Since the 1990s, electricity distribution networks in many countries have been subject to incentive regulation. The sector regulators aim to identify the best performing utilities as frontier firms to determine the relative efficiency of firms. This paper develops a nested latent class (NLC) model approach where unobserved differences in firm performance are modelled using two zero inefficiency stochastic frontier' (ZISF) models nested in a latent class stochastic frontier’ (LCSF) model. This captures the unobserved differences due to technology or environmental conditions. A Monte Carlo simulation suggests that the proposed model does not suffer from identification problems. We illustrate the proposed model with an application to Norwegian distribution network utilities for the period 2004-2011. We find that the efficiency scores in both LCSF and ZISF models are biased, and some firms in the ZISF model are wrongly labelled as inefficient. Conversely, inefficient firms may be wrongly labelled as being fully efficient by the ZISF model.