Even if energy efficiency represents a crucial issue for the sustainability of the manufacturing industry, the companies, especially SMEs (Small and Medium Enterprises), need to be encouraged in investing their time and resources for this goal. One of the means to facilitate this effort is to provide some reference performance indicators for the energy efficiency of their production processes. On the other hand, since the enterprises are very heterogeneous and make use of very different production processes and materials, the values of these reference indicators, such as the Specific Energy Consumption (SEC), have a high variability and risk not to be representative for all the factories. The dispersion of the data has to be decisively decreased in order to represent a useful indicator for a factory. This goal is pursued here by means of an energy consumption characterization model starting from the data of 140 factories, which have been collected through self-analysis software tools in the framework of the European project SET (Save Energy in Textile SMEs). The approach proposed here is based on the separation of the auxiliary energy uses from the production process energy consumption. The analysis refers to the textile industry and it is performed with a focus on the electrical consumption in yarn production factories. The outcomes show a correlation with some production variables, such as the raw materials, and allow to reduce the relative errors of the reference indicators from about 80% to about 30-40%. Moreover, the results are complementary and consistent with respect to the literature and enrich the public available data for textile industry.