The aim of this study was to evaluate the effect of increasing accuracy of genomic selection, achieved by larger training set, on the economic efficiency of breeding programme in dairy cattle. A deterministic model compatible with the Iranian Holstein population, was developed. Gene flow method was applied to estimate genetic and economic parameters of simulated programme over 20 years for milk production. By increasing the number of individuals in training set from 500 to 3500, the accuracy of genomic evaluations increased from 0.62 to 0.82. However, maximum economic efficiency (13.83) was achieved at the accuracy of 0.72 using 1000 individuals in the training set, indicating that the genomic accuracy has an optimum value with respect to economic efficiency. A range of genotyping costs, from 50 to US$400, were examined and the corresponding optimum accuracy was estimated. Although the genotyping cost has a negative effect on the economic efficiency, but it had a minor effect on the optimum accuracy. Varying the heritability from 0.18 to 0.36 did not affect the optimal accuracy. The results of this study demonstrated that achieving higher accuracies of genomic value through allocating more individuals in training set, would not necessarily lead to more economic efficiency. The effect of varying genotyping cost on optimal structure of genomic selection programme could be negligible. The optimal size of training set would not be affected by heritability, although the maximum economic efficiency of selection programme could be affected. ARTICLE HISTORY