The real-time detection of crowded spots in access networks is considered nowadays a necessary step in the evolution of mobile cellular networks as it can be of great benefit for many use-cases. On the one hand, a dynamic positioning of contents and computing resources in the most crowded regions can lower connection latency and data loss and can allow us to have a seamless service provided for the users, without performance degradation across the network. On the other hand, a dynamic resource allocation among access points taking into account their loads can enhance the user's quality of service and indeed network performances. In this context, using real mobile data traces from a cellular network operator in France, provided us with a temporal and spatial analysis of user content consumption habits in different French Metropolitan areas (Paris, Lyon and Nice). Furthermore, we put to use a real-time crowded spot estimator computed using two user mobility metrics, using a linear regression approach. Evaluating our estimator against more than one million user databases from a major French network operator, it appears to be an excellent crowd detection solution of cellular and backhauling network management. We show that its error count definitely decreases with the cell load, and it becomes very small for reasonable crowded spot load reaching S. Hoteit Ecole d'ingénieurs du numérique ISEP 2 Sahar Hoteit et al. upper thresholds. We also show that our crowded spot estimator is time and city-independent as it shows a stable behavior for different times of the day and for different cities with different topographies. Furthermore, compared to another crowded spot estimator from the literature, we show that our proposed estimator offers more suitable and accurate results in terms of crowded spot estimation for the three selected areas.