AbstractFew studies have investigated the factors that influence roadkill occurrence in developing countries. In 2013, we monitored a 100‐km section of the road (comprising the R572 and R521 regional highways and the D2662) that pass through the Greater Mapungubwe Transfrontier Conservation Area in South Africa, to assess the possible factors influencing roadkill. Over a period of 120 days, and across the three ecological seasons, we recorded 981 roadkills (rate = 0.08 roadkill/km/day) from four vertebrate taxonomic groups. We generated predictive models of roadkill from one combined data set that considered eight variables identified from the literature as potential correlates of roadkill. The model that included the distance of the fence from the road, habitat type adjacent to the road, and the presence of a hill in the road (i.e., elevation) or a bank on the side of the road best explained roadkill occurrence. More roadkill was predicted to occur in both open and dense mopane and dense mixed bushveld habitats, on a hill, when there was a bank on the side of the road, and as the distance between the road verge and a fence decreased. Our model provides some insight into the significant predictors of roadkill occurrence and is therefore a valuable tool in identifying sites of high‐potential roadkill frequency and formulating mitigation measures for reducing road mortalities.