Predation is a constant risk for most primates, impacting demography, population dynamics, activity patterns and social behaviour. Data are limited on both the rates of predation and its spatial and temporal variability. We present long-term observations of Cryptoprocta ferox predation on rainforest sifakas in Madagascar, Propithecus diadema at Tsinjoarivo (22 group years) and Propithecus edwardsi at Ranomafana (73 group years), derived from intensive observations based on ongoing behavioural studies. Average per capita offtake rates are relatively low (0.06–0.07), but temporal variability is high (kills are clumped in time). This is consistent with Cryptoprocta ecology; individual home ranges are much larger than sifaka ranges, and individuals may hunt in a subsection of their range until prey density is decreased, then move on. These results have broad implications. First, in terms of the evolution of anti-predator strategies, it now becomes important to ask: (1) whether average or peak predation rates determine the strength of selection and (2) whether antipredator strategies (e.g. vigilance, sleeping site selection) fluctuate interannually, reflecting recent experience. Second, in terms of population ecology, Cryptoprocta may have disproportionately large impacts on the (small) sifaka groups, even driving groups to extinction (as observed at both sites). Third, the disappearance of groups has important implications for conservation. When this happens in continuous forest (as at Ranomafana), home ranges will likely be re-filled over time, whereas in isolated forest fragments (as at Tsinjoarivo), recolonization is less likely. Thus, conservation planners should consider predation as a potentially important proximate cause of extirpation in fragmented landscapes, even when resource density and quality could otherwise sustain populations. Considering the effects of predation can be useful in (1) decisions regarding the allocation of limited conservation resources, including which landscapes to invest resources in and (2) investigating ways to increase resilience of prey species.