Use of smartphone apps to detect driving risks and/or provide warning information to drivers is a hot topic in traffic safety research. However, scientific evidence on the effectiveness of such apps to reduce dangerous driving is still limited. To provide additional insights, this study examines whether and how driving risks diagnoses and real-time warning information provided through a smartphone app could assist drivers to control their vehicles on expressways safely. For this research purpose, a 3-month driving experiment was conducted in Japan, where the different combination and utilization of app functions were investigated. During the experiment, a series of questionnaire surveys were also carried out in response to six different app versions. A bivariate ordered probit model was estimated to jointly represent speed limit compliance and acceleration/deceleration simultaneously. Model estimation results show that with the warning information provided via the smartphone app, it is more effective to improve drivers’ performance on speed limit compliance than on acceleration/deceleration control. The app improves speed limit compliance and acceleration/deceleration control of careless drivers but only improves speed limit compliance of irritable and excessively confident drivers and acceleration/deceleration control of aggressive drivers. Finally, implications of analysis results to traffic safety measures are discussed.