The paper presents am optimization framework to obtain cost effective, high performance blends of transformer oils. The challenge in blending involves finding the particular blend composition, which can meet the product specifications at lowest overall cost. Recent investigations on mixtures of insulating fluids (mineral oil and ester liquid) as an alternative to pure oils, have shown encouraging results and a rough estimate of the optimal blend composition has been given based on experiments carried out for a few mixture compositions. However, for precise determination of the optimal composition the experimental approach can incur high cost because of the potentially large number of trials involved in finding the optimal blend. This paper suggests a complementary approach for precisely predicting the optimal mixture composition using optimization techniques. Utilizing the mathematical model of properties (as function of composition only) estimated from those few initial experiments, the blending problem is formulated and solved as a multi-objective, non-linear goal programming optimization. The proposed approach is demonstrated for the same mixture of mineral oil and ester liquid and a more precise estimate of the optimal composition than the one found directly through experimentation as reported in literature is obtained.