Floor failure on the longwall face is a complex problem due to the dynamic interaction between the surrounding strata and powered roof supports involving multiple variables. It can also be highly problematic and costly when not well understood as there are significantly less remedies available post event. This was the early experience at the case study mine where significant lost time was incurred due to unplanned delays in the first two panels. To detect and prevent such occurrences is clearly the best approach, hence a tool to predict both floor failure occurrence and magnitude prior to mining was developed. Several parameters were determined as the key drivers for the failure events that had been experienced at this site. These parameters were, in list of influence from highest to lowest, material strength, retreat rate and the presence of water. The parameters were evaluated using a semi quantitative risk rating system ‘Heave Risk’ and given a weighting per their apparent influence. This ‘expert system’ methodology was selected as the use of risk rating systems in underground coal mining is well established. The heave risk model was first applied to previously mined panels. It accurately identified all areas of known major floor heave, with clearly defined areas of low and high risk apparent. The heave risk model was then used prior to mining and evaluated post mining based on observed conditions. It was found to also be highly accurate for prediction. The variability of each parameter and its influence on the overall risk rating is illustrated and discussed.