Many different leak detection methods exist to detect leaks in pipelines within short time and to receive a precise localization of the leak. Internal model based systems represent a particular group of these systems and have the aim to optimize the sensitivity, accuracy, robustness and reliability for desirable results. The capabilities of the leak detection system and its integrated modules are in many cases only partially known. Furthermore each application has its own characteristics, such as transported liquid or gas, length and run of the pipeline as well as different operating conditions, which could influence the leak detection systems performance. In addition, the DCS / SCADA and the measurement instrumentation have to be considered, which determine the signal quality and data refresh rate. Finally different regulations and standards such as TRFL, API and CSA demand further requirements regarding leak detection. The paper shows how optimal results can be achieved by using model based leak detection systems adapted to the individual application conditions. Using various practical examples showing heavily varying preconditions, possible capabilities of the leak detection system will be presented. On top of that transient conditions will be taken into consideration and it will be illustrated how this can be compensated to a large extent by implementing statistical leak classification into a model based leak detection system.