Based on the data from two double-deck rail car egress experiments, presented project aims to estimate the effects of various parameters on the evacuation time. This paper focuses on a microscopic analysis of the time headways showing that several headway-based statistics may illustrate the macroscopic behavior as well as may explain observed phenomena. Applying this approach, we have concluded that the evacuation movement may be decomposed into two phases with different performance and that the efficiency of evacuation is given by their proportion. Moreover, we have observed that the total evacuation time is affected mainly by the frequency of extremely large headways, i.e. the performance significantly depends on few slowest persons.