The explosive growth of connected objects is certainly one of the most important challenges facing operators' network infrastructures. Although it has been foreseen for a very long time, it is still not clear how to support such huge number of devices efficiently. A smarter planning of dedicated access slots would certainly limit the burden but remains insufficient since some equipments react to events which cannot be timed. Moreover, barring some IoT devices from accessing the network is very efficient ; nevertheless, efficiency is generally linked to precise knowledge of the number of contending devices. Though, before connection establishment, the terminals are invisible to access points and, therefore, it is very difficult to estimate their number. A lower bound of backlogged devices can be determined. However, underestimating this number may lead to a congestion collapse whereas an overestimation implies underutilization of resources. In this paper, we propose a lightweight change to the standard to accurately reveal the state of network congestion by overloading connec-tions' requests with the number of access attempts (number of times the device has been barred as well as number of attempts). Using such information, we propose an accurate recursive estimator of the number of devices. The obtained results demonstrated that the proposed solution not only makes it possible to estimate the number of equipments much better than existing techniques, but also allows determining precisely the number of blocked equipments.