cairn.com | The Price of Knowing Where Everyone Is

My daughter stopped me before I could apologize. She had spent the afternoon at a polling place helping people register to vote, and I was late to pick her up.

My phone buzzed.

"I know." she texted. "You're at Trader Joe's."

She was right, but I hadn't mentioned Trader Joe's. How did she know?

Then I remembered: months earlier I had agreed to share my location with my family. On her screen, I was a blue dot miles away near a familiar parking lot.

I drove home wondering who else knew where I was.

The question took me back to 2019. For most of a decade I worked for Los Angeles County Metro, and one of my jobs was to study how millions of people moved through the county. My best intentions aside, I was the person on the other end of the blue dot.

The instrument

At Metro we learned to see the city in two ways at once. The first was the farecard, each tap a record of someone's journey. The second was a commercial location dataset of twenty-two million anonymous devices moving through five counties around Los Angeles. Together they provide information about where people go that rider surveys could never provide.

Our queries were practical. Which short trips lose to the car because the walk and the wait eat the trip? Which corridors bus and rail actually win, and by how much against driving? Where the frequent riders live, strung along the rail lines, and how much of each district's travel does Metro capture? And, from the bought map, the big one: how the whole county fills and drains across a weekday.

To answer them, every comparison ran through a routing engine I had helped to build based on OpenTripPlanner (OTP). This is the same trip planner that, on a rider's phone, said Take the westbound 704, arriving in seven minutes. I had spent years teaching a machine to answer one person's question: How do I get there? Now the same machine helped answer a city's: Where is everyone going?

Stay locations and their spatial distribution in the region. CDR data for a 2-month sample period in Boston were used to show the kernel density estimation of the pass-by locations.<br>We didn't build the dataset. We bought it.

There were no names in our data, and we didn't need them. Each rider became an ID, each ID dissolved into aggregates, into flows and pressures and the shape of a morning. That was the point of the work, and, I told myself, its mercy.

We bought four months of commercial location data the way you might buy any product from a vendor. The surprise is not that a transit agency studied movement. Transit is movement. The surprise is that our citizens' movements were for sale at all - that somewhere a company had already gathered the comings and goings of twenty-two million devices, cleaned the data, and offered it as a product.

The farecard data were different. We didn't buy them. Every boarding left a tap. Over four months, those boarding taps accumulated into one hundred thirty-three million records. At the time, though, they revealed only origins. Metro did not begin requiring exit taps until 2025, so in 2018 and 2019 riders' destinations were invisible to us until we bought commercial location data.

Outlines of a life

Neither dataset knew my name. It knew only a persistent pattern of movement. A device sleeps here. It spends its weekdays there. It leaves at the same hour every morning and comes home after dark. It travels beside the same few other devices, day after day. The name is missing, but the outline of a life is already there.

A sequence of phone events becomes a set of places a person is likely to be found.<br>One day of that tells you where someone was. Four months of it tells you who they are.

Time is information.

Home falls out of a simple rule: the place where a device rests on weeknights, between evening and morning. Add a workplace, a gym, a church, the school run, and a person's week resolves out of the noise like a face emerging from static.

The farecard data did the same for riders. Its heuristics identified each card's home and up to three "regular" places it returned to. The authors would not call any of those places work. A regular stop could be a job or it could be a campus, and you cannot tell a commuter from a student by the shape of a week. That care matters, and I am glad we took it. It does not change the larger picture. Careful or careless, the data still drew the outline of a life.

A week of the city's phone activity: the weekday rhythm, peaking in the late afternoon.<br>The patterns held tight because people do. The top five percent of riders accounted for more than sixty percent of the trips. We are creatures of route. Most of what a system can know about you, you repeat.

A life in motion singles you out faster than a fingerprint

How anonymous is anonymous? Less than the word promises.

Those simple rules that turn a phone's movements into likely places - home, work, the grocery store, the school run - didn't come...