This work presents a methodology to choose the convex structure for nonlinear gains in approaches based on Takagi-Sugeno models. The proposal faces the problem of numerical complexity in LMI-based approaches within this area; it gives a finer convex structure for controller and observer matrices depending on two criteria: less number of LMI constraints or more decision variables while keeping the same number of LMIs in traditional approaches. The advantages of the proposal are illustrated via numerical examples.