The energy performance of residential buildings depends on a large number of interrelated factors. The present paper outlines an approach to developing a building thermal simulation model through real-time data and sensitivity analyses. To this end, three existing multi-family apartment buildings in Pristina, Kosovo, were selected. Initially, thermal simulation models were created using multiple data sources. Model outputs were further evaluated via comparison with available and measured data. Consequently, the most influential input parameters were identified and adjusted to calibrate the models. The resulting calibrated models can be deployed to investigate the potential of alternative retrofit measures.