This paper deals with the estimation of a random coefficient model. The virtue of this approach is that it considers firm heterogeneity, which conventional SFA models do not. Applying the model to Polish farms, the results indicate that the conventional random and fixed effect models overestimate the inefficiency score. In addition, the reasons for inefficiency are analysed. It is shown that despite the fragmentation of Polish agriculture, there is no evidence for scale inefficiency. Moreover, inefficiency could partly be attributed to factors, which affect the management input and requirements on farms.