Urban road congestion may take the form of bottleneck congestion when specific points on the road network face a capacity constraint. Bottlenecks induce queuing and involve the propagation of congestion to nearby areas. Congestion may also take the form of frictional congestion as vehicles slow each other down when they travel next to each other. These two forms of congestion imply different calculations for the social cost of congestion and differ in their policy implications. For instance, bottleneck congestion calls for a narrow focus on bottlenecks with respect to both tolls and capacity expansions whereas frictional congestion calls for varying tolls everywhere in the network and a broader approach to road improvements. In practice, very little is known about the relative importance of bottleneck vs. frictional congestion. We collect granular traffic data from Google Maps for several large Asian cities and develop a new methodology to detect bottlenecks and assess their importance in overall congestion.