Analyzing roadkill patterns is challenging due to roadkill data’s hierarchical and conditional nature. An animal on the road may become road-killed, and if so, its carcass may or may not remain on the road at the time of the survey, and the observer may or may not notice it. If we ignore this hierarchy of processes, we risk missing the true drivers of spatiotemporal observation patterns and drawing incorrect conclusions about the impact of roads on animal populations. Roadkill rates alone are poor indices of road mortality impact on populations because they are influenced to an unknown extent by variations in detection probability and the number of individuals vulnerable to vehicle collisions. Hierarchical models provide a suitable framework for improving our understanding of the effect of roads on animal populations by disentangling abundance and distribution from individuals’ risk of dying on the road.