Process discovery techniques aim at extracting a process model from an event log designed as a data structure that minimally contains a multi-set of event sequences. This paper studies the implications of applying newly developed methods and techniques in the field of process mining to reallife scenarios. The main problem consists in processing the collected data in order to use it for discovering the model of the observed process. The analysis of this problem revealed three main aspects: the need to properly represent the event data, to associate events to process instances and to determine the highlevel actions corresponding to the observed events. Each of these aspects has been investigated and a system architecture has been proposed based on their solution.