One of primary assumptions on DEA is that all DMUs for evaluation should be homogeneous. In comparative analysis among DMUs with relative efficiency measurement, it should be evaluated under identical conditions by ruling out external environmental influences. In this study, a measurement of innovation efficiency using the three-stage approach is performed. The approach employs DEA to measure relative efficiency and Tobit regressions to control external variables affecting innovation activity. The approach applied to firms in Korean Innovation Survey: Service Sector 2003 and 2006. Final efficiency scores of the approach represent net efficiency of the innovation. This study found that there is a increasing on technical efficiency of third stage, and it has difference with first stage significantly. Besides, a decrease on standard deviation of third stage is found. It means DMUs biased lower due to unfavorable condition and ones biased higher due to favorable condition are fallen into an identical operating environment through the approach. A measurement of net efficiency, excluding external effects, ensures the homogeneity of DMUs so that improves the reliability in terms of its analysis results. This study is expected to provide a direction and to be a valuable reference to further evaluation of innovation performance in Korean service sector.