This study approaches the measurement of energy efficiency development in the manufacturing sector at different aggregation levels from a production-theoretic structure, using the method of Data Envelopment Analysis (DEA). Using data from the German and Colombian Annual Surveys of Industries for the years 1998 to 2005, the analysis compares the energy efficiency performance in the German and Colombian energy intensive sectors (EISs) at three levels of aggregation, and then applies several alternative models. The results show considerable variation in energy efficiency performance in the EISs of both countries. Comparing the results across models, it was found that, in the German industrial sector, the three measures of energy efficiency were similar, indicating that an appropriate combination of technical efficiency and cost minimisation are necessary for energy efficiency improvement. In the Colombian industrial sector, the highest energy efficiency measured was from the cost minimisation model, suggesting that the relative energy prices have not generated the right incentives to improve energy efficiency. A second-stage regression and correlation analysis reveals that, in German EISs, energy costs and investments have played an important role in energy efficiency performance and decreasing CO2 emissions. In Colombian EISs, inter-fuel substitution was the most significant variable. Finally, the results of DEA models show a significant correlation with the traditional energy efficiency measure, indicating that the energy efficiency measured through DEA could be complementary to the energy intensity in analysing other key elements of energy efficiency performance in the industrial sector. In addition, energy efficiency is one of the quickest and the most efficient strategies for reducing energy demand and CO2 emissions.