When you use mapping software for directions, the app typically gives you two trip times: how long it would take to reach your destination driving the speed limit or a couple of miles per hour faster (a traffic-free utopia, of sorts) and the time it would take in current traffic. But wait a minute — how does a computer, which has never been stuck behind a jackknifed semi on the Beltway, know what traffic is like?
The data source is rather simple: you, and others like you. For example, if you have a GPS-equipped smartphone and enable the “my location” function on Google Maps, the company can monitor how fast you’re traveling. (The data are decoupled from the users’ identity so that no one is tracking your movements, according to Google.)
When your smartphone feeds information into the system, Google adjusts its understanding of how long it takes to travel that stretch of road at that time, on that day of the week, at that time of year, under those weather conditions. (Google uses not only its own data but also information from partner traffic compilers such as Inrix to develop its historical data.)
Historical data are important because it’s hard to acquire enough real-time information to make traffic forecasts. Think of the complications involved. There aren’t that many people with smartphones driving any given block at a given moment, and not all of them will have enabled their data-sharing software. In addition, some of the phones will be red herrings with respect to traffic. Some, for example, will be in the pockets of walkers or in the baskets of bicycles. Others might be in a car that has pulled over or is looking for a parking spot.
When you plug a destination into your smartphone, the program’s first estimate of travel time is based on this historical information. Only then does it begin to add in information from users traveling your route right now. Different programs use different methods to combine the data that hundreds or thousands of phones are transmitting. Unfortunately, the algorithms are tightly held secrets. Just as search engine companies won’t talk much about how they analyze a query, traffic mapping companies won’t say how they integrate real-time traffic data into their software.
The companies are willing to reveal a few of their tricks. When you enter a destination in Google Maps, for example, the app relies more heavily on up-to-date traffic information for the segments that are close to your current location than it does for the distant parts of the trip, since current traffic conditions might not mean much in an hour or two. (Google recently announced that its software would offer en route alerts that recommend a different path when traffic conditions change; this change helps integrate real-time data with information about the more distant parts of a route.)
Waze has some additional features. (Waze was purchased by Google last June but continues to operate an independent app.) Through the app, users can inform other Waze users whenever they see something notable — debris in the right lane, a crash slowing traffic, etc. — on a stretch of road. When other users travel the same road segment, the app asks them to confirm the earlier report. (Waze counts around 300,000 active users in the Washington area.) The whole thing may seem a bit old school — like calling in a car accident to the local television station — but it improves traffic forecasts in two ways.
Active reports enable Waze to add a third layer to its calculations. It starts with historical traffic information, adds in the current speed of users, then uses incident reports to forecast how traffic might change in the near future. More weight is given to incidents that have been confirmed by multiple users. In addition, active-user participation improves the algorithm that estimates current traffic conditions. The system assumes that someone reporting traffic incidents is actually traveling, rather than, say, sitting at a gas station. When running the algorithm that estimates current traffic speeds, it takes these users’ data more seriously than those of people who are passively using the app. (Waze’s active-reports feature raises concerns about distracted driving, but the company insists that reporting traffic incidents is more akin to operating the radio than composing a text message.)
Traffic mapping has made substantial progress in recent years, moving from one guy in a helicopter to a crowdsourcing network, but no company has entirely figured out a way to get precise or real-time information to frustrated drivers sitting in traffic. So for now, be patient and keep your eyes on the road.