This paper proposes a methodology to calibrate multi-zone air flow models in an existing building. The methodology has been demonstrated with the widely used multi-zone model CONTAM and applied to two buildings, one synthetic and one actual. The methodology builds on previous work, with the calibration process enhanced by the use of carbon dioxide tracer gas test data. The proposed methodology includes performing an airflow-based sensitivity analysis to identify influential system drivers, a tracer gas-based sensitivity analysis to identify macro-zones for model reduction, and finally, a process of tuning the model parameters. It was found that for mechanically ventilated commercial buildings, calibrating multi-zone models by reconciling differences between measured and predicted tracer gas behavior is more strongly impacted by room air change rates than by airflow path leakage parameter tuning.