A method, in a base station, provide a mechanism for observing random access performance in a Long Term Evolution network. The random access performance is defined by an access delay. The access delay is a delay from the initial random access attempt until access is granted. The method comprises receiving a respective report from a plurality of user equipment, where each of the respective user equipment has performed a random access process and received a random access grant. The reactive report includes access delay related random access information. The method includes determining a respective access delay, based on each of the access delay related random access information. The determined access delays are used for the estimation of access delay statistics. The random access performance is observed based on the access delay statistics.