Civilian fatality figures are a limited, if important, data point that influences the ability of researchers to study patterns of violence and evaluate policy responses intended to end violence. However, across datasets that track such violence there are significant differences in how and what is counted, this has direct bearing on how atrocity endings are understood and what policies might best be applied. There are often good reasons for data variation that cannot always be resolved. Nonetheless, it is important to understand and itemize the factors that influence these differences. Highlighting the relationship between the evidence base for and the construction of research consensus about civilian fatality figures, this paper draws on case study research to demonstrate how variations in the evidence base that defines an ending can fundamentally alter the narrative about historical mass atrocity events, with significant implications for how protection is conceptualized.