IEEE 802.11p/DSRC (Dedicated Short Range Communication) is considered to be a promising wireless communication standard for enhancing transportation safety and efficiency. However, IEEE 802.11p-based Vehicle-to-Vehicle (V2V) communication is still unreliable because of the complicating factors of high vehicle speed and complex radio environments. In this paper, we performed a data-based evaluation of V2V communication reliability, using real-world measurements in a typical urban expressway in Beijing. With respect to the characteristics of the urban expressway and our experimental data, we found road slope and traffic density to be the major environmental factors having a significant impact on the V2V communication's Line-Of-Sight (LOS) conditions. On the basis of these two factors, we propose a fuzzy classification method for the LOS conditions, and separate the realtime communication environments into different LOS cases. For each LOS case, we quantify the metrics as received signal strength indication, packet delivery rate, and communication latency. The results reveal that the communication reliability in urban expressways is very unstable because of the changing LOS conditions. This study provides a useful reference for the IEEE 802.11p-based cooperative systems in urban expressways.