Over the last 10 years, new tools called Urban Building Energy Models (UBEMs) have been developed that are capable of evaluating energy-saving measures for large building stocks through dynamic simulations. However, these tools require reference buildings—also called archetypes—to provide reliable results. Such archetypes are virtual buildings that represent a significant part of the building stock, and their definition requires geometrical and thermal information that should be retrieved from a range of sources. The Project of National Significant Interest PRIN 2020 “URBEM” project aims to provide a national database of archetypes for use with UBEM tools. This paper first presents the statistical methodology applied to the analysis of available data sets regarding residential buildings located in Calabria and Sicily. These datasets consisted of 720 dwellings in Calabria and 274 dwellings in Sicily, respectively. Then, based on a filtering and clustering procedure involving different climate zones and construction periods, exemplary archetypes are derived that could eventually be exploited by public authorities to develop energy policies and incentives to improve the efficiency of real estate assets, involving the renovation of the building envelope, the use of more efficient mechanical systems and the exploitation of renewable energies.