With the proliferation of mobile devices and the increasingly complicated nature of the consumer journey, a growing proportion of marketers have adopted cross-device tactics to ensure their digital marketing is effectively reaching consumers. According to Winterberry and the IAB, last year over 50% of marketers said that cross-channel measurement and attribution was occupying the most time and resources. This sentiment has increased this year.
While marketers don’t need to be convinced of the importance of cross-device targeting, there is confusion around the qualities that marketers should be seeking from the graphs that empower and enable their cross-device efforts. An Econsultancy study from March 2016 found that less than half of marketers in North America reported the ability to understand cross-device behavior, understand cross-channel purchase paths and tailor creative to meet those behaviors.
Accordingly, there are a number of questions that marketers should be asking about their device graphs to ensure they’re being used to their full potential.
What type of data is used for building the device graph?
Most device graphs are comprised of both deterministic and probabilistic data inputs. In fact, most device graph providers have a limited deterministic truth-set footprint that they combine with a probabilistic model to maximize scale. Graph customers should ask their providers about the breakdown between deterministic/probabilistic and the composition of the various inputs. While more deterministic is better than less, not all deterministic data is equally valuable and some seemingly deterministic data may be deterministic but inaccurate.
How does the device graph account for cookie challenges on both desktop and mobile?
Also, even if a device graph appears to have a sufficient amount of cookies, it is important to understand how many of these are “Phantom” cookies that have already been deleted. Some device graph providers claim to have billions of cookies included in their graph, but a closer look reveals that a significant number of the cookies could in fact have long been deleted by the user or by software installed by the user.
How does your device graph account for fraudulent mobile ID’s?
Ensuring that a methodology is in place to prevent the use of inaccurate and/or fraudulent MAIDs in the device graph is essential for ensuring quality and accuracy. While fraud is a problem that we often hear associated with site traffic and ad serving, it also impacts mobile advertising IDs, which can be spoofed to create fake users. Ensuring that your device graph vendor has methods in place to prevent incorporation of these fraudulent and/or inaccurate IDs will go a long way.
How accurate is the data provided in the graph?
The accuracy of data within a device graph is not confined to preventing fraud. There are many variables that will impact the accuracy. One is how well the device graph can account for new devices since a user won’t have the same phone, tablet or computer for more than a few years. Another is how the device graph accounts for multiple users on the same WiFi. For example, when users are in an office environment sharing the same WiFi, there could be a match between coworker A’s browsers to co-worker B’s mobile ID.
Because new inputs enter into a device graph all the time, it is important for a device graph provider to test itself constantly—- 24 hours a day, 7 days a week.