What is a device graph, and how does it work?
The term "device graph" refers to a map of the connections between users and their multiple devices. For example, consider a hypothetical user or customer who owns a personal computer, a mobile phone and a tablet.
Before device graph technology became common, the analytics data for that user's behavior would be siloed, rolled up into a view of "mobile" performance, "desktop" performance, etc.—even if the same user was interacting with your ads on all of their devices, their data would be split up into different contexts.
But with the device graphs frequently used in internet advertising today, those numbers can be associated back to one user. This is what powers "cross-device" advertising technology, and part of what powers our attribution, which can associate an ad impression on one device (often mobile, but not always) to the activity we receive from your pixel integration.
Some companies (like the larger social media networks and ISPs) build their own based on a deterministic model—they can easily determine which device matches which user based on that person using the same social media account on multiple devices. Since this is valuable context to have for advertising purposes, these companies generally don't share this data set with other advertising or analytics providers, so multiple third-party device graph providers have sprung up to serve the broader market. These providers most often use what's called a probabilistic model, which uses machine learning to observe patterns in user browsing and behavior data to calculate which users are associated with which devices. They are able to achieve this through their own data collection, as well as partnerships with other data providers. We leverage one of these third-party device graph provider to increase the accuracy and cross-device scale of our SmartAds measurement.