Patent Watch: Tracking People Via Cell without Compromising Privacy

I received an e-mail from Jeremy Wood that shared his patent pending method “for tracking cellphones to generate useful demographically-keyed data on the movement of people, without compromising anyone’s privacy.” The patent application will be public in April, but Wood shared it and some slides with me and is interested in feedback.

Below I’ll do my best to explain the method try to think through the value of the data collected. To me that’s the real make or break of the method: does it create data that will be in demand?

How it Works

The method involves tracking a statistical sample of cellphones with their owners’ consent.  The tracking is done in short bursts, not 24/7. The most interesting part is the “fuzzing up” of data aspect. The location information is geocoded and using detailed maps determined to be in a public place or a private one. (More on those terms in a moment.) If the device is not in a public place, the location of the device is “fuzzed up” to an area (census tract, ZIP Code, etc.).
The collected location information, including where the device “spends the night,” is used to assign it a probable demographic profile.

On the Privacy Front

The key ideas behind the method include:

- anonymous movements in public are not private
- the data are anonymous if they don’t contain any private information (like name, address) and don’t show the device entering locations that are private
- the value of location is highest when people are in public places

Wood is working with a Prof. David Lazer of Harvard’s Kennedy School of Government.  Wood describes Lazer as “a privacy expert who has published research using tracks from cellphone movements.”

More on Blurring

First what are public areas? They include large public landmarks (parks, malls), major roads and railways. Lower volume public places includes local points of interest (restaurants, say) and secondary roads. These might be treated as public or private, depending on the risk of privacy being compromised. Everything else is considered a non-public place and will be blurred in “real time.”

Now, in the non-public areas how will blurring work? In dense residential areas blurring up to a census tract may be “good enough.” In non-residential areas the density of cell phones will be used to determine how much to fuzz the data.

Results

Thus from the data collected, it will be possible to write probable stories for the anonymous individuals’ tracks. Wood shared this graphic of two samples.

Uses

Wood suggests this data may work better than existing data for things like:

- targeting direct mail
- competitive analysis - see from where your competitor’s visits originate
- surrogate for demographic data where none is available (shoppers at high end stores tend to from high end neighborhoods)
- predication of destinations for location-based ads
- shortcuts (might want to advertise here if this is how many people travel)
- indoor tracking (would need Wi-Fi or other locating mechanism)
Risks

Wood identifies several, not the least of which is that the patent may not be granted for one or more reasons. Others include lack of demand for privacy protection, cost to produce the data, carriers may not wish to participate, and a few more.

Thoughts?

This is a quick overview that leaves out many details. Could this work? Would there be a business model? Would you as a cell phone carrier “buy in?” What incentive would be needed? Are other permission-based systems already collecting data like this? Are you a data user that might find such data valuable? Comment here and if you like I can put you in touch with Wood.