Currently there is limited design guidance for calculation that involves an unchannelled balcony spill plume. The current guidance for unchannelled spill plume can be obtained from CIBSE (TM19) and BS PD 7974 (Part 2). The guidance is a result of an analysis performed by Law (1995) on a set of experiments performed by Hansell et al (1993). Recently a series of unchannelled balcony spill plume experiments have been performed by Harrison (2009) at 1/10th-scale as part of a research project. Harrison (2009) developed a simplified design formula to calculate the mass flow rate from unchannelled spill plume. Results from Harrison suggested that an unchannelled balcony spill plume is a complex problem and his experiments were limited by the experimental facility. Harrison (2009) recommended that this problem should be supported with computer modelling for design. Research from Harrison (2009) also resulted in a correlation that can be used to calculate the effective width of an unchannelled balcony spill plume. This research is primarily a numerical simulation of the unchannelled balcony spill plume using Fire Dynamics Simulator 5 version 5.3.0 (FDS 5). FDS 5 was used to model unchannelled balcony spill plume experiments performed by Harrison (2009) at 1/10th-scale. Results from Harrison‟s (2009) experiment were used to verify the results from the FDS simulation. A good match was obtained between experimental results and FDS simulation results. The 1/10th-scale model was extended to full-scale using the scaling laws. Then, the model was used to perform a series of unchannelled balcony spill plume simulations with variables of balcony breadth, compartment opening, fire size and height of rise of the spill plume. Mass flow rate readings from simulations were used to improve the effective width correlations developed by Harrison and an improved effective width correlation was developed. Unchannelled balcony spill plume mass flow rate prediction with the revised effective width correlation was compared with mass flow rate predictions using the effective width correlation from Law (1995) and Harrison (2009).