User localization is one of the key service-enablers in broadband mobile communications. Moreover, from a different point of view, next steps towards automatic network optimization also depend upon the capability of the system to perform real-time user localization, in order to obtain the traffic distribution. The aim of this paper is to get deeper into the feasibility and accuracy of different localization mechanisms ranging from triangulation to database correlation. Call tracing data extracted from a real operating mobile network have been used to assess these algorithms after the execution of an extensive measurements campaign. Results show that enhanced triangulation offers the best performance even outperforming other more sophisticated mechanisms like fingerprinting, without introducing any change in the network and without requiring any special characteristic of the user equipment. Indeed, the lack of precision of channel estimates, which for the same position could differ up to 10 dB, introduces a large uncertainty that harms localization mechanisms based on database correlation. Finally, this paper identifies the areas for improvement in triangulation to reach its maximum potential, provides details for its implementation and analyzes the performance of the different proposed enhancements.