Deep-retrofit planning for existing buildings demands high accuracy in energy modelling prediction that minimises the gap between actual and simulated scenarios. A large set of interacting variables and uncertainties in energy performance modelling causes perturbations that can be minimised via model calibration. In this work, a novel multi-stage automated calibration methodology was developed using a case study of a partially-retrofitted university building ( > 35 yrs old) in Ireland. The methodology enables the analysis of models for Indoor Environmental Quality (IEQ) variables along with energy demand. Due to the higher number of uncertainties in the model, a sensitivity analysis was conducted on the model that is both calibrated and validated as per ASHRAE Guide 14 indices of Cv(RMSE) and NMBE. The calibration process was automated using the optimisation algorithm NSGA-II with two sets of reference data i.e. monthly utility and hourly indoor air temperature. Results demonstrate that using only utility data for calibration did not result in accurate predictions of the thermal environment; thus, a second stage was used to improve the model prediction giving a Cv (RMSE)(hourly) = 17.0-25.5% and NMBEhourly = 3.6-10.0% for indoor air temperature across multiple zones. This paper demonstrates an effective staged approach for creating calibrated models of old buildings under high uncertainty that can be used to influence large-scale decision making for retrofits focused on improving indoor environment quality and energy performance.