The ever increasing complexity of commercial mobile networks drives the need for methods capable of reducing human workload required for network troubleshooting. In order to address this issue, several attempts have been made to develop automatic anomaly detection and diagnosis frameworks for mobile network operators. In this paper, the latest improvements introduced to one of those frameworks are discussed, including more sophisticated profiling and detection capabilities. The new algorithms further reduce the need for human intervention related to the proper configuration of the profiling and anomaly detection apparatus. The main concepts of the new approach are described and illustrated with an explanatory showcase featuring performance data from a live 3 G network.