Cross-device tracking, explained

Those are the buzziest of buzzwords, so let’s make sense of them.

First, what is cross-device tracking?
It’s a catch-all for the multiple methods for tracking and identifying Web users across devices. Figuring out that a smartphone user and desktop user are the same person makes it easier to not only target them with ads but also to track whether a given ad seen on the desktop resulted in a purchase on a mobile device — or vice versa. It’s a hard task.

“Deterministic versus probabilistic” sounds like a graduate thesis, though.
Deterministic tracking is the more simple of the two techniques because it relies on logged-in user data to identify people across devices. Logging into Facebook or Google on your desktop and phone sends a pretty clear signal on both devices that you are the same person. Any platform, or even publisher, that collects email addresses or asks users to sign in on their devices can use deterministic linking to target ads.

That seems like it would be pretty effective. Why would you use anything else?
While deterministic tracking is effective, it isn’t without its drawbacks. The problem is that the technique is exclusively within the domain of the players who have the data — — i.e., the big tech companies. This creates so-called “walled gardens,” which are great for those that own them but less great for those outside of them.

“If you look at the vast number of brands, agencies, marketing cloud companies, marketing tech and ad tech companies that touch a consumer on their digital path, each and every one of them need cross-device linking,” said Tapad CEO Are Traasdahl. “An open marketplace is a far more viable solution, and it is necessary for a healthy ecosystem overall.”

OK, so deterministic tracking is effective but limited in scope. What’s the deal with probabilistic?
Increasingly, tech companies are putting their bets behind “probabilistic tracking,” a far more complex method that uses a litany of algorithmic acrobatics to track users across devices. As its name suggests, probabilistic tracking leans heavily on probability: Companies such as Tapad and Drawbridge collect billions of data points, including IP addresses, browsing patterns and even device proximity — and then run predictive tests on them to figure out whether the devices belong to the same person. This method has a 97.3 percent chance of accurately identifying users across devices.

This sounds like a privacy nightmare.
You’re not alone in that impression. The Federal Trade Commission has had similar concerns, which it plans to address at a public hearing in November. Its main question: Should this whole process be more transparent to Web users?

Please excuse me while I go clear my browsing history.
Right behind you.