For cities trying to understand how to get more people to walk and bike, one place to start is data about existing pedestrians and cyclists—but the traditional tools to figure out how many people are biking on a particular route, like sensors embedded in roads, are both expensive and time-consuming to use.
A new tool from the mobility analytics company StreetLight Data maps bike and pedestrian data in much more detail, using machine learning to analyze location data from mobile phones and understand how many people are walking, biking, or driving on any particular street throughout the day.
“From a city’s perspective—particularly as you get into large cities and the recent press around how prone they are to accidents—it’s very difficult to connect the dots between where the bike traffic is and where the bike accidents are happening,” says Martin Morzynski, VP of marketing and product management at StreetLight Data. Most cities have few, if any, bike sensors. San Francisco, which has the most extensive network of sensors for bikes in the U.S., has only 75 installed in a city with more than 1,000 miles of streets and where around a million residents and commuters move around each day. When the city bought three new sensors in 2015, it cost $187,000 to buy, plan for, install, and maintain them, and took two years of lead time. Other cities rely on having people at a street corner manually counting commuters, a method that misses larger patterns over time and across the city as a whole.
StreetLight Data’s software, which it had already been using to analyze car traffic, pulls anonymized location data from apps on mobile phones, using machine learning to analyze the speed of travel, distance, and other factors to know if someone is walking down a sidewalk, biking, or driving. By looking at larger patterns, cities can begin to understand where accident rates are highest, or where it might make sense to install new bike lanes. If cities see that a lot of cars are making short trips on a particular route—over distances that someone could easily walk or bike—that could be another place to prioritize building new infrastructure. Data about when people bike most can also be used to time traffic signals differently.
“The ability of cities to be able to understand the movement of people and bikes is really impaired,” says Morzynski. “Not until you get into location-based services and GPS data like ours do you actually get to have a comprehensive perspective on literally every street across the city.”