Car sharing is a new mode of transportation that is gaining increasing popularity with its promise to reduce traffic congestion, parking demands and pollution in our cities. Despite this potential, the properties of car sharing systems, e.g., in terms of spatiotemporal characterisation of how customers use the service, remain largely unexplored in the research literature. In order to fill this gap, in this work we analyse one month of online car-sharing map data from a large station-based carsharing operator in France, which has 960 stations and more than 2700 electric cars. First, we study the spatial and temporal patterns of station utilisation, uncovering a dichotomy in station usage (stations that attract cars mostly in the morning vs. stations attracting cars mostly in the evening). We also find that this dichotomy is linked to the destination (residential or business) of the zone in which the station is located. In addition, we statistically model the users' demand in terms of drop-off and pickup rates, and the parking times of vehicles. Finally, we propose a classifier that exploits simple average statistics (average pickup rate and car availability of a station) in order to understand whether the station is profitable or not for the operator.