Geographical features on bus routes impact a bus’s performance, and as a consequence affect human mobility through cities. Analysis of these geographical features is non-trivial because they often must be manually recorded, limiting the ability to extract these features on a large scale. This paper proposes a novel method of extracting features from crowd-sourced OpenStreetMap (OSM) data and compares this method to the ground truth data for 539 stop pair segments in Dublin, Ireland. This paper also proposes algorithms to detect turns and the direction taken by buses at roundabouts, using the angle between points on the segment lines. Statistical analysis was performed, and elastic net linear regression models were developed with a subset of the route features to show their effect. The results show over 97% accurate identification of most individual features using the novel technique, with most errors resulting from OSM quality issues. The features that most negatively affected the average speed and reliability of the bus with statistical significance (p < 0.025) were: retail land use, turns, traffic lights, and roundabouts. The average speed limit and the length of the segment had a positive impact on the average speed but not on the reliability. This method can be used with any bus performance metric to obtain a deeper understanding of the dynamics of bus travel, provide detailed information for bus travel time simulations and more accurately predict bus journey times to improve scheduling on the overall bus network.