Prescribed burning is a widely used strategy in forested landscapes to reduce the risk from wildfires to human lives and valued assets. The ability for managers to undertake prescribed burns is contingent on fuel, weather and operational constraints. In practice, not all areas nominated to be burnt get completed, and within burns that occur, not all the area is burnt. There has been limited research into rates of successful implementation of prescribed burns for fuel management and the degree to which the outcomes of prescribed burning are predictable. To better understand this, we undertook a quantitative analysis of prescribed burn outcomes in the State of Victoria, Australia between July 2009 and July 2015, investigating the likelihood of burns being completed and the time between nomination and completion. For those completed, we used Generalized Additive Models to investigate patterns of burn coverage, both for individual points within burns and aggregated to the burn level. We found a large proportion (29.6%) of scheduled prescribed burns were never burnt, and the time between nomination and completion was highly variable. The size of the nominated burn area was found to influence both the likelihood of completion and the likelihood of burns being postponed. Of burns that were completed, the proportion of area burnt was highly variable (mean 74%, median 84%). Patterns of burnt area were predictable: our point level model could predict the likelihood of any point within a burn being burnt with 66% accuracy from fuel type, aridity and temperature seasonality. The model was more confident in predicting what would remain unburnt than what would burn. Our burn level model explained 55% of variation in the proportion burnt from area weighted fuel type, average aridity, temperature seasonality and the size of the nominated burn area. An ability to foresee the outcomes of prescribed burn programs is important for managers in designing effective risk reduction programs.