In this paper, we examine the application of SFA method with time-invariant inefficiency and assess its estimation of inefficiency when applied to cross section and panel data. By using simulation methods, we look at the effect of unobserved heterogeneity on the estimates of inefficiency in both cross section and panel. In the presence of unobserved heterogeneity and significant variance in the inefficiency term, stochastic frontier estimation of inefficiency can be significantly different in panel and in cross section. This finding accords with analysis of actual data from the postal sector. We then suggest an estimation method for cost frontier when inefficiency is time-invariant and with unobserved heterogeneity.