Based on a microhistoric analysis of a serious accident in 1947, this paper sheds light on the challenges of accident management on the bus network in Paris, from the 1920s to the 1980s. Contrary to many transport systems, the bus has the specificity of being both a system linked to an operator marked by industrial logics, and one of the systems evolving in a shared public space with many others. This configuration explains why the role of information in this sector is crucial, although it is difficult to collect. The paper follows these information challenges to understand how different ways of formulating, analyzing and using accident data coexist. First, they aim to establish the responsibilities of the accident and feed any subsequent procedures. Then, once multiplied by the number of accidents occurring on the network, they make it possible to estimate the extent and nature of these accidents. Finally, they are at the base of prevention policy, which can use them for areas of improvement and measure the effects of interventions. These pluralities of information lead to a plurality of their uses, which are superimposed. The available tools played a key role in the establishment of an effective statistical apparatus from the 1950s onwards, but the intellectual context opened by the interaction between data collection and the implementation of psychotechnical methods from the 1920s onwards promoted their implementation. There has been a reinforcing effect between psychotechnical approaches and statistical approaches that leads to the hypothesis that the isolated accident is difficult to anticipate, predict, know and treat, while the mass of accidents opens the way to quantitative methods with a more visible and controllable efficiency.