How many marchers made up the Million Man March? We will never know for sure. Controversy swirled in the wake of the 1995 event, with estimates based on photos and video of the crowd ranging from 400,000 to 2 million people. Getting accurate head counts for massive events, be they peaceful protests or destructive riots, isn’t much easier today. The number of people protesting during the Arab Spring in 2011 on the streets in Yemen (above), for example, is impossible to know. It doesn’t matter how powerful the cameras are in satellites or drones overhead if it is dark or smoky or if the crowds are moving in and out of cover. Rather than visually counting heads, it might be possible to count the streams of digital data flowing from the mobile phones in everyone’s pocket: phone calls, tweets, and Web browsing, all of which can be geolocated. The problem is that no one knows how reliable those streams are for estimating the head count. When you have a million people in one place, what proportion can be expected to use their phones in various ways over a given time period? Now, researchers have made a stab at that calibration. Using crowds of known sizes—at an airport and a stadium in Milan, Italy—a study published in Royal Society Open Science reports that mobile phone data predicted the actual number of people on the ground with an average error rate of just 13%. It wouldn’t have worked on the Million Man March, when the Internet was a novelty and smart phones just a dream, but we’ll be ready for the next one.