Leveraging Open-Source Points of Interest Data: Insights from Large-Scale Travel Surveys

Transportation research and travel demand modeling now heavily rely on the precision and comprehensiveness of Points of Interest (POI) data. This study evaluates POI data quality through analysis of a random sample of 1,034 locations from five Quebec travel surveys that were manually validated, comparing data from Google Places, OpenStreetMap (OSM), Overture Places, and the Property Assessment Roll against validated precise entrance locations. Results show that while Google Places provides the highest coverage (99.2%), followed by the Property Assessment Roll (92.1%), both OSM and Overture demonstrate similar coverage levels (72.8% and 72.4% respectively) with a median distance error between declared and validated locations between 20 and 50m. Analysis of user interactions reveals that device type has minimal impact on accuracy, while zoom level significantly influences precision, with median distance errors increasing from 20 meters at zoom level 18 to over 140 meters at zoom level 14. Map clicks and marker drags show similar accuracy patterns.