Research Reports

Document Type


Publication Date



travel behavior, travel survey, smartphone, mobile device, multimodal, data

Digital Object Identifier (DOI)


This project evaluated the accuracy and precision of multimodal trip origins and destinations collected using OneBusAway (OBA), an open- source mobile transit app. Accuracy and precision were determined by comparing OBA-collected data with ground truth data manually collected in College Station, Texas, and Tampa, Florida. Researchers made 286 trips (1,568 activity observations) across all devices. The median values for distance and time error for all modes across all devices for detected trips were 47 meters and 59 seconds, respectively. The distance error was larger at the origin than at the destination primarily because the user was in motion during the time delay between the start of a new activity and the device acquiring a location. The mode with least median distance error (32 meters) was walking, while bus travel had the largest (110 meters). The average false negative rate (OBA not detecting a trip) was 29%. It is likely that Tampa-based trips in urban canyons with less dwell time between trips resulted in higher false negative rates. In general, the Samsung devices were noisier and less accurate when detecting activity transitions when compared to the Google Pixel devices. However, across all device models, the median values of time and distance error were relatively similar (25-75 meters, 31-82 seconds). Finally, the accuracy of trip mode classification was very high (99%-100%) except for an outlier (94%) for bicycle in College Station. Researchers also conducted stakeholder interviews and found that multimodal trip data (as collected by OBA) are considered important and useful although additional processing is needed to transform the raw data. The data were identified to be best suited for trip counts for origin-destination and mode detection analyses and meet the time and distance accuracy and precision requirements stated by the stakeholders for these categories. Generally, all stakeholders welcomed the addition of more data to supplement existing datasets.

Policy Brief.pdf (250 kB)
Policy Brief