To provide a solution for efficient and large-scale characterization of push-to-talk service performance in a mobile communication environment there is described a method and apparatus for passive evaluation of push-to-talk traffic in a communication network. After capturing (S10) data transactions in the communication network there follows a step of extracting (S12) interpretations of data transactions being relevant for push-to-talk, traffic classes, and users, from captured data transactions. Then a correlation of extracted interpretations of data transactions being relevant for push-to-talk, traffic classes, and users is stored (S14) in a traffic data base. This allows to calculate (S16) at least one performance indicator with respect to at least one selected push-to-talk session as a function of information stored in the traffic database.