The sensitivity of computer fire modelling using results from NIST's Fire Dynamics Simulator (FDS) to a set of input parameters related to fire growth has been analyzed. The scenario simulated is the real-scale Dalmarnock Fire Test One and the modelling results are compared to the measurements. Fire size and location, convection, radiation and combustion parameters were varied in order to determine the associated degree of sensitivity. Emphasis is put in the prediction of secondary ignition and time to flashover. In this context and while keeping the HRR constant, simulations of fire growth are significantly sensitive to location of the heat release rate (HRR), fire area, flame radiative fraction, and material thermal and ignition properties. The simulations are relatively insensitive to the heat of combustion (while keeping the HRR constant), the soot yield and the heating from the smoke layer. The results indicate that the future development of successful fire forecast methodologies of fire growth using CFD must focus on the global HRR as well as the important parameters identified here.