Functional renormalization group (RG) method is suitable to perform renormalization non-perturbatively. However, the approximated RG equations depend on the choice of the so called regulator function. Thus, optimisation is required in order to increase the predicting power of the functional RG approach. As a result of a general optimisation strategy, the Litim-Pawlowski method, one obtains Litim's regulator which is one of the most frequently used regulator function. It is a continuous function with a compact support but non-differentiable, hence, it confronts to the gradient expansion. Another optimization scenario is based on the Principle of Minimal Sensitivity which can be used at any order of the gradient expansion but regulators cannot be compared to each other. In my talk I show how a compactly supported smooth regulator which reduces to all major types of regulator functions in appropriate limits can solve these problems.