This paper presents the effects of mixing non-homogeneous bicomponent granular structures by the flow method with the use of an inverted cone insert as a supporting element. The analysis of homogeneity of mixtures after 10 con­secutive flows was carried out based on computer image analysis. Based on these tests, the use of a nonlinear regression for the modelling of two selected parameters of the process was proposed. Spatial dependence was described by the relation between the variance of the tracer distribution (dependent variable) and density ratio of the mixed components, and the diameter of the cone insert used (independent variables). The mathematical description was made with the use of a ‘nonlinear regression’ module. The results obtained in these tests proved that the use of supporting insert improves the degree of mixability of granular materials. The two-dimensional model in the form of a quadratic function may be a formula for describing the influence of selected parameters on the homogeneity of a granulate mixture before the beginning of the mixing process.