An experimental campaign was carried out to evaluate the influence of CNG and gasoline on the exhaust emissions and fuel consumption of a bi-fuel passenger car over on-road tests performed in the city of Naples. The chosen route is very traffic congested during the daytime of experimental measurements. An on-board analyzer was used to measure CO, CO2, NOx tailpipe concentrations and the exhaust flow rate. Throughout a carbon balance on the exhaust pollutants, the fuel consumption was estimated. The exact spatial position was acquired by a GPS which allowed to calculate vehicle speed and the traffic condition was monitored by a video camera. Whole trip realized by the vehicle was subdivided in succession of kinematic sequences and the vehicle emissions and fuel consumption were analyzed and presented as value on each kinematic sequence. Moreover, throughout a multivariate statistical analysis of sequences, the driving cycles characterizing the use of vehicle were identified.Finally, comparison between regulated emissions of CNG and gasoline configurations was performed qualitatively by the analysis of speed and emission profiles belonging to the same cluster of cycles. Frequency distribution of mean values of CO2, CO, NOx emissions, and fuel consumptions, respectively with CNG and gasoline fuel type are presented. They are very well differentiated, both in the range of values than in the mode of frequency distribution. Particularly CO2 emissions relative to gasoline fuel show a range of higher values respect on CNG fuel. Related to CO emissions, values obtained with CNG fuel also result much low. Concerning NO emissions, the richer combustion occurring when CNG is fuelled, due to different ECU tuning respect to gasoline, discourages the production of NOx.Moreover it was realized a statistical distribution analysis evaluating emission and fuel consumption mean values in each cluster. This analysis allows to compare emission and fuel consumption distribution between two different fuel types in the same traffic condition. To better underline difference due to fuel type, some statistical moments such as mean, standard deviation and range of regulated emissions are presented, cluster by cluster, to better evaluate the dispersion for each subgroup. The reducing effect computed in pair on each subgroup is clearly evident and repeatable.