Power quality management systems can provide information related with voltage sags as the duration, remnant voltage and affected phases associated with a point of measurement. Given a strong relation of these parameters with severity and energy indexes of voltage sags an improvement of the segmentation and characterization process is needed. For this reason a new method of analysis of voltage sags is presented in this paper. This new method uses a single parameter for the segmentation and classification process which is generated from the tensor concept of voltage signals. This parameter has a constant behavior under ideal conditions and oscillations during the occurrence of voltage sags. The classification process is performed from the maximum and minimum peaks of the parameter under a voltage sag condition. Comparisons between characterization results of the proposed method and a power quality management system in an industrial distribution system are shown. The shortcomings in the characterization that may result from the segmentation stage and its subsequent impact on the calculation of voltage sags indexes on an industrial circuit of the local utility feeder are shown as well.