the most interesting.In this context, this work aims to respond to these issues by introducing a methodological procedure giving assistance inthe optimal sizing of electrical machines in particular, and in optimization in general.In this procedure, a number of numerical processing tools are implemented, to offer the opportunity to understand morefinely, typically through 2D graphical representations, the complexity of optimization design problems. In other words, thiswork is an attempt to find an intermediate configuration between the pure optimization approach and the one based solelyon modeling, in the sense that these models are first represented graphically in order to understand the complexity of thecorresponding optimization problem.The application of this approach is illustrated by the presentation of the main steps of the optimal design of a PermanentMagnet Synchronous Machine.