Blending operations are widely employed in the petroleum and chemical process industries in order to maintain products within specifications. It is often impractical to analyze stream compositions when the number of components present is very large (e. g., gasoline blends). In such cases, there is a need for mixing rules that will allow prediction of the final blend properties from knowledge of constituent stream properties and the ratios in which they are blended. It is shown that this can be done if the q-fractions (r, s) mixture rule applies and the composition descriptors can be expressed in terms of a second, known mixture property. This principle is illustrated for the surface tension of the system benzene + n-hexane + cyclohexane with the refractive index chosen as the pure component composition descriptor. In essence, the q-fractions (r, s) model is a composition-adjusted weighted double power mean. It is a general form that includes several simpler models as special cases. The prediction accuracy of this family of models is demonstrated using surface tension, viscosity and refractive index data for the ternary system benzene + cyclohexane + hexane. Parameter confidence intervals were determined using the bootstrap.