Predation ecology and evidence-based conflict management strategies require reliable and accurate identification of individual predators. Identifying predators is, however, complex, as they are secretive and individual identification is difficult. Trace DNA that predators leave behind at kill sites might provide an effective strategy to identify them but remains poorly assessed. We use non-invasive genetic samples from kill sites to assess their utility for predator identification. We systematically investigated 198 livestock kills in two of Central India's critical source tiger populations: Kanha and Bandhavgarh Tiger Reserves. We collected 342 saliva, 33 scat and 395 shed hair samples as potential sources of predator DNA. Individual tigers were identified using up to 123 SNP markers. All sample types were highly successful (>95 %) in identifying tigers. Individual tigers were identified at 85 % of all kill sites using at least one sample. Shed hair samples were most effective for individual identification, followed by saliva and scat. Sample type and sampling season were the primary determinants of the number of SNPs typed per sample and individual identification success. Based on the collection site and type of sample in identification, we classified species and individuals into three categories: true predator (high confidence as predator), circumstantial predator (medium confidence) and predator uncertain (low confidence). Individuals were classified as a true predator at 72 sites, circumstantial predator at 34 sites, and predator uncertain at 49 sites. We suggest that when paired with systematic kill site investigation and sample collection, these methods can be effectively used to understand predation ecology better and facilitate evidence-based conflict management.