This paper is a contribution towards an operational use of large floating car data (FCD) in traffic 47 management. The work focuses on a practice-ready approach on highways. The goal is to detect 48 in real-time the end of a perturbation. As an entire highway network is not fully equipped with 49 cameras or loop detectors, FCD have the potential to help better detect the end of a moving 50 bottleneck. This specific zone represents a significant road safety risk and there is a need for 51 better real-time detection of the end of congestion. To address this issue, real world data are 52 analyzed from a French freeway with recurrent congestion patterns. After discussing the quality 53 and precision of FCD, a dynamic spatial segmentation of the network highlights the relevance of 54 this data source from an operational standpoint. Further to the empirical network 55 characterization, a systematic detection algorithm is introduced, which is able to detect the 56 queue-end in real-time with a 500m precision. Based on the assumption of a growing penetration 57 rate of FCD in the coming years, the algorithm uses only FCD with very simple detection rules 58 and few parameters. The method is validated on real congestion cases and the results prove the 59 accuracy of the detection. A discussion about the precision of FCD and recommendations for 60 road operators are introduced.