Restoration outcomes are notorious for being unpredictable. Similar restoration actions, carried out on seemingly similar sites, can result in variable outcomes and this unpredictability challenges our capacity to meet specific goals. In turn, this points to the need to develop predictive capacities in restoration. But what will this look like and what challenges must be overcome to develop restoration into a predictive science? We outline an approach for predicting restoration outcomes and describe key challenges that must be overcome, to put this vision into practice. We then describe a research project, where we are putting these ideas into practice, to predict the outcomes of longleaf pine woodlands restoration.